Presented by Rohit Bagalkot: September 29, 2016
Everyone says programmatic is the future of media buying. But what exactly is it? How does programmatic work? And what do things like DSP and DMP mean? This class will go through the basics of programmatic buying, specifically a brief history of how programmatic came to be, the flow of programmatic, its components, and some behind-the-scenes look at how it is optimized.
2. Programmatic Media 101
• Market Trends
• What is Programmatic?
• The Programmatic Evolution
• Programmatic @ TMK
• The Programmatic Buying Process
• The Trade Desk
• The Future
8. 8
Programmatic Media Buying Defined
The use of software to automate the
purchasing of digital advertising
• Prioritizes the audience rather than the site
• Simplifies point of access for inventory
• Accesses robust data for smarter decisions
• Provides real-time operational controls
9. 9
Programmatic Media Buying Defined
Right
Audience
Right
Medium
Right
Time
Right
Message
• Who are we
speaking to?
• How are they
related to the
product/service?
• How deep is the
relationship?
1st Party data,
3rd party data,
Demo,
Behavioral,
Attitudinal etc.
• How can past
behavior tailor
creative?
• How can real
time location, site
data inform
better creative?
Dynamic
Creative using
pub site, user
location etc.,
site behavior
• Is it Social?
Native? Mobile?
Video?
• How can the
user journey
inform the choice
of medium?
Programmatic
Native, Video,
Rich media,
banner & social
channels
• Can we reach
the user when
they’re most
likely to buy?
• How can time of
day, location
behaviors,
seasonality
leveraged?
Programmatic
location based
data providers,
time of day
bidding etc.
15. 15
Programmatic Evolution
Centralized
to Trading Desks
Decentralize
to Planning Teams
Early programmatic buying required
fluency in cumbersome platforms and
advanced quantitative skills.
As buying technology evolved, the tools
became much easier to use and media
teams have become more analytical.
16. 16
Lifting the Veil on Transparency
By eliminating a trading desk as a buying intermediary, agencies can provide greater cost
transparency within a traditionally black box pricing system:
Centralized
Model
Decentralized
Model
Publisher SSP DSP
Publisher SSP DSP
Cost of Media 10-15% 11-25%
Trading Desks
?
Advertiser
Cost of Media 10-15% 13%
Advertiser
18. 18
Starting with Our Search Team
Search teams were technically the first programmatic buyers, experienced with
• Bidding on auctions
• Optimizing in real time
• Managing with UIs
• Integrating with planning teams
• Providing transparency to complicated topic
19. 19
Evolve Into a Programmatic Agency
• More strategic recommendations
• Closer eye on performance
• Providing transparency to complicated topic
• Less middlemen
20. 20
Misconceptions about Programmatic Media
• Too complicated
• Due to opaque business model of trading desks
• Only remnant inventory
• Opening with top publishers like Forbes, Conde Nast, Hearst
• Programmatic = retargeting
• Only one type of programmatic buying
• Lower CPMs than traditional buys
• Not necessarily when targeting exact audience at right time
• No publishers
• Publisher relationships very important for private market places
• Standard desktop banner ads only
• Includes video, mobile, social native, rising star, skins
22. 22
The Publisher Waterfall
Publisher ad servers show ads based on a priority listing known as “the waterfall”
Publisher Direct and
Automated Guaranteed
PMP
Open Exchange
Advertisers need to consider the implications for media buys
Priority
ScaleandEfficiency
23. 23
Types of Programmatic Buys
Open Exchange
(RTB)
Private Marketplaces
(PMP)
Automated
Guarantee/Direct
Definition
Purchase of ads
through real-time
auction
Restricted auction to
select buyers via
DealIDs
Software enabled
direct buys to help with
RFP and IO process
Publisher/Advertiser Many to Many Some to Some One to One
Pros
Higher scale with
typically cheaper pricing
More premium
inventory and
transparency
Streamline traditional
planning process
Cons
Typically less premium
inventory
Requires time for
negotiation and set-up
Tech in early stages
and cumbersome
26. 26
Programmatic Process: Find Winning Bid
Publisher
Publisher
Ad Server
2a. When ad is needed on a webpage,
publisher requests ad from its ad server
DSP Advertiser
27. 27
Programmatic Process: Find Winning Bid
SSP
2b. If programmatic has priority on
waterfall, ad server reaches out to
SSP for ad
Publisher
Publisher
Ad Server
DSP Advertiser
28. 28
Programmatic Process: Find Winning Bid
2c. SSP requests bids from DSPs of
different advertisers and chooses
highest bidder
SSPPublisher
Publisher
Ad Server
DSP Advertiser
29. 29
Programmatic Process: Place Ad
3. Winning advertiser’s ad server
provides advertiser to publisher
SSPPublisher
Publisher
Ad Server
DSP Advertiser
Advertiser
Ad Server
31. 31
Programmatic Players
Demand-Side Platform (DSP): software used to buy advertising in an automated fashion
Supply-Side Platform (SSP): software used to sell advertising in an automated fashion
Data Management Platform (DMP): data warehouse used to analyze and target audiences
33. 33
Our DSP: The Trade Desk
Problem with typical DSPs
• Line-item system: Campaigns are set-up in a line-item system. When optimizations are needed,
separate campaigns need to be made, which creates major operational efficiencies
• Example: To bid more on “News” sites, a new campaign would be created and managed.
• Black box optimizations: Most DSPs use automated algorithms for optimizations. Unfortunately,
this typically don’t provide transparent insights into what worked and what didn’t.
The Trade Desk’s Solution: Bid Factors
• Multiple bids can be made within a single campaign, removing the need for separate line items
• Example: Bidding more on “News” sites can be done within the same campaign.
• Bid factors empower advertisers with full control, and thus transparency, over optimizations
34. 34
DSP Levers
Targeting Control
Audience
• Demographic
• Interests
• Behaviors
• Geography
Site
• Whitelist
• Blacklist
• Category
• Content
• Budget
• Bids
• Time of Day
• Day of Week
• Frequency
• Recency
• Brand Protection
35. Audience composition of 3rd party data
Example: look-a-like modeling
Online actions on proprietary assets
Example: site retargeting
35
Targeting with Data
Data
1st Party
Retargeting
2nd Party
Publisher
3rd party
DMP
Predictive
Audience
1st Party
Onboarding
Online cookies matched with hashed PII
Example: emails onboarded by LiveRamp
Publisher proprietary data used via PMP
Example: C-Level Execs who read NYT
Segments built by DMPs like BlueKai
Example: Dog lovers, in-market for BMW
36. 36
Things to Consider
• Agency Fee
• DSP Fee
• Lack of Make Goods – Viewability and Ad Fraud
• Team Support
• Not Required by Highly Recommended