This document discusses optimizing real-time bidding (RTB) campaigns. It provides examples of analyzing media plans and campaign performance from major online spenders. Key aspects that influence campaign performance are identified as the buying algorithm, data quality, and inventory quality. The document explores how DSP technologies can optimize bidding to achieve targeted CPC goals. It also discusses different data sources available for targeting, their typical costs, and how controlling maximum CPM bids and inventory quality are important. Overall, the document advocates controlling one's inventory, data, and optimization to improve RTB campaign performance.
8. But this is not only methodical difference, it is a reality
Programmatic from buy side Programmatic from sell side
then we have strong market imbalance
ADWORDS SKLIK
eTARGET other DSPs
When we have in
Doubleclick
AdX RUBICON PUBMATIC ADMETA
APPNEXUS iBILLBOARD many
others
ADWORDS SKLIK eTARGET 1.43 billion Kč
and in 0.05 billion Kč
other DSPs
9. What happens in a situation of imbalance on market
where every impression is auctioned?
Content network "RTB"
optimistic pessimistic optimistic pessimistic
Estimated spend 1 430 000 000 Kč 1 130 000 000 Kč 70 000 000 Kč 50 000 000 Kč
Estimated impressions 8 339 029 763 14 480 352 594 13 605 815 484 24 605 815 484
CPT 171 Kč 78 Kč 5 Kč 2 Kč
avg. CTR 1,7% 1,7% 0,2% 0,2%
CPC 10,09 Kč 4,59 Kč
11. Single utmost reason for programmatic
Programmatic gives possibility to target individual impressions
Non-programmatic Programmatic
12. There is a fundamental
problem in perception of RTB
Adwords is RTB
Adwords is subset of RTB and this is
subset of programmatic
Programmatic
RTB
Adwords
18. What we mean by
Exchange
• Adwords
• Sklik
• Etarget
• …
Ad Network
• Adactive
• Impression Media
• Spirit Media
• …
SSP
• Rubicon
• Admeta
• Doubleclick AdX
• Pubmatic
• …
19. Maybe this is a reason why programmatic market is not clear
Doubleclick AdX Sklik Etarget SSP Ad Network Programmatic score
Seznam HP x x x x x 0
Youtube.com √ x x x x 1
Rajce.net x √ √ √ x 3
Centrum.cz √ x √ √ x 3
Novinky.cz x √ x √ x 2
Bazos.cz √ x x x x 1
Heureka.cz √ x x x x 1
Blesk.cz x x √ √ x 2
Nova.cz √ √ x x x 2
CSFD.cz √ x x √ √ 3
iPrima.cz √ √ x √ x 3
25. Programmatic buying has three key pillars
Buying algorithm
optimization
Buying algorithm
optimization
Buying algorithm
Data Quality Ionpvteimntiozrayt iQouna lity
Let’s focus on details of these pillars
27. Optimizing assignment for DSP
Targeted CPC – what possibilities DSP has to
achieve particular CPC when we pay for CPM
How big budget do we need for optimized CPC
(same applies to CPA)
28. Simple math on algorithmic optimization
Expectation (campaign goal) 12 Kč CPC
CPM = 1000 x CTR x CPC
Basic CTR 0.15%
Maximum allowable bid 18 Kč CPM
20% improvement = (basic CTR – optimized CTR) /optimized CTR
Optimized CTR 0.18%
Maximum allowable bid 25.2 Kč CPM
Basic Impressions with 0.15% 538664 impressions = 807 clicks
Optimized Impressions 0.18% 574066 impressions = 1207 clicks
Minimum impressions for 95% confidence interval
107732 for 18 Kč CPM
114813 for 25.2 Kč CPM
Minimum budget for 95% confidence about 20% CTR improvement 4832 Kč
Standard error
CPC = CPM / (1000 x CTR)
0.012%
CPC between 18 Kč – 12 Kč
Minimum impressions for 99% confidence interval
269332 for 18 Kč CPM
287033 for 25.2 Kč CPM
Minimum budget for 99% confidence about 20% CTR improvement 12081 Kč
Standard error
CPC = CPM / (1000 x CTR)
0.007%
CPC between 15 Kč – 12 Kč
29. Simple math on algorithmic optimization
To have 95% confidence that with target of 12 Kč per click and default click rate
0.15% we can safely try to improve click rate by 20% and therefore risk actual click
somewhere between 18-12 Kč we need to invest 4832 Kč
Now what parameters which DSP has under control influence
click rate (independent variable)
30. DSP degrees of freedom
Targeted inventory (actual sites, placements)
Bid (maximum, minimum, variable)
Chosen creative
Geography of user
User attributes*
Frequency
Day and time
System parameters (Device, OS, browser)
Connection speed
ISP
Simplified math for our particular example = 3
parameters (geography + daytime + creative) with 3
attributes (Prague vs Brno, Morning vs Evening,
Creative versions) = testing budget = 3x3x4832 =
43.488 Kč
32. If you are on demand side you better look like this
Or be really intelligent.
