Cost of Not Having FAIR Research Data

Cost of not having FAIR
research data
Cost-Benefit analysis for FAIR research data
Written by PwC EU Services

March – 2018
Cost-Benefit analysis for FAIR research data - Cost of not having FAIR research data
European Commission
Directorate-General for Research and Innovation
Directorate A — Policy Development and Coordination
Unit A.2 — Open Data Policy and Science Cloud
Contact Athanasios Karalopoulos
E-mail Athanasios.Karalopoulos@ec.europa.eu
RTD-PUBLICATIONS@ec.europa.eu
European Commission
B-1049 Brussels
Manuscript completed in March 2018.

The information and views set out in this report are those of the author(s) and do not necessarily reflect the official opinion of
the Commission. The Commission does not guarantee the accuracy of the data included in this study. Neither the Commission
nor any person acting on the Commission’s behalf may be held responsible for the use which may be made of the information
contained therein.
More information on the European Union is available on the internet (http://europa.eu).
Luxembourg: Publications Office of the European Union, 2018
PDF ISBN978-92-79-98886-8 doi: 10.2777/02999 KI-02-19-023-EN-N
© European Union, 2018.

Reuse is authorised provided the source is acknowledged. The reuse policy of European Commission documents is regulated by
Decision 2011/833/EU (OJ L 330, 14.12.2011, p. 39).
EUROPEAN COMMISSION
Cost of not having FAIR

research data
Cost-Benefit analysis for FAIR research data
Written by PwC EU Services
Directorate-General for Research and Innovation

Open Science
Contents
1. INTRODUCTION........................................................................................... 5
1.1. Context 5
1.1.1. The value of FAIR 5
1.1.2. On-going FAIR initiatives 6
1.1.3. Challenges for the implementation of the FAIR principles 6
1.2. Intended audience 7
1.3. FAIR versus open research data 8
1.4. Structure of the report 8
2. APPROACH .................................................................................................. 9
2.1. Overview 9
2.2. Indicators identification 9
2.2.1. Impact on research activities 10
2.2.2. Impact on collaboration 14
2.2.3. Impact on innovation 15
2.3. Selected indicators 16
2.4. Indicators quantification 17
2.4.1. Indicator #1: Time spent 18
2.4.2. Indicator #2: Cost of storage 20
2.4.3. Indicator #3: Licence costs 22
2.4.4. Indicator #4: Research retraction 23
2.4.5. Indicator #5: Double funding 23
2.4.6. Indicator #6: Interdisciplinarity 23
2.4.7. Indicator #7: Potential economic growth 24
3. COST CALCULATION .................................................................................... 26
3.1. Indicator #1: Time spent 26
3.2. Indicator #2: Cost of storage 27
3.3. Indicator #3: Licence costs 27
3.4. Indicator #4: Research retraction 28
3.5. Indicator #5: Double funding 29
3.6. Indicator #6: Interdisciplinarity 29
3.7. Indicator #7: Potential economic growth 30
4. CONCLUSION .............................................................................................. 31
BIBLIOGRAPHY ................................................................................................. 33
ANNEX I FAIR PRINCIPLES ............................................................................... 36
2
List of figures
Figure 1: Four foundational characteristics of FAIR .................................................... 5
Figure 2: High-level representation of the methodology followed ................................. 9
Figure 3: Research activities ..................................................................................11
Figure 4: Distribution of volume per number of repositories .......................................21
Figure 5: Cost breakdown ......................................................................................26
List of tables
Table 1: Selected indicators per area ......................................................................16
Table 2: Indicator #1 calculation ............................................................................27
Table 7: FAIR Principles and facets against a (meta)dataset ......................................36
3
EXECUTIVE SUMMARY
Technological advancements have made research and science more data intensive and
interconnected, with researchers producing and sharing increasing volumes of data. In
their effort to produce high quality data, researchers have to follow good data
management and data stewardship practices such as the FAIR principles.
However, there has been no thorough analysis to determine the value of not having FAIR
research data, within and across scientific disciplines, both in economic and non-
economic terms, and to contrast it against the current situation where a majority of
research data is not adhering to the FAIR principles.This report aims to fill these gaps by
estimaing the cost of not having FAIR research data for the EU data market and EU data
economy.
Our analysis relied on available studies that have focused on the quantitative value of
research data that is findable, accessible, interoperable and reusable.
By looking at the impact of FAIR on research activities, collaboration and innovation,
indicators were identified, defined and then quantified. Seven indicators were defined to
estimate the cost of not having FAIR research data: Time spent, cost of storage, licence
costs, research retraction, double funding, interdisciplinarity and potential economic
growth.
To estimate the first five indicators, we first assessed the inefficiencies arising in research
activities due to the absence of FAIR data. From the different levels of inefficiency, we
computed the time wasted due to no having FAIR and the associated costs. Secondly, we
estimated the cost of extra licences that researchers have to pay to access data that
would otherwise be open with the FAIR principles. Thirdly, we looked at the additional
storage costs linked to the absence of FAIR data. Unaccessible data leads to the creation
of additional copies of the data (e.g. by journals or partner universities) which would
otherwise not be required if the FAIR principles were in place. With insufficient data to
estimate the last two indicators, we provided mostly qualitative considerations and
findings instead.
Following this approach, we found that the annual cost of not having FAIR research data
costs the European economy at least €10.2bn every year. In addition, we also listed a
number of consequences from not having FAIR which could not be reliably estimated,
such as an impact on research quality, economic turnover, or machine readability of
research data. By drawing a rough parallel with the European open data economy, we
concluded that these unquantified elements could account for another €16bn annually on
top of what we estimated. These results relied on a combination of desk research,
interviews with the subject matter experts and our most conservative assumptions.
Moreover, while building on top of other available studies and being heavily reliant on
existing material, we have come to realise ourselves how important is to have FAIR
research data. Not only the time invested in this study could have been reduced by a
significant amount, but the content could have been enhanced if more material had been
accessible and reusable.
Finally, by estimating the qualitative and quantitative costs of not having FAIR data, this
report will enable decision makers to make evidence based decisions about efficient ways
to support the real-life implementation of the FAIR data principles. Researchers and
research institutions will now be able to weight the cost of not having FAIR versus the
cost of implementing the FAIR principles.
4
1. INTRODUCTION
1.1. Context
Technological advancements have made research and science more data intensive and
interconnected, with researchers producing and sharing increasing volumes of data. In
their effort to produce high quality data, researchers have to follow good data
management and data stewardship practices. Beyond proper collection, annotation and
archival, good data management and stewardship include the notion of long-term care of
valuable digital assets, either alone or in combination with newly generated data. Good
data management and stewardship is not a goal in itself but rather is a key conduit
leading to easier and simpler data and knowledge discovery and evaluation, and to
subsequent data and knowledge integration and reuse in downstream studies.
To maximise the value of science, research data should have four foundational
characteristics 1; they should be:
Figure 1: Four foundational characteristics of FAIR
Findability, Accessibility, Interoperability and Reusability – the FAIR principles – intend to

define a minimal set of related but independent and separable guiding principles and
practices, which enable both machines and humans to find, access, interoperate and re-
use research data and metadata.
1.1.1. The value of FAIR
The implementation of the FAIR principles can bring direct and indirect benefits to
research stakeholders, from funders to researchers, and can have a positive impact on
the quality and the ROI of research itself. Studies supporting the implementation of the
FAIR principles talk about their positive impact on:
• Reducing duplication in research, in terms of time, effort and funding;
• Rigorous management and stewardship of digital resource helping researchers
adhere to the expectations and requirements of their funding agencies 2;
• Scaling up research findings based on integrated and analysed existing data from
multiple disciplines and regions 3;
• Enabling research to focus more on adding value activities such as interpreting the
data rather than on searching, collecting or re-creating existing data 4; and
1 FAIR Principles described by GO-FAIR, https://www.go-fair.org/fair-principles/; FAIR Principles described by

