Research Fro rontier of f Real-Time Bid idding based Dis ispla - - PowerPoint PPT Presentation

▶
research fro rontier of f real time bid idding based dis
SMART_READER_LITE
LIVE PREVIEW

Research Fro rontier of f Real-Time Bid idding based Dis ispla - - PowerPoint PPT Presentation

Sep 07, 2023 466 likes •945 views

Research Fro rontier of f Real-Time Bid idding based Dis ispla lay Advertising Weinan Zhang University College London w.zhang@cs.ucl.ac.uk http://www0.cs.ucl.ac.uk/staff/w.zhang August 2015 Bas asic RTB Pro rocess Data User

slide-1
SLIDE 1

Research Fro rontier of f Real-Time Bid idding based Dis ispla lay Advertising

Weinan Zhang University College London w.zhang@cs.ucl.ac.uk http://www0.cs.ucl.ac.uk/staff/w.zhang August 2015

slide-2
SLIDE 2

Bas asic RTB Pro rocess

2

RTB Ad Exchange Demand-Side Platform Advertiser Data Management Platform

  • 0. Ad Request
  • 1. Bid Request

(user, page, context)

  • 2. Bid Response

(ad, bid price)

  • 3. Ad Auction
  • 4. Win Notice

(charged price)

  • 5. Ad

(with tracking)

  • 6. User Feedback

(click, conversion)

User Information

User Demography: Male, 26, Student User Segmentations: Ad science, London traveling

Page User

slide-3
SLIDE 3

Model Bidding Str trategy

  • A function mapping from bid request feature space to a

bid price

  • Design this function to optimise the advertising key

performance indicators (KPIs)

Bid Request

(user, ad, page, context)

Bid Price

Bidding Strategy

3

slide-4
SLIDE 4

Bidding Str trategy in in Pra ractice

Bid Request

(user, ad, page, context)

Bid Price Bidding Strategy

Feature Eng. Whitelist / Blacklist Retargeting Budget Pacing Bid Landscape Bid Calculation Frequency Capping CTR / CVR Estimation Campaign Pricing Scheme

4

slide-5
SLIDE 5

Bidding Str trategy in in Pra ractice: : New Per erspective

Bid Request

(user, ad, page, context)

Bid Price Bidding Strategy

Utility Estimation Cost Estimation

Preprocessing Bidding Function

CTR, CVR, revenue Bid landscape

5

slide-6
SLIDE 6

Dis iscussed Topics of f This Tal alk

Fundamentals ls

  • CTR/CVR Estimation
  • Bid Landscape Forecasting
  • Bidding Strategies

Advances

  • Arbitrage
  • Unbiased Training and Optimisation
  • Conversion Attribution
slide-7
SLIDE 7

CTR/CVR Est stimatio ion

  • A seriously unbalanced-label binary regression problem

– Negative down sampling, calibration

  • Logistic Regression

[Lee et al. Estimating Conversion Rate in Display Advertising from Past Performance

  • Data. KDD 12]
slide-8
SLIDE 8

CTR/CVR Est stimatio ion

  • Follow-The-Regularised-Leader (FTRL) regression

[McMahan et al. Ad Click Prediction : a View from the Trenches. KDD 13]

Closed-form solution

slide-9
SLIDE 9

CTR/CVR Est stimatio ion

  • Factorisation Machines

– Explicitly model feature interactions – Empirically better than logistic regression – A new way for use ser pro rofilin iling

  • GBDT+FM

[Oentaryo et al. Predicting response in mobile advertising with hierarchical importance-aware factorization machine. WSDM 14] [http://www.csie.ntu.edu.tw/~r01922136/kaggle-2014-criteo.pdf]

slide-10
SLIDE 10

Deep Learning Models [our working project]

slide-11
SLIDE 11

Bid Lan andscape Forecasting

Auction Winning Probability Win probability: Expected cost: Count Win bid

slide-12
SLIDE 12

Bid Lan andscape Forecasting

  • Log-Normal Distribution

[Cui et al. Bid Landscape Forecasting in Online Ad Exchange Marketplace. KDD 11]

Auction Winning Probability

slide-13
SLIDE 13

Bid Lan andscape Forecasting

  • Price Prediction via Linear Regression

[Wu et al. Predicting Winning Price in Real Time Bidding with Censored Data. KDD 15]

– Modelling censored data in lost bid requests

slide-14
SLIDE 14

Bidding Str trategies

  • How much to bid for each bid request?
  • Bid to optimise the KPI with budget constraint

Bid Request

(user, ad, page, context)

