A Tale of Two Sensors: Measuring accuracy and

ease-of-use through human-centered design

A 5-household user test in urban Uganda reveals new insights on how best to
design an offline PAYG LPG product

EXECUTIVE SUMMARY

With support from FCDO’s Modern Energy Cooking Services, M-KOPA Labs has
examined the suitability of pay-as-you-go (PAYG) financing to enable East African
homes to affordably purchase LPG cooking fuel to displace traditional charcoal fuel.

In the past two years, it has explored a suite of

off-line technologies to test accurate, remote
monitoring of customers’ LPG usage without
requiring IoT functionality, thus challenging the
need for a ‘smart valve’ solution which remains
unproven as a market viable product. Following
months of testing various low-cost gas
monitoring solutions (see M-KOPA’s A
Comparative Analysis of Enabling Technology
for PAYG LPG report), Labs sought to test the
highest potential offline sensors in a real
household application.

This paper summarises the findings following a

5-household trial over the course of six weeks in
Kampala, Uganda. Labs selected households of
M-KOPA Uganda staff members who indicated
a willingness to participate in this trial gave daily
Figure 1: M-KOPA staff member and test readings of bottom-mounted ultrasonic sensors
user cooking with her offline PAYG LPG and handheld weighing scales via SMS to
prototype cookstove
gauge LPG usage. They subscribed to PAYG
payment plan, requiring a micro-payment sent
via mobile money to assimilate the PAYG experience, understanding that they would
not pay if they did not like the service. At the end of the trial, the five test users
participated in focus group discussion to relay their experiences shifting from
predominately electric and charcoal use to wholly LPG use, as well as impart
recommendations to further shape M-KOPA’s PAYG LPG concept.

TEST OBJECTIVE & APPROACH

The purpose of this trial was to investigate the effectiveness of low-cost sensors to
gather customer usage data when administering a PAYG LPG service model. The
primary questions we intended to answer through this trial are as follows:

• Which sensor holds the most potential, in terms of accuracy, if a PAYG LPG
proposition is scaled?

1
 • What is the superior customer experience with administering PAYG LPG, in
terms of simplicity for customers?
To answer these fundamental questions, Labs designed an experiment to assimilate
a PAYG LPG experience for target customers eager to displace charcoal use with
this modern cooking fuel.

The remainder of this section outlines the key components of the PAYG LPG
prototype, and how Labs designed a user trial to understand the degree of ease and
accuracy of offline gas monitoring to underpin a pay-as-you-cook business model.

Prototype
The intent of an offline PAYG LPG negates the need to attach an IoT-enabled ‘smart’
valve on the LPG canister to monitor daily usage and synchronise payments to
customers’ accounts, thus unlocking continued use of LPG gas. The challenge with
this technology, however, is that the added hardware, telemetry, and other material
costs for a locking valve amount to $43 - although some industry experts note it is
closer to $200. Therefore, the smart locking valve's high cost relative to that of the
LPG cylinder erodes affordability for customers and scalability for providers.

Instead, the M-KOPA solution looks to collaterliase customers’ existing PAYG assets
(smartphones) to secure a line of credit to enable a pay-as-you-cook micro-payment
model with simple, low-cost monitoring technologies. These technologies –
illustrated in Figure 1 below – were selected following a series of lab testing in Labs’
Woking, UK facility and its Kenya-based Large Appliance Testing Station.

The various components, as outlined below, enabled Labs to evaluate the degree of
measurement accuracy for the ultrasonic sensors and the handheld scale, as well as
the degree of ease for the customer to report daily usage. Thus, Labs could validate
the hypothesis that lower cost bottom-mounted ultrasonic sensors can capture more
accurate market information via a smartphone app to underpin a PAYG LPG
business model.

Four sensor brand tests to further examine their ease and accuracy:

• Truma • Mopeka
• Gas watcher • Smart Propane
Figure 2: Diagram of components comprising the PAYG LPG prototype for user testing

2
Payment Plan
The experiment was designed to ensure customers remained active in the trial by
having ‘skin in the game’ and a clear incentive to purchase a second cannister on a
PAYG basis, if needed. Therefore, users were asked to send a nominal daily
payment to M-KOPA, thus assimilating the PAYG repayment experience.
Recognising the price sensitivity of Ugandans and the competitiveness of charcoal
fuel, Labs selected an attractive daily payment that would ensure customers did not
revert to charcoal use – expected to cost Kampala households £0.35 daily.

Payment Plan Daily Rate

Deposit UGX 20,000 / £4.00
Daily payment UGX 800 / £0.16
Days 28
Total UGX 42,400 / £8.42
Table 1: Test PAYG LPG Payment Plan

This payment plan was uploaded onto M-KOPA’s proprietary PAYG platform to verify
and monitor customers’ micro-payments for continued LPG use and participation in
the trial.