Force 11, https://www.force11.org/group/fairgroup/fairprinciples; and (Wilkison, Dumontier, & Mons, 2016).
2 (Wilkison, Dumontier, & Mons, 2016)
3 (Bonino da Silva Santos, et al., 2016)
4 http://visit.crowdflower.com/rs/416-ZBE-142/images/CrowdFlower_DataScienceReport_2016.pdf
5
• Enhance the science infrastructure to support knowledge discovery and
innovation.
1.1.2. On-going FAIR initiatives
Multiple initiatives are currently working towards the implementation of the FAIR
principles. We are listing the main ones below:
• The GO FAIR movement 5 which federates existing national initiatives that
committed to the implementation of the FAIR principles. The GO FAIR movement
makes collective choices for the implementation in terms of standards, good
practices and protocols and proposes a technical infrastructure, change
management and training for data stewards.
• Having recognised the importance of a FAIR-enabled data ecosystem for the
implementation of the European Open Science Cloud 6, the European Commission -
together with stakeholders and assisted by the FAIR Data Expert Group 7 - is
working towards an action plan which aims to make research data FAIR and allow
researchers to be able to do cross-disciplinary scientific data sharing and reuse.
This action plan (aka the FAIR data action plan) covers all types of digital research
objects and it is a collaborative instrument guiding the integration of the FAIR
principles at European level, across borders and/or disciplines. In addition, the
FAIR data action plan creates the conditions for the development of coherent
national and/or discipline specific plans for the implementation of the FAIR
principles at national and/or discipline level.
• The study on the possible implementation of the FAIR principles in Denmark,
which was commissioned by the Danish Agency for Science and Higher Education.
This study offers a cost-benefit analysis for a future implementation of the FAIR
principles for public-funded research in Denmark.
• Numerous universities and research institutes, which have openly embraced the
FAIR movement and are promoting the FAIR principles. Examples include the
Dutch Techcentre for Life Science 8, Leiden University 9, Utrecht University 10, and
Maastricht University 11.
• ELIXIR which is a distributed infrastructure for life science information. ELIXIR 12 is
committed to enable the availability of FAIR research data within the framework of
EOSC 13. EXILIR Nodes together with EMBL-EBI 14, coordinate and integrate
Bioinformatics resources across Member States (i.e. by providing databases,
analysis tools, interoperability services, etc.) with the end goal of making
information freely available to the science community.
1.1.3. Challenges for the implementation of the FAIR principles
The fact that the FAIR principles are not common practice yet is due to numerous
reasons. Some concern the lack of awareness in the research community 15 about how to
share data, in which format, what information or metadata should be provided etc.
Others touch upon existing cultures and behaviours in conducting research, from
research funders forbidding researchers to share their data to researchers not even
considering that the data they produce can be valuable for others, the lack of attention
5 https://www.go-fair.org/
6 European Cloud Initiative - Building a competitive data and knowledge economy in Europe (COM(2016) 178
final)
7 The Commission has established the Expert Group on FAIR data to support the Research and Innovation
policy development on Open Science.
8 https://www.dtls.nl/fair-data/fair-data/
9 https://www.universiteitleiden.nl/en/research-dossiers/data-science/leiden-silicon-valley-of-fair-data
10 https://www.uu.nl/en/research/research-data-management/guides/costs-of-data-management
11 https://www.maastrichtuniversity.nl/research/data-science-um/research
12 ELIXIR Is a distributed infrastructure comprising 180 leading universities and centres of excellence
13 https://www.elixir-europe.org/system/files/elixir_statement_on_fair_data_management.pdf
14 European Molecular Biology Laboratory (EMBL) – European Bioinformatics Institute (EBI)
15 Interview with Barend Mons, 2018-01-17
6
given to the preparation of a data management plan, missing metadata 16, various
competing standards for research data and metadata, and the lack of persistent
identifiers for data, datasets and metadata 17.
Moreover, many researchers and organisations are still reluctant to apply the FAIR
principles and share their datasets because of real or perceived costs, including time
investment and money.
In order to drive the implementation of the FAIR principles in Europe, the European
Commission together with a number of pioneering European research stakeholders is
taking activities and measures that aim at raising awareness about costs and benefits of
FAIR data, and is encouraging funding bodies to set guidelines or support the
development of an infrastructure for publishing FAIR data.
This current study is positioned exactly in this context. Although the discussion is very
often around the costs and benefits of the FAIR principles’ implementation both in
economic and non-economic terms (and even there a global view is still to be
developed), few think about the current costs and opportunity losses from not having the
FAIR principles implemented (or having very small partial implementations in specific
disciplines and countries). Assessing the cost of not having FAIR data will provide
quantified facts to back-up the utility of the FAIR principles and will help convincing
research stakeholders to invest in their implementation.
This study presents and applies a quantitative methodology for estimating the cost of not
having FAIR research data in Europe as well as cost estimations. The methodology is well
documented, easily applicable in different countries and disciplines and hence easily
repeatable in the future.
The scope of this study comprises direct and indirect costs resulting from not
implementing with the FAIR principles in Europe. This includes research carried out by
public, private and non-governmental organisations. For this reason, the report did not
look into the cost to implement FAIR research data in Europe nor the potential loss of
revenue incurred by the researchers as a result of making data free and open.
1.2. Intended audience
This report is intended primarily for:
(1) Research funders: to raise awareness on the cost of not having FAIR
research data, e.g. funding redundant work because earlier research was not
FAIR. With this report, funders can better appraise the importance of FAIR
and include where relevant compliance to the FAIR principles as a condition
for researchers to obtain funding.
(2) Research (data) infrastructures: to raise awareness about the impact of not
having FAIR research data on e.g. the cost of storage, and other activities of
research infrastructures.
(3) Research performing organisations: to raise awareness about the cost of not
having FAIR research data for researchers, e.g. in terms of research quality,
additional time spent on research, and citations. By identifying the cost of
not having FAIR research data, this report helps researchers to justify
investing into compliance with the FAIR principles.
16 (Zahedi, Haustein, & Bowman, 2014), and (Parsons, Grimshaw, & Williamson, 2013)
17 (Johnson, Parsons, Chiarelli, & Kaye, JISC Research Data Assessment Support - Findings of the 2016 data
assessment framework (DAF) surveys, 2016), Stehouwer & Wittenburg, 2014 and Tenopir C. , et al., 2011
7
1.3. FAIR versus open research data
This study adopts the H2020 Programme 18 definition of research data, i.e. information, in
particular facts or numbers, collected to be examined and considered as a basis for
reasoning, discussion, or calculation. The types of data covered by this definition are
first, the underlying data (the data needed to validate the results presented in scientific
publications), including the associated metadata (i.e. metadata describing the research
data deposited) and second any other data (for instance curated data not directly
attributable to a publication, or raw data) including also the associated metadata.
FAIR data and open data are two distinct concepts which are however coming closer and
closer. (Mons, et al., 2017) distinguish the concept of FAIR from the concept of open,
saying: “In the envisioned Internet of FAIR Data and Services, the degree to which any
piece of data is available, or even advertised as being available (via its metadata) is
entirely at the discretion of the data owner.” The reasoning behind is that “FAIR” only
speaks of the need to:
• Describe a machine-readable or human process for accessing discovered data;
• Openly and richly describe the context within which those data were generated, to
enable evaluation of its utility;
• Explicitly define the conditions under which they may be reused; and
• Provide clear instructions on how they should be cited when reused. 19
However, research, data is not truly reusable unless it is open, i.e. available under an
open licence and at marginal costs (in most cases at zero cost), and openness comes
often hand in hand with the implementation of the FAIR principles. As part of the Open
Science Agenda, the European Commission defined the ambition to make FAIR and open
access the default for data sharing in scientific research. Moreover, the Council of the
European Union concluded 20 that “as open as possible, as closed as necessary” is the
underlying principle for optimal reuse of research data.
In this study, we therefore consider the cost of not having FAIR and free open access
research data whenever discussing the FAIR principles.
1.4. Structure of the report
The remainder of this study is structured as follows:
• Chapter 2 “Approach” lays out our approach for identifying indicators to measure
the cost of not having research data, describes each indicator and the
methodology for quantifying them.
• Chapter 3 “Cost calculation” presents the results of the quantification exercise for
the cost of not having FAIR data.
• Chapter 4 “Conclusion” discusses the main findings and observations of the study.
18 (Directorate-General for Research & Innovation (European Commission), 2017)

19 (Data Citation Synthesis Group, 2014)
20 (Council of the European Union, 2016)
8
2. APPROACH
2.1. Overview
Figure 2: High-level representation of the methodology followed
In this study, we started by identifying and selecting economic indicators which drive the
cost of not having FAIR research data. These indicators - linked to the causes of the cost
occurring due to not having FAIR research data - were categorised and quantified to the
extent possible. Furthermore, this approach is based solely on secondary data 21.
In order to quantify the indicators, assumptions and assertions were made to allow us to
estimate the cost of not having FAIR research data. The method for estimating the
indicators and the necessary assumptions are described.
Different criteria were considered for identifying and quantifying the indicators:
• Preciseness: the indicator is related to specific activities, which are defined in

clear terms;
• Reliability: the indicator is based on facts and consistently measurable over
time, which ensure the sustainability of this estimation;
• Measurability: the indicator can be quantified using existing tools and methods;
and
• No overlaps: each indicator can be measured independently from the other
indicators.
2.2. Indicators identification
In order to measure the cost of not complying with the FAIR principles, we first needed
to assess their impact on research data. Therefore, we detailed in Annex I the various
21 Secondary data refers to information that already exist, in our case information that we collected rather than
created.
9
facets of each principle, and reviewed the metrics proposed by recent related works 22 for
estimating the compliance of research data to the FAIR principles. These facets and
metrics are used for identifying the economic indicators linked to non-FAIR research
data. Economic indicators not related to the FAIR principles were discarded.
We projected the impact of the FAIR principles on research data to researchers,
repositories and publishers, and innovation. The identification of potential economic
indicators for the costs of not having FAIR research data also benefited from the desk
research and interviewing experts.
The following three areas regroup the different economic indicators identified as part of
the cost of not having FAIR research data. These areas are applicable to all sectors (i.e.
academic, private, public, non-profit).
• Impact on research activities: This category of indicators relates to activities
taking place during research. It concerns, but is not limited to, research
development, planning of the research project, creation of the data, collection of
the data, pre-processing and cleansing of the data, integration and transformation
of the data, etc. In these activities, time wasted due to research data not being
FAIR is identified and quantified. This category also includes the duplication of
storage for research data that is not FAIR (i.e. research data being stored
unnecessary on different repositories). This duplication is quantified in terms of
storage costs. In addition, we also include in this category the costs faced by
researchers for finding, accessing and analysing research data that is not open.
For instance, time spent to obtain the data (e.g. registration process), associated
fees and licensing costs, etc.
• Impact on collaboration: here is an opportunity cost for the research
community but also individuals and organisations using research data. This cost
manifests itself in the form of missed opportunities for collaboration or cross-
fertilization between research groups within and across disciplines. It is important
to note that the estimation of this cost is very sensitive to the assumptions taken.
• Impact on innovation: This category relates to the impact of research data not
being FAIR on innovation and consequently on the European economy. Ample
empirical evidence demonstrates that research is a “key driver of productivity and
economic growth” 23. Likewise, non-FAIR research data can have an impact on the
number of patents filed, missed opportunities in terms of new business or
products, lower job creation, etc.
In the remainder of this study, we differentiate between academic and non-academic
researchers, the latter encompassing the private, public and non-profit sectors. For
academic staff, the term ‘researcher’ includes all people for whom research is one of the
core activities such as professors, PhD students, post-doctorates, assistant professors
etc. For non-academic staff, the term ‘researcher’ includes all people who are not
academics but qualified as ‘research intensive’ (i.e. whose core activity is research).
2.2.1. Impact on research activities
When analysing existing studies on the cost related to research activities, we identified a
set of activities to which the cost indicators can be linked. These activities have been
compiled from existing literature about data life cycle 24 and research activities 25.
Most of the costs faced by research stakeholders can be linked directly to one or multiple
of the following activities (see also Figure 2):
22 (Dunning, de Smaele, & Böhmer, 2017)