Ideally
both
33. While ideal look on supply side is this
Or being really intelligent.
Ideally
both
34. RTB = CPM? Or CPC? Or CPA?
“ Doesn't it make more sense to buy everything on CPC base?
”
Well it depends...
Bigger is the base CPC and click rate more expensive is to buy on CPC base
For example in case of 0.5% CTR CPM price of 15 Kč CPC buy is
400% more expensive than opposite version of 15 Kč CPM buy.
39. Decision influencing information flowing to bidder
User data including unique ID of user known to DSP,
site info,
device info,
placement data,
bid floor
40. Capabilities of adwords, sklik, etarget, “RTB”
Adwords Sklik eTarget “RTB”
Formats
text + IAB hlavní +
dynamické inzeráty +
engagement ads + video
inzeráty
text + podmnožina IAB
hlavní Podmnožina IAB hlavní
Podle rozhodnutí media
(libovolný)
Buying models CPC, CPM, CPA, visibility,
true view CPC CPC
Nativně: CPM
DSP modely: CPA, CPC,
visibility,
Targeting
Google zájmy
Google context
Look-a-like
1st party
1st party
Kontext
1st party
Kontext
1st party
Publisher data
Behaviorální
In-market
Networks Doubleclick AdX Sklik eTarget + SSP Libovolný SSP zdroj
41. DSP vs. trading desk vs. retargetter vs. ad exchange
DSP
Doubleclick Bid
Manager, MediaMath,
Appnexus
Gives full transparency in buying
TD
Sociomantic, DataXu,
Xaxis, Criteo
Gives limited transparency in
buying, can apply arbitrage
“Retargetter”
PerfectAudience,AdRoll
Focused on 1st party data
targeting, usually does part of the
buying through 3rd bidder
Ad Exchange
Sklik,Adwords,eTarget
Allows buying on it’s own
inventory only
45. An example of Adwords campaign
Targeting - interests
Categories: Home and garden, Construction and structure,
Agriculture and forestry, Doors and windows, Gardening
Default maximum CPC 10 Kč
Final CTR: 0.14%
Average final CPC: 5.8 Kč
52. Data sources
Buying platform
packaged data
(visibility,
arbitrage,
demographics)
Buying platform
native data
Platform
integrated data
(data exchange
deal as source -
DMP)
Publisher
packaged data
(private deals)
53. Czech programmatic marketplace - data
Context (Doubleclick Bid Manager native, Appnexus-third-party
Grapeshot, CPEX-third-party Demdex, sklik- nativní, Adwords-native)
Visibility (external – DBM pre-negotiated, others e.g. Peer39)
Behavioural (BlueKai, Exelate, Eyeota, HiFive)
Weather (TVTY)
Brand Safety (DBM native, Integral Ad Science, DoubleVerify, Grapeshot)
Synchro TV (Xaxis Sync, TVTY)
54. Example of typical costs of data
Seznam behavioural - +20% to basis CPM (= cca 2-30 Kč CPM)
CPEx Audience packaged data – 1.9 USD including media space, costs of data unknown (= 41.50
Kč CPM)
Grapeshot semantic context data – 0.12 EUR CPM (= 3.4 Kč CPM)
Adwords behavioral, interests, context – part of CPC arbitrage – price unknown
BlueKai interests – 0.32 USD CPM (= 7 Kč CPM)
Eyeota bizographics – 1.20 USD CPM (= 26 Kč CPM)
DBM Demographics – free inside platform
…
55. Control your inventory, data and optimization
and make you campaigns fly
Feel free to download this presentation at http://www.programmatic.cz/rtb_explained/