Engagement & Data Collection

Upon delivery the 6kg canisters and standard meko grill to customers’ homes,
customers received a tutorial on proper lighting of their LPG grill and a safety
demonstration from the TOTAL Gas representative. Users were also briefed on the
requirements of the pilot by their M-KOPA Engagement Officer, entailing daily
response to an SMS survey to gather the following: meals cooked each day and
readings from both the bottom-mounted ultrasonic sensor and hand-held weighing
scale. Users were also informed that an Engagement Officer is available, should
they run into any issues or need a new canister.

At the end of the study, Labs held a focus group discussion with test users to hear
and collect honest feedback on the ease of use and issues arising from the sensors
and scales, which is outlined below.

User Profiles and Pre-trial Cooking Norms

Test users are all M-KOPA Uganda junior staff members and representative of the
target Kampala urbanite, who aspires to cook with LPG yet predominately cooks with
electricity when it is available yet relies on charcoal as a low-cost and accessible
substitute. We saw a mix of participants living solo or in households with up to 12
members, with an average of 5.6 members per household. Users were instructed to
cook as they normally would, yet account for any adjustments required to
accommodate for LPG’s fast cooking ability.

When asked what cooking fuels they cook with most, all users noted they cook with
charcoal – on average 6 times per week. One respondent notes he cooks with
kerosene daily, while three users cited cooking with electricity daily, if the grid is
available. Only one participants cooked with LPG on a regular basis prior to this trial,

3
yet noted it is very expensive. On a scale of “very unsatisfied” to “satisfied”, only one
participant rated their normal cooking experience as “satisfied” – unsurprisingly, the
user who mostly cooks with electricity and LPG. The remainder rated their
experience as “very unsatisfied” or “unsatisfied” – for reasons relating to cost and
inability to cook when the grid is down or charcoal is wet from the rains.

It should not come as a surprise then that staff members who feel the pinch of
cooking with charcoal on wallets would self-select to participate in trial that would
allow for a break from this consumptive spending, while upgrading to a premium
cooking fuel.

RESULTS OF THE PAYG LPG COOKING TRIAL

This section contextualises the user experience relative to the measurement
accuracy for a range of offline measurement devices to monitor LPG use.

Degree of Technology Accurateness & Ease of Use

Handheld weighing scales:

Of the two solutions, our test users

preferred the handheld weighing
scale for its simplicity. As an analog “If someone is home, they can read the scale
device, users were never troubled if for me to input. Whereas I couldn’t input the
they were not home and needed to sensor reading if I’m not home because my
phone must be close. That’s a challenge if
know the current gas level, as
you’re away.”
anyone in the household could take
- M-KOPA Test user
the measurement.

Of course, M-KOPA, or any PAYG provider, could scale an analog device and
confidently rely on user-reported data to finance a PAYG LPG cannister. In the
future, this technology must be a ‘smart’, i.e. able to digitally collect and report
readings via Bluetooth, which negates the key benefit customers illustrated above.

Moreover, one user noted the handheld sensor became increasingly inaccurate,
providing irregular measurements. After a couple of weeks, the handheld scale
provided a measurement that was higher than previous day. Even when weighing it
multiple times, he would see different readings – leading to confusion and frustration.
We suspect his scale had low batteries, which were replaced during the trial which
corrected the issues.

Bluetooth ultrasonic sensors:

Two-out-of-five test users erred more toward our digital solution – an ultrasonic
sensor installed directly onto the canister that automatically synced and pushed gas
level readings to a smartphone app. The real benefit to users was the ‘set it and
forget it’ ease that didn’t require any human intervention to gauge gas levels, besides
being within close proximity for short-range wireless communications. Interestingly, it
was the single-person households that preferred this option – as they could not
benefit from another household member weighing the cannisters, as mentioned
above.
4
This method was not without its technical issues either. Multiple participants needed
new batteries from constant use over the six-week course. The project Engagement
Officer sent new cells once reported, but in a real-life situation, customers may take
days or weeks to replace the specific cell battery – or may not realise that is the
issue as there is no battery level reading – thus disrupting continued LPG use. This
risk could be mitigated by designing or procuring a weighing scale that can be
recharged via a standard USB port.

Additionally, the sensor readings themselves led to confusing amongst some users.
Some sensors provided users with an estimate of remaining cook time, while others
only saw the remaining gas level. The former was very much preferred as users
because it leaves no guessing game as to whether or not users have sufficient gas
to cook that day’s meals.
Accuracy Comparison:
It is important to remember that this trial was conducted with only five participants,
which is too small of a sample to draw any causal conclusions. Even more, for
unforeseen reasons, one participant (the Smart Propane user) was away from home
too many days to conclusively use data to draw preliminary correlations.
Firstly, Truma proved to be most the reliable sensor in terms of precision and
accuracy. The drawback of this sensor, though, is that it requires more battery
power, which was evidenced in the pilot. Mopeka came in second for accuracy,
however we have reason to believe the sensor’s calibration went awry prior to
delivery, which led to its questionable results. (We will seek to re-test this in the
future.) Lastly, the Gas Watcher sensors provided the largest error in measurement
and did not prove to be easy to calibrate during lab tests.
Safe to say, these three sensors require more extensive testing before any can be
eliminated or confirmed as the right fit for purpose.