https://via.hypothes.is/https://www.force11.org/group/fairgroup/fairprinciples
http://datafairport.org/fair-principles-living-document-menu
(Wilkison, Dumontier, & Mons, 2016)
https://www.force11.org/group/fairgroup/fairprinciples
https://via.hypothes.is/https:/content.iospress.com/articles/information-services-and-use/isu824#ref019
23 (Directorate-General for Research and Innovation (European Commission), 2017)
24 (Data Life Cycle | DataONE, n.d.)
25 (Ziker, 2013)
10
• Research development
• Planning of a research project
• Creation and collection of the data
• Pre-processing and data cleansing
• Integration of the data
• Analysis of the data
• Redaction of the report
• Registration and publication
• Peer review
The FAIR principles do not have direct impact on all those activities, which is why
research development, planning of the research project and redaction of the report are
left out in our analysis.
Figure 3: Research activities
2.2.1.1. Creation and collection of the data
Creation and collection of the data is the step in the research life cycle where data is
created (i.e. through observations, experiments or simulations), and potentially useful
and existing data is found and obtained. This step also includes the performance of
veracity checks on the data.
In this activity of the research life cycle, not having FAIR has an impact on the time,
storage costs and licence costs required to find, access, manipulate and reuse
(meta)data underpinning research papers and studies which has not been made
(properly) available, for different reasons:
• Time lost due to unfindable, not human understandable or with unstructured or
incomplete metadata. Many researchers do not publish the minimum set of
metadata for their studies, required for other researchers to find, access and
reuse it. Metadata of poor quality hampers findability. Furthermore, as described
in Annex I, the metadata should be rich enough for external readers to directly
understand the necessary information about the study and its supportive data.
11
• The cost of time from the human expertise needed to read and understand the
data. Having the data well documented can result in a much lower overhead in
terms of capturing the essential information inside the dataset and work with it
(i.e. sufficient understanding of the field research).
• The lack of a single point of access 26 is increasing the time required by
researchers to authenticate to different publication channels, to search into each
channel etc. The FAIR movement is advocating a single point of access, where you
could query for metadata which would instantly point to the layer where the data
are located, thus reducing considerably time to access (meta)data. In the event of
unavailable existing data underpinning published findings, time is required to re-
create a dataset.
• Data related to a specific research or a whole research discipline is often stored in
isolation, in information silos. It often occurs that data is partially available from
different points of access 27, causing the researcher to spend more time to retrieve
all the data he/she needs. This researcher may also need to transfer data from
one repository to another which would increase overall storage costs.
• Data is duplicated across one or more repository, which causes researchers to
spend additional time selecting the right datasets.
• Making the research data compliant to the FAIR principles will have a positive
impact on the open access of scientific publications and will consequently reduce
the access fees (i.e. licence costs) faced by researchers which will enhance the
citation ratio (i.e. reuse) of their publications 28.
• Finally, unstructured metadata makes it hardly machine-readable which is one of
the many perspectives advocated by the FAIR principles. Non-machine-readable
and non-interoperable metadata, which cannot be indexed by search engines
hampers greatly the findability of the associated data 29.
Machine analysing data for pattern instead of human capital could reduce the time
spent by researchers and all the costs associated (e.g. licence costs to access
specific software, external expertise, data transfer due to remote access
impossible) to it down to almost zero. The current state of play forces researchers
to read and understand the metadata manually, and hampers automated data
search and discovery.
Moreover, difficulties in assessing data quality and integrity because of un-adequate
(meta)data makes it impossible for researchers to share and analyse research data in a
trusted environment across technologies, disciplines and borders.
2.2.1.2. Pre-processing and data cleansing
Pre-processing and describing the data is the step in the research life cycle where
inspection and checks are performed to improve and ensure data quality. Data modelling
and data cleansing also take place in this step.
Not having FAIR has an impact on the time necessary to create, curate, cleanse,
reprocess and keep the data and metadata up to date. Maintenance of low quality
metadata incurs significant costs and can be a source of errors. This cost increases with
time for various potential reasons: human factors, standards changed etc.
For instance, time is needed to run quality checks (e.g. run scripts/queries) on both
existing and raw data collected. Similarly, time is required to improve the quality of
(meta)data and to transform data when necessary (e.g. to identify and fix problems).
26 E.g. EOSC ambitions to be “a one-stop-shop to find, access, and use research data and services from
multiple disciplines and platforms.”
27 A single point of access should not be confused with a single repository. The first would only harvest and
store the metadata while the second refers to all types of data.
28 (Van Noorden, 2013)
29 (Stehouwer & Wittenburg, 2014)
12
FAIR research data should be machine readable, thus reducing the time required for
checking quality. The underlying reasons are:
• Curation methodologies are often invented in a moment of need and consequently
cannot be replicated for other metadata, which results in inefficiencies in the
future; and
• Lack of documentation for the wide heterogeneity of pre-processing operations
being carried out, depending on the data that has been generated.
2.2.1.3. Integration of the data
Integration of the data is the activity in the research life cycle where data from disparate
sources is aggregated to form one homogeneous dataset that can be analysed. As
research data volume is surging 30, data integration becomes necessary to be able to
aggregate data from multiple sources (and from different formats).
Not having FAIR has an impact on the time required to correlate/transform two or more
datasets in a machine-readable and harmonised format (from proprietary formats to
standardized ones), or time required to understand that two different datasets cannot be
integrated due to non-interoperable vocabularies (i.e. metadata) used to define datasets.
As a rule of thumb, in a data analysis project, data cleansing of poor quality data can
take up to 80% of the total effort.
It should be noted that bridging the gap between semantic classifications and
annotations is a challenge due to the dynamic nature of research, the dynamics of
classifications and unhandy tools 31.
In the private sector, integration costs are particularly high in the case of joint research
ventures, as time to integrate heterogeneous data is considerable 32. This may sometimes
lead to the non-utilization of datasets.
Moreover, experts in some communities developed their own guidelines on metadata.
While guidelines provide a high degree of flexibility, and thus representational capacity, it
first requires experts to make use of them 33 and secondly, the guidelines are often not
aligned with recognized standards, hindering human and machine readability.
2.2.1.4. Analysis of the data
Analysis of the data is the activity in the research life cycle where the data is processed
with the ultimate purpose to extract useful information, to elicit insights and eventually
to formulate observations and conclusions. Analysis of the data or data modelling is often
performed using specific software’s, tools or methodologies.
In this step of the research life cycle, not having FAIR has an impact on the time required
to analyse data that is not properly documented or is not complete and at the right level
of quality. According to a recent paper, “there are growing concerns about replicability
and reproducibility of research results: a recent survey indicated that more than 70% of
researchers have tried and failed to reproduce another scientist’s results, and more than
half agree that there is a significant crisis of reproducibility” (Baker, 2016). It also
pertains to the time required to verify the findings of other studies in order to decide
whether the methodology and the results are correct and could be the object of a
publication.
2.2.1.5. Registration and publication
Registration and publication is the step in the research life cycle where data is accurately
and adequately described using the relevant metadata. The (meta)data is then registered

31 Peter Wittenburg
33 Second year report on RDA Europe analysis programme
13
(i.e. stored) online e.g. on web servers and published using the adequate channels. Not
having FAIR has an impact on both time and storage costs, namely:
• Time to register (meta)data and to maintain a research repository with
(meta)data not FAIR, or to keep a specific database up-to-date with a certain
level of interoperability.
• Storage costs related to the volume of data duplicated in different repositories,
independently from the preservation mechanism used. For example, data
inaccessible or wrongly described could lead to duplications within and between
repositories.
The publication and indexation processes and requirements vary between repositories,
increasing the time spent on those activities.
Without accessible data and publications, journals, universities and funders often each
hold a separate copy of the underlying research data on a repository. This unnecessary
redundancy drives the overall costs of storage up.
2.2.1.6. Peer review
Peer review is the step in the research life cycle where research work is evaluated by
other researchers with relevant expertise or experience before the work can be endorsed,
published or presented.
In this step of the research life cycle, not having FAIR has an impact on the time spent
because of redundant reviews. Reviews of research data and publications are rarely
shared, and even when they are, the lack of industry-wide standard means that each
entity solicits its own reviews before making a decision 34.
Moreover, each peer review can be time consuming due to the lack of data supporting
the conclusions of the article reviewed.
According to the trade association for academic and professional publishers 35, peer
review accounts for 15 million hours of time a year. The process takes on average up to
two or three persons and the average acceptance rate is about 50%. Peer review is often
seen as a quality insurance tool for improving the quality of research studies. However,
the process is often redundant across the scientific community and therefore sometimes
unnecessary.
2.2.2. Impact on collaboration
Collaboration applied to science is all about combining research data and findings from
different organisations within and across disciplines to produce new and better outputs.
Hence, it is much impacted due to research data not being FAIR. A survey 36 conducted at
the University of Sheffield showcased that 31% of the academic staff agreed with the
following statement: “Lack of access to data generated by other researchers or
institutions has restricted my ability to answer research questions”. Another study 37
refers to 50.1% of the respondents having restricted ability to answer scientific questions
due to a lack of access to data generated by other researchers.
Furthermore, a recent study 38 put forth that 73% of academics surveyed said that having
access to published research data would benefit their own research. In correlation with
that figure, survey respondents indicated that sharing research data is important for
doing research in their field. Having FAIR research data could thus enable the reuse of
research data and hence collaboration. A better collaboration could indirectly and
positively impact cross-fertilization 39 and thus innovation.
34 (American Journal Experts, n.d.)

35 (Ware & Mabe, 2012)
36 (Cox & Williamson, 2014)
37 (Tenopir, et al., 2011)
38 https://www.elsevier.com/about/open-science/research-data/open-data-report
39 Cross-fertilization refers to the mixing of data from different disciplines to produce a better result.
14
The probability of finding reports likely depends on their availability in indices such as
Google Scholar and Scopus or in particular research journals and portals. By increasing
the usage of publicly available dissemination channels, FAIR increases the probability for
reports to be found and reused.
In addition, a lot of papers remain uncited 40 because of poor collaboration. Data reuse
can also have a huge impact on citation and thus collaboration and cross-fertilization as
shown by a recent study 41. It has been proven that studies that made data available in a
public repository received more citations than similar studies for which the data was not
made available.
Implementing the FAIR principles could partially increase the collaboration between
scientific communities. For instance, the survey conducted at the University of Sheffield
found that 64% of the respondents would use the data of others, if it was easily
accessible. Between 50 and 60% of the researchers surveyed would be willing to share
data with others via a central data repository without restrictions. However, quantifying
the direct impact of FAIR on collaboration is challenging as one would have to assess the
value of the research that would have been done with FAIR against that of the research
that has been done without.
Because of this, we look at the impact of not having FAIR on collaboration by using the
following proxies:
• Research retraction: research which is retracted as a consequence of not having
FAIR cannot contribute to the progress of science. This includes research that is
retracted because of errors, non-reproducibility, fraud, plagiarism, etc 42. We
postulate that the FAIR principles would decrease fraud and increase research
quality which would be observed in a decrease of the number of articles retracted.
• Research (funding) duplication: redundant research does not contribute to
science.
• Cross-fertilization: research made possible thanks to the FAIR principles which
would not have been possible otherwise (e.g. because the data on which it is
based would not have been reusable).
2.2.3. Impact on innovation
The impact on innovation represents an opportunity cost of not having FAIR research
data. In a survey published in ‘Data Sharing by Scientists: Practices and Perceptions’ 43,
64% of respondents answered that the lack of access to data generated by other
researchers or institutions is a major impediment to progress in science. Another study 44
supports this idea with a figure up to 43%, with the sole difference that they mentioned
a major impediment to progress in their subject discipline. From these observations, we
conclude that having FAIR research data would have a positive impact on innovation as
the opposite is claimed to slow down progress. Beyond research and science, it can be
reasoned that businesses are facing pitfalls too as data availability, use and combination
are crucial for developing new and differentiating existing services and products 45.
Not implementing the FAIR principles impacts innovation in different ways, here is a non-
exhaustive list:
• Lack of access to high value data prevents the development of innovative services
and the creation of new business models.
40 (Davis, 2012)
41 (Piwowar & Vision, 2013)
42 (Fang, Steen, & Casadevall, 2012) and (Wager & Williams, 2011)
44 (Cox & Williamson, 2014)
45 (Wittenburg, Costs of FAIR Compliance and not being FAIR compliant, 2017)
15
• Non-FAIR research data will also prevent the use of Machine Science 46 and what it
entails. Data not being machine readable, it will not be possible for a machine to
process the growing quantity of data and extract insights at a level the human is
not capable of doing.
• Innovation and progress are made possible by building upon the results of
previous work. If research data is not available, new research will always add to
the same baseline, thereby hampering innovation.
• Inability to easily identify possible research collaborators, partners and experts
due to research data not being FAIR.
• Loss of data stemming from poor data management or the absence of a clear
retention policy.
• Lack of clarity about licences and data use conditions prevents the use of non-
FAIR data in value-added projects, due to litigation and licence violation risks.
Overall, the impact of FAIR on innovation translates itself in terms of unrealised
economic growth and job creation.
2.3. Selected indicators
Based on the above three areas and our analysis, the indicators we identified are as
follows:
Areas Indicators
1. Time spent
Impact on research activities 2. Cost of storage
3. Licence costs
4. Research retraction
Impact on opportunities for further research 5. Double funding
6. Cross-fertilization
Impact on innovation 7. Potential economic growth (as % of GDP)
Table 1: Selected indicators per area
Indicator #1: Time spent
In order to cost the time wasted during research activities, we estimate the total time
spent by research staff to perform the research-related activities listed above. From that
total time we derived the time wasted due to non-FAIR data and transformed it into a
financial value using the average salaries for researchers per country, factoring in
differences for academic and non-academic research.
Indicator #1 measures a direct cost faced today by researchers and funders.
Indicator #2: Cost of storage
The cost of storage relates to the electronic storage costs for additional redundant copies
of the data that would otherwise not be needed if the data was FAIR. The total cost of
storage is estimated based on the number of copies, the size of datasets and the storage
duration.
Indicator #2 measures a direct cost faced by research organisations today.
46 Machine Science is defined by the increasing use of computational resources in research-related activities. In
our case, Machine Science specifically refers to machine-readability and reusability of data.
16
Indicator #3: Licence costs
This indicator looks at the licence costs to use data which could have been made
available as open and FAIR data.
Indicator #3 measures a direct cost paid today by research organisations and funders.
Indicator #4: Research retraction
Research retraction measures the cost of research which would not have been retracted
if the FAIR principles had been respected.
Indicator #4 measures an indirect cost faced by researchers today.
Indicator #5: Double funding
While measuring what could have been, i.e. the benefits of more cross-fertilisation, is not
directly possible. We postulate that non-FAIR research leads to duplication of research
projects and the costs linked to redundant research mirrors at least in part the value of
the research that could have been funded instead. This indicator does not include the
duplication of some research activities between different research projects. These are
included in indicator #1: time spent.
Indicator #5 measures a direct cost faced by funders today.
Indicator #6: Interdisciplinarity
Interdisciplinarity refers to the added value of new research combining several academic
disciplines which would be possible thanks to the FAIR principles, compared to the value
of research that would be done otherwise without FAIR.
Indicator #6 estimates an opportunity cost due to the unrealised benefits of FAIR.
Indicator #7: Potential economic growth
Potential economic growth refers to the GDP growth and the number of jobs that would
be created if the FAIR principles were applied widely. By unlocking the value of research
and facilitating the progress of science, FAIR has a positive impact on innovation which
translates itself into job creation and higher GDP.
Indicator #7 estimates an opportunity cost due to the unrealised benefits of FAIR.
2.4. Indicators quantification
For quantifying the selected indicators, we started by collecting data at the

microeconomic scale, namely data applicable to individual research projects and
activities. Once we quantified the costs of not having FAIR data at the micro level, we
extrapolated those results to the level of the European research economy.
Because of the extrapolation from micro to macro level, error margins and estimation
intervals at the micro level may cause the global estimate to be inaccurate. Therefore,
we took the most conservative estimates with regard to the costs of not having FAIR
data and the frequency at which these costs occur. Indicators which are not clearly
quantifiable have been excluded. Consequently, the actual true cost of not having FAIR
data may be significantly higher, due to e.g. unquantifiable spill-over effects or
qualitative externalities that remain unaccounted for.
When collecting the data for the microeconomic indicators, we applied the following
criteria to maximize the representativeness of the data:
• Discipline specificity: data related to specific research discipline will be preferred
over aggregated data for e.g. a research programme. This will allow us to
17
extrapolate information more accurately, as different discipline have different
weights.
• Geographical coverage: our datasets cover multiple countries (e.g. Denmark,
Finland, Netherlands, UK), mostly European. Knowing the research expenditures
and research areas of each country allows us to extrapolate information more
accurately. But due to a shortage of quality resources, results from non-European
regions (e.g. Australia) were included in our report when the research
infrastructure was similar.
• Reusability: data pertaining to cases which are not likely to reoccur have been
avoided to the extent possible.
2.4.1. Indicator #1: Time spent
In order to estimate the time 47 wasted on specific research activities, it is essential to

have an approximated idea of the time spent by researchers on each research activities.
The cost of time spent was identified by multiplying the time wasted by not having FAIR
data with the average wage of academic and non-academic researchers respectively. This
calculation will take into account the number of researchers and the average wages for
each of the 28 Member States.
𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑜𝑜𝑜𝑜 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 × 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊
The time wasted by researchers is identified by factoring the time spent by researchers
on research activities 48 and the inefficiencies due to non-FAIR data.
𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 = 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 × 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟
The share of time spent by researchers on research activities has been studied by (Ziker,
2013) and (Tenopir, et al., 2011) among others. For our calculations, we used an
average compiled from seven studies, the two referred above and (Nur Farah Naadia
Mohd Fauzi, Abd Rashid, Ahmad Sharkawi, Farihah Hasan, & Aripin, 2016), (Court,
2012), (Cheol Shin, Arimoto, Cummings, & Teichler, 2014), (Houghton & Gruen, 2014).
Moreover, we also dissociate between scientific papers (reports) and supporting data in
our calculations wherever relevant.
Overall, the time inefficiencies due to the lack of FAIR data have been estimated to
3.12% for academic researchers according to the share of time spent on each of the
following research activities and the associated ineffeciences. Knowing which share of
time is spent on research by academic and non-academics researchers 49, we
extrapolated this figure to 4.47% of time inefficiencies for non-academic researchers.
2.4.1.1. Creation and collection of the data
We estimated that 31.52% of time spent for finding data could be saved if the FAIR
principles were applied. This figure is based on aggregated statistics 50 about the use and
quality of the metadata (i.e. type of standards, if existing) and assumptions on the time
lost depending on the availability and quality of the metadata.
The time for finding secondary data is directly linked to the quality and richness of the
available metadata. Consequently, we estimate the time in accordance with four levels of
metadata availability:
• No metadata available;
• Metadata not following any standards;
47 For the remaining of this section, the term ‘Time’ will be used in equations as a percentage of time but for
convenience, we will keep the notation ‘Time’.
48 The research activities are defined in section 2.2.1.
(Royal Veterinary College, University of London, n.d.)
(Parsons, Grimshaw, & Williamson, 2013)
18
• Metadata following local/proprietary specifications; and
• Metadata following international recognised standards such as DataCite 51, DCAT-
AP 52 Dublin Core 53, DDI 54, or SDMX 55.
To the time for finding the right secondary data, we add the time lost in searching for
data in multiple access points. Each researcher loses time for accessing relevant data
from journals where an authentication is required. This time lost consists off the time
required to authenticate to the journal as well as the time for managing an account (e.g.
creating, authenticating, changing password). We assume this time to be rather low.
However, implementing the FAIR principles would decrease the number of times a
researcher will be required to authenticate himself or herself.
2.4.1.2. Pre-processing and data cleansing
While the interoperability and reusability principles of FAIR can facilitate the (pre-)
processing of existing data, the time saved in direct correlation to the FAIR principles was
mainly due to a reduction in the effort necessary for the initial cleansing of the data.
However, the time spent in this step is often hard to dissociate from the time for
transformation and integration activities. We therefore estimated the cost of time for this
step together with the step integration of the data.
Furthermore, we understood from interviews with researchers and experts that most of
the time saved by FAIR pertains to other research activities and the impact on the
production of data by researchers is marginal.
2.4.1.3. Integration of the data
The probability of being allowed to reuse an article and the time required to identify this
authorisation depends on the licence provided and the easiness with which a researcher
can find this licence. From there, we estimated the average time required to identify and
understand a licence for secondary data and estimated that FAIR could help reduce this
time by 1.46%.
While we found that the time needed to identify the licence is low, it should be noted that
every time a dataset or document cannot be reused, the time spent finding it has
essentially been wasted.
Furthermore, the probability of being able to integrate multiple sources of secondary data
together directly depends on the formats and quality of the data to be integrated and the
objectives pursued by the research. For estimating the feasibility of the integration, we
look at the percentage of data replicable, if the format of the data is machine readable or
not, and if data quality issues can be resolved or not.
• Replicable data:
o Machine readable; or
o Non-machine readable, for which we differentiate pdf from other
formats.
• Non-replicable data:
o Resolvable; or
o Non-resolvable.
For each of the data quality levels above, we estimated the time lost for being able to
reuse the data and how FAIR would impact this time. For example for resolvable data
51 http://schema.datacite.org/
52 https://joinup.ec.europa.eu/page/dcat-ap
53 http://dublincore.org/specifications/
54 http://www.ddialliance.org/Specification/
55 https://sdmx.org/
19
with quality issues, we based ourselves on a survey from (Biology, 2015) with the time
required to fix bad quality data.
The impact of FAIR on the time and probability of interoperating data is twofold. Firstly,
FAIR pushes researchers to use data formats which are “formal, accessible, shared, and
broadly applicable for knowledge representation” 56 for human and machine readability.
As many reports are only accessible in pdf format, the machine readability is limited.
Secondly, the reusability FAIR principle stipulates that data must be “richly described
with a plurality of accurate and relevant attributes” 57 enabling a human or a machine to
“decide if the data is actually useful in a particular context” 58.
Overall, we estimated that FAIR could help reduce time spent for integrating the data by
28.66%.
2.4.1.4. Analysis of the data
The impact of not having FAIR on the time spent analysing the data is difficult to
dissociate from the time spent due to interoperability and reusability issues faced when
integrating the data.
Due to the lack of specific data regarding the impact of FAIR on this activity, we
estimated the additional time needed specifically for analysis to be insignificant.
2.4.1.5. Registration and publication
When it comes to registration and publication, we found that the main impact of FAIR is
not on the time spent registering and publishing data, but on the number of redundant
copies being made due to inaccessible data. These aspects are quantified by indicator
#2: Cost of storage.
2.4.1.6. Peer review
We assumed that 10% of time spent on peer reviews could be saved if the FAIR
principles were applied. FAIR impacts peer review time by increasing the quality of
research and results reproducibility. While we consider that the need for peer reviews
would remain mostly, FAIR will help reduce the time needed to perform the reviews.
We then factor this assumption with the amount of peer reviews and the amount of time
spent on peer reviews from existing literature to determine the cost of not having FAIR.
The FAIR principles would have a second impact on the time spent by researchers on
peer reviews. Every time a published report is retracted from a journal for reasons such
as errors in the methodology or misconduct of the authors, the time external peer
reviewers had spent on this article is wasted. As the FAIR principles would have an
impact on the amount of reports retractable as described in indicator #3: Research
retraction, it would also contribute to reduce the time wasted by peer reviewers.
2.4.2. Indicator #2: Cost of storage
For publishers and data repositories, the FAIR principles would reduce the cost of storing
data by reducing the need for redundant copies. In order to quantify this cost, we use
the following information:
• Research data volumes per researcher in Europe per year 59.
• The proportion of datasets that are stored in multiples places 60.
• Cloud and University storage costs
56 https://www.go-fair.org/fair-principles/i1-metadata-use-formal-accessible-shared-broadly-applicable-
language-knowledge-representation/
57 https://www.go-fair.org/fair-principles/r1-metadata-richly-described-plurality-accurate-relevant-attributes/
58 https://www.go-fair.org/fair-principles/r1-metadata-richly-described-plurality-accurate-relevant-attributes/
59 Based on sources such as (Addis, 2015) and (Van Tuyl & Michalek, 2015)
60 (Parsons, Grimshaw, & Williamson, 2013)
20
Based on different pricing models 61, we noted that the cost for storing data permanently
has a linear relationship to the volume of data stored.
The following assumptions 62 were used to quantify the added cost of storage due to non-
fair data:
• Empiric evidence 63 shows that on average data is stored in 4.63 repository. In a

perfect open and collaborative world, the data would not need to be stored in
more than one repository. However, FAIR is not a panacea for unnecessary
redundancy. In order to remove unnecessary redundancy, new internal rules and
cultural changes are also required in research (funding) institutions. While FAIR
can contribute to making this change happen, we take the conservative
assumption that implementing the FAIR principles will directly reduce the number
of redundant copies of the data by 20%. The graph below shows the distributions
of the number of copies (i.e. repositories used) in blue and the average sizes of
datasets per number of copies.
Use of repositories and dataset sizes

20,00% 2,5
18,00%
16,00% 2
14,00%
12,00% 1,5
Frequency
Terabytes
10,00%
8,00% 1
6,00%
4,00% 0,5
2,00%
0,00% 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Number of copies
Frequency of use per number of repositories used
Dataset sizes per number of copies (assuming most efficient data management)
Figure 4: Distribution of volume per number of repositories
If we assume that the largest datasets are better managed (because the storage
costs are higher) and map the distributions like presented above 64, then having a
single copy of the data would reduce the total stored volume by 25.67%. This
implies that a 20% reduction across the board is both feasible and conservative as
in reality, not all of the largest datasets would be kept on a single repository.
• Backups which are made as part of the operation of an infrastructure or a
repository are not regarded as separate copies of the data. For our calculation, we
consider backups as a technical measure to prevent the loss of data on one copy
of the data.
• As we rely heavily on surveys and data from the UK (according to the literature,
UK being one of the most advanced countries in terms of data management) we
assume that the data on which we based our assumptions and calculations can be
extrapolated to the EU-28;
• We understand permanent storage as a minimum of 20 years;
61 https://aws.amazon.com/s3/pricing/ and (Royal Veterinary College, University of London, n.d.)