Truma Sensor v. Scale [kg]

8
6
4
2
0
22-Feb
23-Feb
24-Feb
25-Feb
26-Feb
27-Feb
28-Feb
01-Mar
02-Mar
03-Mar
04-Mar
05-Mar
06-Mar
07-Mar
08-Mar
09-Mar
10-Mar
11-Mar
12-Mar
13-Mar
14-Mar
15-Mar
16-Mar
17-Mar

Sensor Fill Level [kg] Remainging gas [kg]

Linear (Sensor Fill Level [kg]) Linear (Remainging gas [kg])

5
 Mopeka sensor vs Scale [kg]
8
6
4
2
0
19-Feb 20-Feb 21-Feb 22-Feb 23-Feb 24-Feb 25-Feb 26-Feb 27-Feb 28-Feb 01-Mar 02-Mar

Sensor gas [kg] Remaining gas [kg]

Linear (Sensor gas [kg]) Linear (Remaining gas [kg])

Gas Watcher Sensor vs. Scale [kg]

15

10

0
20-Feb
21-Feb
22-Feb
23-Feb
24-Feb
25-Feb
26-Feb
27-Feb
28-Feb
01-Mar
02-Mar
03-Mar
04-Mar
05-Mar
06-Mar
07-Mar
08-Mar
09-Mar
10-Mar
11-Mar
12-Mar
13-Mar
14-Mar
15-Mar
16-Mar
17-Mar
Sensor Fill Level Remaining gas [kg]

That said, a review of these comparison analyses has deepened M-KOPA’s

Hardware Engineering team on the following technical necessities, especially as they
relate to customer needs and behaviours with these measurement devices:
• Sensor calibration must be conducted and checked prior to installation as
providers cannot rely on customers to perform the calibration at home due to
complexity;
• Ultrasonic sensors show promise as an accurate measure of remaining gas;
and
• Final solution must have multiple sensors to account for different positions of
the canister to optimise measurement accuracy.

Before and After Comparison: Fuel Choice, Frequency of Use and Satisfaction
Unsurprisingly, prior to the trial, users used a diverse mix of cooking fuel and stack
fuels on a daily basis, as is the norm in Uganda given the comparatively low prices of
charcoal and availability of (albeit fragile) electric grid access.

Given these users are located in urban and peri-urban settings, we saw a larger
reliance on electricity for cooking, anecdotally evidenced as the primary source
unless the grid is down in which case users resort to charcoal or an alternative.
Users also noted a preference for using biogas in their individual fuel stack.

Additionally, users that rely mostly on electricity for cooking report higher levels of
satisfaction with their current cooking stack, while those reporting a higher use of
charcoal report a lower satisfaction score. They further contextualised that their poor

6
satisfaction relates to affordability and frustratingly slow cooking times, especially if
the charcoal is unable to heat up from dampness or rainwater.

Figure 3: Before and after mix of commonly used fuels by frequency, and satisfaction
rating with current fuel stack

Once users had access to LPG cooking, more than half of the participants
maintained cooking with their previous primary cooking fuel. Why? Because those
with larger families could not feed their families cooking on only one burner.
Moreover, when probed further during the focus group, participants cited some
difficulties regulating the heat and speed cooking with LPG, prompting some to

7
revert to their customary mix for greater control of the cooking experience and
comfort knowing food would ruin from burning.

Bottomline:
While the sample size of this study was small, Labs uncovered important insights for
to further shape the offline PAYG LPG concept, particularly we consider the user-
friendliness of these devices and their match with lifestyles and tech/digital literacy.
What is clear is, customers value simplicity and the transparency that an app-
enabled sensor provides to inform cooking expenses and fill level for cooking. Even
more, customers want a predictability with: payments, refills, timeline of use, and
even cooking temperatures (i.e., a more user-friendly dial to avoid burning). These
user requirements will directly feed into the ultimate product design based on needs
and wants (see Appendix) to promote adoption amongst target users.
Labs will be deepening this study in further lab tests through a collaboration with
Clean Cooking Alliance, specifically working to further evaluate the technological and
economic pros and cons of an online PAYG gas locking valve or offline alternative as
the appropriate solution.

8
Appendix: Mapping Urban Uganda Needs and Wants to Product Requirements