62 Arkivum, Estimating Research Data Volumes in UK HEI, 2015.
63 (Parsons, Grimshaw, & Williamson, 2013)
64 This assumption also gives the most conservative figures.
21
• The total volume of data is estimated based on existing studies which use buckets
for asking researchers about the data volumes used. These buckets do not have a
linear distribution, e.g. 1-50 MB, 50-100MB, 100-250MB etc. 65 As a consequence,
the extrapolations made from such surveys are kept within each of the buckets
used;
• Next to the non-linear distribution, the last proposition of the buckets is most of
the time open-ended, e.g. >1TB. This may create of bias in our calculation for the
storage of the largest files, since one answer in this category could be 2TB or
10TB, with a significant influence on the average.
The cost pricings per volume of data fluctuate with time and vary according to the pricing
model of the repository. In this study, we use the pricing model offered by an UK college
to its own researchers 66, and the data storage pricing models of AWS 67, Azure 68 and
Google 69. Finally, due to the limited availability of figures on volumes of data per
researchers and redundancy, the scope of our calculation is limited to academic research.
Extending our results to the whole research sector in EU would assume that all
researchers in Europe produce the same quantity of data per year.
2.4.3. Indicator #3: Licence costs
First, we calculated the percentage of reports which specify a reusable licence in their
metadata and we looked at the places in which the licence was actually indicated.
Different uses of licences were evaluated:
• If there is no licence, the researcher either:
o Do not reuse the report and its data; or
o Reuse the report and its data without knowing if he or she can.
• If there is a licence in the metadata of the report or of the data, it can be:
o A standardised machine readable licence; or
o A local human readable licence; and
o An open licence; or
o A closed licence which imposes a fee for reusing the data.
In order to quantify the licence costs associated to not having FAIR and open access, we
gathered statistics on the use repartition and we looked at which percentage of data
could be made open. By aggregating studies 70 on open access to data, we consider that
71.5% of the data could be made open. This leaves 28.5% of the academic research data
which must remain closed due to privacy and security reasons.
Secondly, we factored the percentage of data which is already open today. According to
the European Commission 71, in 2015, 48% of the data is available in open access (i.e.
gold and green open access). Similarly a recent study estimated, also for 2015, that 45%
of the data is available in open access.
Thirdly, we looked at the subscriptions and access costs for research institutions. We
found data for public research organisations in United Kingdom 72 and Finland 73 which we
used to extrapolate the costs of licences for academic researchers in the EU28 based on
the number of researchers per country. This indicator only looks at licence costs incurred
65 (Van Tuyl & Michalek, 2015) and (Addis, 2015)

66 (Royal Veterinary College, University of London, n.d.)
67 https://aws.amazon.com/pricing/
68 https://azure.microsoft.com/en-us/pricing/calculator/#storage1
69 https://cloud.google.com/products/calculator/
70 (Tenopir, et al., 2011) and (Johnson, Parsons, Chiarelli, & Kaye, JISC Research Data Assessment Support -
Findings of the 2016 data assessment framework (DAF) surveys, 2016)
71 https://ec.europa.eu/research/openscience/index.cfm?pg=access&section=monitor#viz1489066430689
72 (Lawson, Meghreblian, & Brook, 2015)
73 (Lahti, 2016)
22
today. It does not include the hypothetical licence costs for data that would have been
reused if it had been available in open access.
2.4.4. Indicator #4: Research retraction
In order to quantify the cost of articles retraction, we first identified the volumes and
reasons for retracting research. Various studies 74 estimate the reasons (e.g. non-
reproducibility, errors, fraud, plagiarism, etc.) and the amount of retracted articles
(0.04% of all articles). Based on discussions with experts and calculations, we assume
that FAIR would help reduce the amount of retracted articles by 51.44%. This percentage
is based on a study 75 looking at the causes for research retraction and our assumptions
for the impact of FAIR on each if these causes (i.e. Errors, non-reproducibility, fraud
(suspicion), multiple submission, plagiarism, other misconduct). We then factored the
time spent by researchers on the retracted research to evaluate its cost.
Besides the decreasing percentage of articles retractable, the FAIR principles would also
impact the average time required before an article is retracted. However, as the cost
related to this average retraction time is difficult to estimate, we did not quantify it.
2.4.5. Indicator #5: Double funding
Our estimations of the costs of double funding focus on public research programmes and
research grants. In order to identify to which extent double funding occurs, we rely on
the work of Harold R. Garner, Lauren J. McIver and Michael B. Waitzkin 76 which estimates
double funding by using full-text algorithms to identify overlaps in grant applications.
From their work, we used the percentage of grants application with suspicious overlaps
and the ratio depicting the average size of the first award compared to the potentially
overlapping one.
Their conclusions for US research programmes are extrapolated to European research
programmes using the average EU contribution for H2020 funded projects together with
the total number of public research grants in EU.
The prevention of double funding relies primarily on anti-plagiarism tools which compare
automatically research with available existing articles and data. This method is only
effective to the extent that the plagiarised research is available in a machine-readable
format, which would be ensured by the application of the FAIR principles. Therefore, we
consider that at least 80% of double funding could be avoided with FAIR, as a few will
likely find new creative methods for plagiarism.
Finally, while some private research-funding organisation also use grants, there is no
data which could be used to support the calculations. As a consequence, our estimates
cover double funding in public funded research.
2.4.6. Indicator #6: Interdisciplinarity
In order to quantify interdisciplinarity, we need to identify the volume of research made

possible through FAIR-enabled cross-fertilisation. However, there are few studies that
attempt to quantify the impact of interdisciplinarity and those that do rely on case
studies 77. For this reason, the impact of FAIR on interdisciplinarity cannot not be
estimated with the data currently available.
However, while there is no data to estimate the value of cross-fertilization, it should also
be noted that a new research project made possible with FAIR would always be replacing
another research project which would have taken place otherwise. The impact of FAIR on
cross-fertilization does not in fact increase the total volume of research performed.
74 (Horbach & Halffman, 2017), (Masic, 2012), (Fang, Steen, & Casadevall, 2012), and (Wager & Williams,
2011)
75 (Wager & Williams, 2011)
76 (Garner, McIver, & Waitzkin, 2013)
77 E.g. (Björkdahl, 2009)
23
Therefore, the impact of FAIR on interdisciplinarity is limited to the additional value of
new research compared to that of the research which would have been done without
FAIR.
2.4.7. Indicator #7: Potential economic growth
In order to quantify the potential economic growth, several approaches were followed,
based on the work of Beagrie & Houghton (2014, 2016) and also the work of the Tauri
Group 78. Therefore, we used the following perspectives to look at the impact of FAIR
research data on innovation, which later on can be transformed into an opportunity cost:
• Spinoff: relying on NASA’s contractor work, we estimated the impact on
innovation through the European spinoff network, as spinoffs stem from the
academic world and are heavily reliant on research data. We assumed that due to
research data not being FAIR a certain share of potential spinoffs does not exist.
However, the multiplication effect of FAIR could not be quantified.
• Efficiency impact: we estimated the efficiency impact achieved due to FAIR
research data. Based on calculations in section 2.4.1, the time lost due to
inefficiencies stemming from non-FAIR data can be regarded as time freed up and
reinvested to do other research-related activities. In other words, due to FAIR
research data (i.e. access to new resources) a researcher could produce a same
level output at a lower cost, which is considered to be an efficiency gain. However,
we could not estimate to which extent the time saved by FAIR would be
reinvested in innovative new research and the value thereof.
• Contingent valuation: contingent valuation consists of estimating the value of
non-market goods and services based on preference theory 79. In this case,
individuals are asked what they would pay for a good or a service in a
hypothetical market situation (i.e. pay for something that is actually free).
Following this logic, the ideal process to quantify the FAIR research data impact
on innovation would be to administer a questionnaire with a hypothetical scenario.
The questionnaire would be created in such a way to arrive at a coefficient
depicting the probability to which FAIR research data would contribute to a
researcher’s work. Later on, the coefficient would be transformed in a willingness
to pay to benefit from FAIR research data 80. However, performing an extensive
survey to collect this information was not possible in the timeframe and context of
our study which relies primarily on existing material.
• Loss of data: we tried to identify volumes of data lost due to not having FAIR
research data and estimate the residual value of the lost data. However, without
information on when and how much data is lost in relation to not having FAIR, we
could not estimate the cost of data lost. Another issue is that not all data retains
value equally over time. For instance, traffic information and weather forecasts
lose most of their value very rapidly, while archaeological data retains value long
after discovery.
• Return on public investment: finally, we also tried to measure the return on public
investment on FAIR research, compared to regular research. According to
Houghton, the social return on public investment of R&D should probably be
somewhere 20-60%. However various problems arise, first research is both
publicly and privately funded, and the distribution is unknown. And secondly, what
is the proportion of the return on public investment that FAIR is contributing to.
The link between innovation and growth has been thoroughly studied by economists and
researchers alike. However, such links are always illustrated with specific case studies,
from which it is impossible to extrapolate. For instance, a case study at an institutional
level would have to be extrapolated at discipline, national and European level. That is not
78 https://www.nasa.gov/sites/default/files/files/SEINSI.pdf
79 Reference theory states that a good or service which contributes to human welfare has economic value.
80 To be also included, the calculation of the cost of (re)creating the data.
24
including the fact that a case study is generally discipline specific, and therefore cannot
be a true reflection of another discipline.
Furthermore, the impact of not having FAIR on economic growth can only be quantified if
the impact of FAIR on innovation (e.g. new science, research projects or patents) can be
quantified accurately and with no overlaps. At the time of writing, there is insufficient
data to support the assumptions needed to quantify the economic benefits of FAIR.
Instead, we will present mostly qualitative findings with regard to the economic benefits
of FAIR.
25
3. COST CALCULATION
We estimate the annual cost of not having FAIR data to a minimum of €10.2bn per
year. The actual cost is likely to be much higher due to unquantifiable elements such as
the value of improved research quality and other indirect positive spill-over effects of
FAIR research data.
The following picture shows the repartition of the cost per indicator compared to the total
likely cost of not having FAIR research data, which includes a rough estimate based on
figures for open data. On the left side of the graph, the impact on innovation, would
account for over 60% of the likely cost of not having FAIR research data, while the
minimum true cost of not having FAIR research data, encompassing indicators #1 to #5
accounts for the remaining 40%.
Likely cost of not having FAIR research data
52.39%
~ €16.9bn €10.2bn
43.81%
3.52%
Indicator #5
0.24%
Indicator #4
0.04%
Indicator #1: Time spent Indicator #2: Cost of storage

Indicator #3: Licence costs Indicator #4: Research retraction
Indicator #5: Research duplication Impact on innovation
Figure 5: Cost breakdown
The right part of the graph shows that for the minimum true cost of not having FAIR
research data, the indicator #1: time spent and the indicator #2: cost of storage account
for most of the minimum true cost of not having FAIR research data.
The following sections present the main input data and assumptions used in the
calculations for each indicator. The indicators and the approach to quantifying them are
described in chapter 2. Our calculations, which rely exclusively on secondary data, are
openly available 81.
3.1. Indicator #1: Time spent
81
https://ec.europa.eu/info/files/cost-not-having-fair-research-data-data-and-calculations_en
26
The cost of time spent because of non-FAIR research is €4.5bn per year. It should be
noted that having data solely for the public sector, the impact of not having FAIR on Non-
Academics is extrapolated from the impact of not having FAIR on Academics. The table
below presents the data that was used to estimate this indicator.
Cost of time lost by Non-Academics Unit 2017

Avg. salary for a researcher in the private sector / Non-Academics EUR (€) €61,864
Avg. salary for a researcher in government / Non-Academics EUR (€) €52,853
Time dedicated to research by Non-Academics Time 50.00%

Impact of not having FAIR on Non-Academics Time 4.47%
Number of researchers in the private sector / Non-Academics # 923,997

Number of researchers in government / Non-Academics # 269,963
Total cost of time lost by Non-Academics €3,221,109,809
Cost of time lost by Academics

Avg. salary for a researcher in higher education / Academics EUR (€) €54,484.1
Time dedicated to research by Academics Time 34.96%

Impact of not having FAIR on Academics Time 3%
Number of researchers in higher education / Academics # 727,348
Total cost of time lost by Academics €1,238,244,725
Total cost of time lost due to not having FAIR research data €4,459,354,534
Table 2: Indicator #1 calculation
3.2. Indicator #2: Cost of storage
The cost of redundant storage because of non-FAIR research is €5.3bn per year. The
table below presents the data that was used to estimate the cost of storage that would
not be needed of the FAIR principles were applied.
Inputs on data usage Unit Value

Average volume of data created per researcher/year TB 2.45
Average cost of storing data per TB/year EUR (€) €122
Average number of repositories where data is stored # 4.63
Average data retention period Years 10
Reduction in the number of repositories needed thanks to FAIR % 20%
Average number of repositories where data is stored after the FAIR
implementation # 3.70
Result: Cost of storage per researcher EUR (€) €2,776
Input on the number of researchers 2017

Total number of researchers in EU28 # 1,921,308
Total cost of unnecessary storage due to non-FAIR research data €5,333,228,576
It should be noted that the average volume of data created per researcher/per year
encompasses all types of data (e.g. audio and video recording, pictures, documents,
spreadsheet, markup/code, etc.).
3.3. Indicator #3: Licence costs
27
The cost of licence costs due to the absence of open access is €360m per year. The
table below presents the data that was used to estimate the value of the cost incurred for
not having full open access to research data.
Inputs on open access Unit 2017

(Research) data currently in open access % 56.09%
(Research) data currently not available in open access % 43.91%
Total proportion of (research) data that could be made open % 71.47%

Closed (research) data which is not open for acceptable reasons (e.g. privacy) % 28.53%
Additional data which would be made open by applying FAIR % 31.38%
Licence costs for the academic research organisations in EU28 EUR(€) €1,141,161,883
Total licence costs due to the absence of open access EUR (€) €358,095,416
Due to a lack of data, we could only estimate the licence costs faced by researchers in
the public sector. The actual licence costs due to the absence of FAIR and open research
data is thus likely even higher.
3.4. Indicator #4: Research retraction
The cost of research retraction because of non-FAIR research is €4.4m per year. The
table below presents the data that was used to estimate the value of time lost due to
research retraction which could have been avoider with FAIR.
Cost of time lost by Non-Academics Unit 2017

Avg. salary for a researcher in the private sector / Non-
Academics EUR (€) €61,864
Avg. salary for a researcher in government / Non-Academics EUR (€) €52,853
Time dedicated to research by Non-Academics Time 50.00%

Time lost due to research retraction Time 0.0085%
Reduction of retracted articles with FAIR % 51.44%
Impact of not having FAIR on Non-Academics Time 0.0044%
Number of researchers in the private sector / Non-Academics # 923,997

Number of researchers in government / Non-Academics # 269,963
Total cost of time lost by Non-Academics EUR (€) €3,144,844
Cost of time lost by Academics

Avg. salary for a researcher in higher education / Academics EUR (€) €54,484.1
Time dedicated to research by Academics Time 34.96%

Time lost due to research retraction Time 0.0059%
Reduction of retracted articles with FAIR % 51.44%
Impact of not having FAIR on Academics Time 0.0031%
Number of researchers in higher education / Academics # 727,348
Total cost of time lost by Academics EUR (€) €1,208,927
Total cost of time lost due to retraction €4,353,772
28
It should be noted that FAIR also benefits the quality of articles that are not retracted.
These benefits could not be estimated but it can be reasoned that retracted articles are
only a small portion of the research which could benefit from FAIR. This was confirmed
during interviews with experts who indicated that while FAIR discourages fraud, it also
improves research quality across the board by improving reproducibility.
3.5. Indicator #5: Double funding
The cost of double funding because of non-FAIR research is €25m per year. The table
below presents the data that was used to estimate this indicator.
Input about research duplication Unit 2017

Total percentage of pairs with suspicious overlaps Percent 0.03%
Average EU contribution in H2020 funded projects EUR (€) €1,783,788
Estimated number of public research grants in EU28 # 189,014
Total number of suspicious overlaps within the hypotheses for grants in
#
EU28 50
Total funds allocated to the suspicious pairs EUR (€) €89,185,346
Average size of the first award of the pair compared to the second # 1.9
Total value of the funds for duplicate grants EUR (€) €30,753,568
Reduction of duplication due to FAIR Percent 80.00%
Avoidable cost of double funding EUR (€) €24,602,854
3.6. Indicator #6: Interdisciplinarity
The potential impact on interdisciplinarity that is missed due to not having FAIR research
data could not be estimated reliably. It is acknowledged that the impact of FAIR on
interdisciplinarity does not in fact increase the total volume of research performed, but
rather enhance specifics aspects of the research itself, such as better output quality,
greater collaboration and reusability, etc. On this basis, we identified several elements
that led us to believe the tangible impact of FAIR on interdisciplinarity, but also the
contribution of cross-fertilization to the cost of not having FAIR research data:
• Interdisciplinarity relies in part on reproducibilicy and requires transparency about
tools, methods and data used. This leads to an increased reliability of the findings
underlying scientific publications. Thanks to FAIR research data, the overall
quality of research would be improved.
• For a vast majority of researchers surveyed 82, lack of access to data and poor
quality of data is restricting interdisciplinarity and preventing good research
quality. Introducing FAIR research data could drive to a greater cross-fertilization.
• FAIR would allow researchers to access disparate data from other disciplines,
thereby giving them the opportunity to gain new insights and thus facilitating
knowledge sharing.
To sump up, more interdisciplinarity through the FAIR principles could bring sciences
disciplines closer and increase the current rate of data reuse 83, allowing to explore new
82 https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3126798/table/pone-0021101-t008/
https://www.elsevier.com/about/open-science/research-data/open-data-report
(Cox & Williamson, 2014)
83 (Womack, 2015)
29
possibilities oustide the traditionnal way of conducting research, which would benefit the
science community and other areas connected to it.
3.7. Indicator #7: Potential economic growth
The potential economic growth that is missed due to not having FAIR research data could
not be estimated reliably just as well. However, we identified a number of elements
which lead us to believe the missed economic spill-over effects of FAIR may constitute
the largest component in the cost of not having FAIR research data:
• FAIR and openness have a direct positive impact on the number of citations. All
things equal, FAIR research will be reused more often and thus have more value.
• Resources in open access constitute the second most used source of information
after collections from one’s own institution, according to survey 84. By increasing
the accessibility and reusability of research data, FAIR would also increase the
societal value of research as open research is reused more often.
• Of the minority of researchers who make their data available electronically for
others, the majority does so because they are required to do so, according to a
study 85. Applying the FAIR principles at institutions or funders’ level would thus
greatly improve the availability of research data, thereby creating value.
• Moreover, empirical evidence 86 demonstrates that the share of publications in
open access has been steadily increasing since the 90s. In the absence of
constraining policy or incentives, this strongly suggests that there are benefits for
making research more accessible.
Finally, as the European data economy is estimated to be roughly the same size as the
European research expenditures (both ~ €300bn 87), a parallel can be drawn with open
data. The likely economic benefits for open data were estimated 88 between €11.7bn and
€22.1bn per year in Europe by 2020 and it could therefore be expected for the economic
benefits of FAIR to fall in the same range.
84 http://www.sr.ithaka.org/wp-
content/uploads/2016/06/SR_Report_UK_Survey_Academics_2015_06152016.pdf
85 http://www.ijdc.net/index.php/ijdc/article/view/10.1.210/393
86 https://ec.europa.eu/research/openscience/index.cfm?pg=access&section=monitor#viz1489066430689
87 (European Commission, 2017) and (Eurostat, 2017)
88 (European Commission, 2017)
30
4. CONCLUSION
Interpreting the overall cost of not having FAIR research data as a single value overlooks
many non-quantifiable benefits of FAIR. Nonetheless, at €10.2bn per year in Europe, the
measurable cost of not having FAIR research data makes an overwhelming case in favour
of the implementation of the FAIR principles.
To put this into perspective, research expenditures in Europe amounted to €302.9bn in
2016. While the minimum true cost of not having FAIR can be seen as only 3% of all
research expenditures, €10.2bn per year is 78% of the Horizon 2020 budget per year
and ~ 400%, of what the European Research Council and European research
infrastructures receive combined.
To top this, figures for the open data economy suggest that the impact on innovation of
FAIR could add another €16bn to the minimum cost we estimated.
While this study does not account for the cost of implementing FAIR, if we assume that
the additional costs allocated for data management are up to 2.5% of all research
expenditures, this would leave a positive balance of ~ €2.6bn per year from the
implementation of the FAIR principles. Moreover, not all the costs for implementing the
FAIR principles would be recurrent. Once the proper infrastructure in place, one could
expect the net benefits from the FAIR principles to increase.
Our study presents results for the EU research economy as a whole, however the cost of
not having FAIR varies strongly from discipline to discipline. In some data intensive
disciplines such as genomics or crystallography, (some of) the FAIR principles have
already been implemented already without the need for a quantified cost-benefit
analysis.
As a result, some deductions can be drawn from the work accomplished:
• FAIR is part of a broader movement which is changing the way science is done,
with the emergence of data stewardship and a growing momentum in favour of
openness. Accordingly, it is essential that the necessary infrastructures and
policies are implemented in order to fully benefit from the FAIR principles and
maximise the value of research data.
• With regards to the cost of not having FAIR research, time spent and cost of
storage are the most significant measurable cost drivers. In other words, FAIR
would have a considerable impact on the time we spent manipulating data and the
way we store data.
• Nevertheless, we are confident that the FAIR principles would greatly contribute to
the science ecosystem and innovation in Europe, but the lack of data linking FAIR
and innovation impedes the quantification of the impact of FAIR on innovation.
The cost of not having FAIR research was computed from five quantifiable indicators:
time spent, cost of storage, licence costs, research retraction and double funding. These
indicators cannot possibly cover all the benefits of FAIR i.e. on innovation and economic
growth. Moreover, very conservative assumptions were used and some limitations apply
to these indicators, for instance:
• With time spent, we did not factor that the time wasted due to not having FAIR
could be reinvested in research which would lead to a certain return on
investment.
• With cost of storage, our calculations cover only academic research, because of a
lack of data for the private sector.
• With licence costs, our calculations cover only the cost borne by public research
organisation, because of a lack of data for the private sector.
• With research retraction, we leave out an unknown quantity of research that is
not retracted but also suffers of poor quality and non-reproducibility to some
degree because of non-FAIR data.
We also could not fully consider the impact of not having FAIR on machine readability
and usability. Data volumes produced are growing and in a few years from now
31
researchers will not be able to deal and process those volumes manually. If the data is
FAIR, machines could assist researchers in finding and analysing research data which
would positively impact the science ecosystem, in terms of time freed up, preciseness,
volumes analysed, and velocity but also in terms of new insights drawn.
Therefore, we are confident that the true cost of not having FAIR research data is much
higher than the estimated €10.2bn per year. To estimate it fully, additional data would
need to be collected, in particular for research in private organisations.
As the first world producer in volume of research data, Europe has placed research at the
core of its development strategy. By unlocking additional value from research data,
adopting the FAIR principles is not only a question of efficiency. Not doing so also
comports a clear cost which is diverting resources from the next scientific breakthrough.
32
BIBLIOGRAPHY
Addis, M. (2015). Estimation of research data volumes per researcher in UK HEI.

doi:10.6084/m9.figshare.1577539
American Journal Experts. (n.d.). Peer Review: How We Found 15 Million Hours of Lost Time. Retrieved from
American Journal Experts: https://www.aje.com/en/arc/peer-review-process-15-million-hours-lost-
time/
Baker, M. (2016). 1500 scientists lift the lid on reproducibility. Retrieved from
https://www.nature.com/news/1-500-scientists-lift-the-lid-on-reproducibility-1.19970
Beagrie, N., & Houghton, J. (2014). The Value and Impact of Data Sharing and Curation. Retrieved from
http://repository.jisc.ac.uk/5568/1/iDF308_-_Digital_Infrastructure_Directions_Report%2C_Jan14_v1-
04.pdf
Beagrie, N., & Houghton, J. (2016). The Value and Impact of the European Bioinformatics Institute. Retrieved
from https://beagrie.com/static/resource/EBI-impact-report.pdf
Biology, A. I. (2015). ASCB Member Survey on Reproducibility. Retrieved from http://www.ascb.org/wp-

content/uploads/2015/11/final-survey-results-without-Q11.pdf
Björkdahl, J. (2009). Technology cross-fertilization and the business model: The case of integrating ICTs in
mechanical engineering products. Research Policy, 1468-1477. Retrieved from
https://www.researchgate.net/publication/46489026_Technology_cross-
fertilization_and_the_business_model_The_case_of_integrating_ICTs_in_mechanical_engineering_pro
ducts
Bonino da Silva Santos, L. O., Wilkinson, M., Kuzniar, A., Kaliyaperumal, R., Thompson, M., Dumontier, M., &
Burger, K. (2016). AIR Data Points Supporting Big Data Interoperability. Retrieved from
https://www.researchgate.net/publication/309468587_FAIR_Data_Points_Supporting_Big_Data_Intero
perability
Charles Beagrie Limited. (2011). User Guide for Keeping Research Data Safe - Assessing Costs/Benefits of
Research Data Management, Preservation and Re-use. Retrieved from
https://beagrie.com/static/resource/KeepingResearchDataSafe_UserGuide_v2.pdf
Cheol Shin, J., Arimoto, A., Cummings, W. K., & Teichler, U. (2014). Teaching and Research in Contemporary
Higher Education. Springer. doi:10.1007/978-007-6830-7
Council of the European Union. (2016, May 27). OUTCOME OF PROCEEDINGS: The transition towards an Open
Science system - Council conclusions. Retrieved from
http://data.consilium.europa.eu/doc/document/ST-9526-2016-INIT/en/pdf
Court, S. (2012, October). An analysis of student:staff ratios and academics’ use of time, and potential links
with student satisfaction. Retrieved from https://www.ucu.org.uk/media/5566/An-analysis-of-
studentstaff-ratios-and-academics-use-of-time-and-potential-links-with-student-satisfaction-Dec-
12/pdf/ucu_ssranalysis_dec12.pdf
Cox, A., & Williamson, L. (2014). The 2014 DAF Survey at the University of Sheffield. Retrieved from
http://www.ijdc.net/index.php/ijdc/article/view/10.1.210/393
Data Citation Synthesis Group. (2014, February). Joint Declaration of Data Citation Principles. (M. M., Editor, &
F. 11, Producer) doi:10.25490/a97f-egyk
Data Life Cycle | DataONE. (n.d.). Retrieved January 2018, from DataOne | Data Observation Network for
Earth: https://www.dataone.org/data-life-cycle
Davis, P. (2012). How Much of the Literature Goes Uncited? Retrieved from The Scholarly kitchen:
https://scholarlykitchen.sspnet.org/2012/12/20/how-much-of-the-literature-goes-uncited/
Directorate-General for Research & Innovation (European Commission). (2017). Guidelines to the Rules on
Open Access to Scientific Publications and Open Access to Research Data in Horizon 2020. Retrieved
from http://ec.europa.eu/research/participants/data/ref/h2020/grants_manual/hi/oa_pilot/h2020-hi-
oa-pilot-guide_en.pdf
Directorate-General for Research and Innovation (European Commission). (2017). The economic rationale for
public R&I funding and its impact. Publications Office of the European Commission. Retrieved from
33
https://publications.europa.eu/en/publication-detail/-/publication/0635b07f-07bb-11e7-8a35-
01aa75ed71a1/language-en
Dunning, A., de Smaele, M., & Böhmer, J. (2017). Are the FAIR Data Principles fair? 4TU.ResearchData.
Retrieved from https://zenodo.org/record/321423#.Wku301WnE3F
European Commission. (2017). Final results of the European Data Market study measuring the size and trends
of the EU data economy. Retrieved from https://ec.europa.eu/digital-single-market/en/news/final-
results-european-data-market-study-measuring-size-and-trends-eu-data-economy
European Commission. (2017). Review of the Directive on the re-use of public sector information (Directive
2013/37/EU). Retrieved from https://ec.europa.eu/info/law/better-regulation/initiatives/ares-2017-
4540429_en
Eurostat. (2017). R & D expenditure. Retrieved from Eurostat: http://ec.europa.eu/eurostat/statistics-

explained/index.php/R_%26_D_expenditure
Fang, F. C., Steen, G. R., & Casadevall, A. (2012). Misconduct accounts for the majority of retracted scientific
publications. Proceedings of the National Academy of Sciences of the United States of America, 17028-
17033. doi:https://doi.org/10.1073/pnas.1212247109
Garner, H., McIver, L., & Waitzkin, M. (2013). Same work, twice the money? Nature, 493, 599-601. Retrieved
from https://www.healthra.org/download-resource/?resource-url=/wp-
content/uploads/2013/11/Garner_Nature_1_2013.pdf
Horbach, S., & Halffman, W. (2017). The extent and causes of academic text recycling or ‘self-plagiarism’.
Elsevier, 11. doi:https://doi.org/10.1016/j.respol.2017.09.004
Houghton, J., & Gruen, N. (2014). Open Research Data - Report to the Australian National Data Service
(ANDS). Retrieved from http://apo.org.au/system/files/53613/apo-nid53613-72236.pdf
Johnson, R., Parsons, T., Chiarelli, A., & Kaye, J. (2016). Jisc Research Data Assessment Support - Finding of
the 2016 data assessment framework (DAF) surveys. doi:10.5281/zenodo.177856
Johnson, R., Parsons, T., Chiarelli, A., & Kaye, J. (2016). JISC Research Data Assessment Support - Findings of
the 2016 data assessment framework (DAF) surveys. doi:10.5281/zenodo.177856
Kejser, U. B., Nielsen, A. B., & Thirifays, A. (2011). Cost Model for Digital Preservation: Cost of Digital
Migration. The International Journal of Digital Curation. Retrieved from
http://www.ijdc.net/index.php/ijdc/article/viewFile/177/246
Lahti, L. (2016). Scientific journal subscription costs in Finland 2010-2015: a preliminary analysis. rOpenGov.
Retrieved from http://ropengov.github.io/r/2016/06/10/FOI/
Lawson, S., Meghreblian, B., & Brook, M. (2015). Journal subscription costs - FOIs to UK universities. Retrieved
from figshare:
https://figshare.com/articles/Journal_subscription_costs_FOIs_to_UK_universities/1186832
Masic, I. (2012). Plagiarism in scientific publishing. Acta Informatica Medica, 208-213.

doi:10.5455/aim.2012.20.208-213
Mons, B., Neylon, C., Velterop, J., Dumontier, M., da Silva Santos, L., & Wilkinson, M. (2017). Cloudy,
increasingly FAIR; revisiting the FAIR Data guiding principles for the European Open Science Cloud.
Information Services & Use, vol. 37, no. 1, pp. 49-56.
Nur Farah Naadia Mohd Fauzi, P., Abd Rashid, K., Ahmad Sharkawi, A., Farihah Hasan, S., & Aripin, S. (2016).
A survey on required time allocated vs. actual time spent by academic in pursuit of key performance
indicators (KPIs) at departme,t of quantity surveying, KAED, IIUM. 42-48. Retrieved from
http://jsrad.org/wp-content/2016/Issue%204,%202016/7jj.pdf
Palaiologk, A. S., Economides, A. A., Tjalsma, H. D., & Sesink, L. B. (2012). An activity-based costing model for
long-term preservation and dissemination of digital research data: the case of DANS. International
Journal on Digital Libraries. doi:10.1007/s00799-012-0092-1
Parsons, T., Grimshaw, S., & Williamson, L. (2013, 02 06). Research Data Management Survey. Retrieved from
http://eprints.nottingham.ac.uk/1893/1/ADMIRe_Survey_Results_and_Analysis_2013.pdf
Peter, Larry, Raphael, Gary, & Beth. (2015). FAIR and RDA DFT: sharing the same key messages. Retrieved
from https://b2share.eudat.eu/api/files/e2d84fea-b4e2-41d1-8d4e-6ce2d315b2b9/comparison-fair-
dft-v2.pdf
34
Piwowar, H. A., & Vision, T. J. (2013). Data reuse and the open data citation advantage. Retrieved from
https://peerj.com/articles/175/?utm_content=buffer9059d&utm_source=buffer&utm_medium=twitter
&utm_campaign=Buffer
Royal Veterinary College, University of London. (n.d.). Costing Research Data Management. Retrieved from
https://www.rvc.ac.uk/research/about/research-data-management/before-a-project/costing-research-
data-management : https://www.rvc.ac.uk/research/about/research-data-management/before-a-
project/costing-research-data-management
Stehouwer, H., & Wittenburg, P. (2014). RDA Europe: Data Practices Analysis. Research Data Alliance Europe.
Retrieved from https://b2share.eudat.eu/api/files/0312c522-968c-43e2-8127-9772d68ba084/iCORDI-
D2%205-final-submit.pdf
Tenopir, C., Allard, S., Douglass, K., Umur Aydinoglu, A., Wu, L., Read, E., . . . Frame, M. (2011). Data Sharing
by Scientists: Practices and Perceptions. PLoS ONE. doi:https://doi.org/10.1371/journal.pone.0021101
Van Noorden, R. (2013). Open access: The true cost of science publishing. Nature. Retrieved from
https://www.nature.com/news/open-access-the-true-cost-of-science-publishing-1.12676
Van Tuyl, S., & Michalek, G. (2015). Assessing Research Data Management Practices of Faculty at Carnegie
Mellon University. doi:10.7710/2162-3309.1258
Wager, E., & Williams, P. (2011). Why and how do journals retract articles? An analysis of Medline retractions
1988-2008. Journal of medical ethics, 9. doi:10.1136/jme.2010.040964
Ware, M., & Mabe, M. (2012). The stm report - An overview of scientific and scholarly journal publishing.
Retrieved from http://www.stm-assoc.org/2012_12_11_STM_Report_2012.pdf
Wilkison, M. D., Dumontier, M., & Mons, B. (2016). The FAIR Guiding Principles for scientific data management
and stewardship. doi:10.1038/sdata.2016.18
Wittenburg, P. (2017). Costs of FAIR Compliance and not being FAIR compliant.
doi:10.23728/b2share.e184bd1ff12d45269de80c3f3e443eb7
Wittenburg, P., Ritz, R., & Berg-Cross, G. (2015). RDA Data Foundation and Terminology - DFT: Results RFC.
Retrieved from https://b2share.eudat.eu/api/files/266e2ea2-6200-4b9d-9d3b-
ed6d03ff93a5/DFT%20Core%20Terms-and%20model-v1-6.pdf
Womack, P. R. (2015). Research Data in Core Journals in Biology, Chemistry, Mathematics, and Physics.
doi:10.1371/journal.pone.0143460
Zahedi, Z., Haustein, S., & Bowman, T. D. (2014). Exploring data quality and retrieval strategies for Mendeley
reader counts. Retrieved from http://www.asis.org/SIG/SIGMET/data/uploads/sigmet2014/zahedi.pdf
Ziker, J. N. (2013, October 13). Time Allocation Workload Knowledge Study, Phase 1 Report. ResearchGate.
Retrieved from
https://www.researchgate.net/publication/308761975_Time_Allocation_Workload_Knowledge_Study_
Phase_1_Report
35
Annex I FAIR PRINCIPLES
The table below describes different criteria required to adhere to the FAIR principles. The
left column lists the 15 facets corresponding to FAIR, while the right column presents
metrics to assess the FAIR compliance of a digital resource 89. The wording ‘(meta)data’ is
used in cases where the Principles should be applied to both metadata and data.
Table 7: FAIR Principles and facets against a (meta)dataset
The FAIR Guiding FAIR Proposed Metrics 90

Principles
To be Findable:
F1. (meta)data are Identifier Uniqueness
assigned a globally
unique and eternally Presence of an URL linking to a registered identifier scheme, which
persistent identifier uniquely identify the digital resource. Some non-exhaustive examples
are RN, IRI, DOI, Handle, trustyURI, LSID, etc.
Identifier Persistence
Presence of an URL linking to a policy that would manage changes in

the identifier scheme.
F2. data have rich Machine-readability of metadata

metadata description
(defined by R1 below) Attributes to optimize their discovery. Structured and rich metadata
attributes such as title, creator, publication date, keyword(s),
temporal and spatial coverage, etc. Ideally an URL should link to a
document that contains machine-readable metadata for the digital
resource.
F3. (meta)data are Resource Identifier in Metadata

registered or indexed in
a searchable resource. Metadata must explicitly contain the identifier for the digital resource
it describes thus an URL of the metadata and the IRI of the digital
resource it describes must be provided.
F4. metadata specify Indexed in a searchable resource

the data identifier.
The persistent identifier of the resource and one or more URLs that
give search results of different search engines must be provided.
To be Accessible:
A1. (meta)data are Access Protocol
retrievable by their
identifier using a An URL describing the protocol whether it is an open and free
standardized protocol, closed protocol, protocol with royalties must be provided.
communications
protocol Access authorization
A1.1 the protocol is In case of restrictions, the protocol by which the content can be
open, free, and accessed must be fully specified (i.e. whether authorization is needed
universally and a description of the process to obtain access to restricted
89
https://www.go-fair.org/fair-principles/
https://zenodo.org/record/321423#.WhV7oFWnE3E
https://www.nature.com/articles/sdata201618
https://www.force11.org/group/fairgroup/fairprinciples
https://via.hypothes.is/https:/content.iospress.com/articles/information-services-and-use/isu824#ref019
90 https://github.com/FAIRMetrics/Metrics/blob/master/ALL.pdf
36
implementable content).
A1.2 the protocol

allows for an
authentication and
authorization
procedure, where
necessary
A2. (Meta)data are Metadata longevity

accessible, even when
the data are no longer An URL to a formal metadata longevity plan must be provided, as
available metadata must remain discoverable, even in the absence of the data.
To be Interoperable:
I1. (meta)data use a Use a Knowledge Representation Language
formal, accessible,
shared, and broadly Necessary use of languages that are capable of representing concepts
applicable language for in a machine-readable manner. An URL linking to the specification of
knowledge such a language must be provided.
representation.
I2. (meta)data use Use FAIR Vocabularies

vocabularies that follow
FAIR principles The metadata values and qualified relations should themselves be
FAIR, for example, terms from open, community-accepted
vocabularies published in an appropriate knowledge exchange format.
An IRIs representing the vocabularies used for (meta)data must be
then provided.
I3. (meta)data include Use Qualified References

qualified references to
other (meta)data Relationship within (meta)data, and between local and third-party
data, have explicit and ‘useful’ semantic meaning.
To be Reusable:
R1. meta(data) have a Aspects of metadata that help one to evaluate how reusable a dataset
plurality of accurate is. Authors should not attempt to define the possible downstream
and relevant attributes. usages, but rather should provide as many attributes as possible,
beyond those required for their anticipated downstream use.
R1.1. (meta)data are Accessible Usage Licence

released with a clear
and accessible data (meta)data are released with a clear and accessible data usage
usage licence licence. An IRI of the licence (e.g. its URL) for the data licence and for
the metadata licence must be provided. Achieving machine-readability
is a plus and is possible by referring to one of the licences at
http://purl.org/NET/rdflicense
R1.2. (meta)data are Detailed provenance

associated with detailed
provenance Provenance information such as who/what/when produced the data
and why/how was the data produced should be associated with the
data. Two URLs must be provided. One pointing to the vocabularies
used to describe citational provenance, the second one pointing to one
of the vocabularies (likely domain-specific) used to describe
contextual provenance.
R1.3. (meta)data meet Meets community standards

domain-relevant
community standards Provide a certification, from a recognized body, saying that the
resource is compliant with the community standards.
37
Getting in touch with the EU
IN PERSON
All over the European Union there are hundreds of Europe Direct Information Centres.
You can find the address of the centre nearest you at: http://europa.eu/contact
ON THE PHONE OR BY E-MAIL

Europe Direct is a service that answers your questions about the European Union.
You can contact this service
– by freephone: 00 800 6 7 8 9 10 11 (certain operators may charge for these calls),
– at the following standard number: +32 22999696 or
– by electronic mail via: http://europa.eu/contact
Finding information about the EU
ONLINE
Information about the European Union in all the official languages of the EU is available on the Europa website at:
http://europa.eu
EU PUBLICATIONS
You can download or order free and priced EU publications from EU Bookshop at:
http://bookshop.europa.eu. Multiple copies of free publications may be obtained
by contacting Europe Direct or your local information centre (see http://europa.eu/contact)
EU LAW AND RELATED DOCUMENTS

For access to legal information from the EU, including all EU law since 1951 in all the official language versions,
go to EUR-Lex at: http://eur-lex.europa.eu
OPEN DATA FROM THE EU

The EU Open Data Portal (http://data.europa.eu/euodp/en/data) provides access to
datasets from the EU. Data can be downloaded and reused for free, both for commercial and
non-commercial purposes.
FAIR research data encompasses the way to create, store and
publish research data in a way that they are findable, accessible,
interoperable and reusable. In order to be FAIR, research data
published should meet certain criteria described by the FAIR
principles. FAIR originated from the current patchy data
management practices in EU, which are not optimal yet. Several
local initiatives, as well as global ones, are making the move
towards an infrastructure supporting the FAIR principles in order to
get the most of research data. This report aims to estimate the
cost of not having FAIR research data for the EU economy based
on a series of measurable indicators, which were defined based on
existing studies and interviews with subject matter experts.
Studies and reports

Cost of Not Having FAIR Research Data

Uploaded by

Copyright:

Available Formats

Cost of Not Having FAIR Research Data

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cost of Not Having FAIR Research Data

Uploaded by

Copyright:

Available Formats

Cost of not having FAIR

Cost-Benefit analysis for FAIR research data

Written by PwC EU Services

Manuscript completed in March 2018.

Luxembourg: Publications Office of the European Union, 2018

PDF ISBN978-92-79-98886-8 doi: 10.2777/02999 KI-02-19-023-EN-N

© European Union, 2018.

Cost of not having FAIR

Written by PwC EU Services

Directorate-General for Research and Innovation

Figure 1: Four foundational characteristics of FAIR

Findability, Accessibility, Interoperability and Reusability – the FAIR principles – intend to

1 FAIR Principles described by GO-FAIR, https://www.go-fair.org/fair-principles/; FAIR Principles described by

1.2. Intended audience

This report is intended primarily for:

1.4. Structure of the report

The remainder of this study is structured as follows:

18 (Directorate-General for Research & Innovation (European Commission), 2017)

Figure 2: High-level representation of the methodology followed

• Preciseness: the indicator is related to specific activities, which are defined in

2.2. Indicators identification

2.2.1. Impact on research activities

22 (Dunning, de Smaele, & Böhmer, 2017)

Figure 3: Research activities

2.2.1.1. Creation and collection of the data

2.2.1.2. Pre-processing and data cleansing

2.2.1.3. Integration of the data

2.2.1.4. Analysis of the data

2.2.1.5. Registration and publication

30 Interview with Barend Mons, 2018-01-17

2.2.1.6. Peer review

2.2.2. Impact on collaboration

34 (American Journal Experts, n.d.)

2.2.3. Impact on innovation

2.3. Selected indicators

Impact on innovation 7. Potential economic growth (as % of GDP)

Table 1: Selected indicators per area

Indicator #1: Time spent

Indicator #1 measures a direct cost faced today by researchers and funders.

Indicator #2: Cost of storage

Indicator #2 measures a direct cost faced by research organisations today.

Indicator #4: Research retraction

Indicator #4 measures an indirect cost faced by researchers today.

Indicator #5: Double funding

Indicator #5 measures a direct cost faced by funders today.

Indicator #6: Interdisciplinarity

Indicator #6 estimates an opportunity cost due to the unrealised benefits of FAIR.

Indicator #7: Potential economic growth

Indicator #7 estimates an opportunity cost due to the unrealised benefits of FAIR.

2.4. Indicators quantification

For quantifying the selected indicators, we started by collecting data at the

2.4.1. Indicator #1: Time spent

In order to estimate the time 47 wasted on specific research activities, it is essential to

2.4.1.1. Creation and collection of the data

2.4.1.2. Pre-processing and data cleansing

2.4.1.3. Integration of the data

2.4.1.4. Analysis of the data

2.4.1.5. Registration and publication

2.4.1.6. Peer review