Bid Price

Bidding Strategy

slide-15
SLIDE 15

Bidding Str trategies

  • Truthful bidding in second-price auction

– Bid the true value of the impression

  • Non-truthful linear bidding

– With budget and volume consideration

[Chen et al. Real-time bidding algorithms for performance-based display ad allocation. KDD 11] [Perlich et al. Bid Optimizing and Inventory Scoring in Targeted Online

  • Advertising. KDD 12]
slide-16
SLIDE 16

Bidding Str trategies

  • Direct functional optimisation
  • Solution: Calculus of variations

CTR winning function bidding function budget

  • Est. volume

16

[Zhang et al. Optimal real-time bidding for display advertising. KDD 14]

slide-17
SLIDE 17

Optimal Bidding Str trategy Solu lution

17

[Zhang et al. Optimal real-time bidding for display advertising. KDD 14]

slide-18
SLIDE 18

Overall Performance – Optimising Cli licks or r Conversions

18

iPinYou dataset [Zhang et al. Optimal real-time bidding for display advertising. KDD 14]

slide-19
SLIDE 19

Dis iscussed Topics of f This Tal alk

Fundamentals ls

  • CTR/CVR Estimation
  • Bid Landscape Forecasting
  • Bidding Strategies

Advances

  • Arbitrage
  • Unbiased Training and Optimisation
  • Conversion Attribution
slide-20
SLIDE 20

Dis iscussed Topics of f This Tal alk

Fundamentals ls

  • CTR/CVR Estimation
  • Bid Landscape Forecasting
  • Bidding Strategies

Advances

  • Arb

rbit itrage

  • Unbiased Training and Optimisation
  • Conversion Attribution
slide-21
SLIDE 21

Dis isplay Advertising In Intermediaries

This work: Intermediary arbitrage algorithms in RTB display advertising.

21

[Zhang et al. Statistical Arbitrage Mining for Display Advertising. KDD 15]

slide-22
SLIDE 22

Intermediary’s Statistical Arbitrage via RTB

  • Statistical arbitrage opportunity occurs, e.g., when

(CPM) cost per conversion < (CPA) payoff per conversion

1000 impressions * 5 cent < 8000 cent for 1 conversion

22

slide-23
SLIDE 23

Sta tatistical Arbitrage Min ining

  • Expected utility (net profit) and cost on multiple

campaigns

CVR estimation winning function bidding function Cost upper bound

  • Est. payoff
  • Prob. of selecting

Campaign i Bid request vol.

23

slide-24
SLIDE 24
  • Optimising net profit by tuning bidding function and

campaign volume allocation

Sta tatistical Arbitrage Min ining

Total cost constraint Risk control

E-Step M-Step

Total arbitrage net profit

  • Solve it in an EM fashion

24

slide-25
SLIDE 25

M-Step: Bidding fu function optimisatio ion

  • Fix v and tune b()

25

slide-26
SLIDE 26

E-Step: Campaign volume allo llocation

  • Multi-campaign portfolio optimisation

where

Portfolio margin variance Portfolio margin mean

Net profit margin

  • n each campaign

26

slide-27
SLIDE 27

Campaign Portfolio Opti timisation Results

27

slide-28
SLIDE 28

Dynamic Portfolio Optimisation

28

slide-29
SLIDE 29

Onli line A/B Test on Big igTree™ DSP

  • 23 hours, 13-14 Feb. 2015, with $60 budget each

29

slide-30
SLIDE 30

Dis iscussed Topics of f This Tal alk

Fundamentals ls

  • CTR/CVR Estimation
  • Bid Landscape Forecasting
  • Bidding Strategies

Advances

  • Arbitrage
  • Unbia

iased Tr Train inin ing an and Opti timis isatio ion

  • Conversion Attribution
slide-31
SLIDE 31

Pro roblem of f Tra raining Dat ata Bia ias

  • Data observation process
  • We want to train the model
  • But we train on the biased data

A bid request Bid Data

  • bservation

If win

[Zhang et al. Learning and Optimisation with Censored Auction Data in Display Advertising. AAAI 2016 Submission]

slide-32
SLIDE 32

Unbiased Tra raining

  • Eliminate the data bias via importance sampling
  • Training target
  • Modelling winning probability via bid landscape
slide-33
SLIDE 33

Unbiased Tra raining

  • Modelling winning probability via bid landscape
  • Only use observed impression data [UOMP]
  • Also use lost bid request data (censored data) [KMMP]

nj: # {winning prices > bj} dj: # {winning prices = bj}

slide-34
SLIDE 34

Exp xperimental Results

  • Winning probability estimation
slide-35
SLIDE 35

Exp xperimental Results

  • CTR estimation: immediate performance improvement
slide-36
SLIDE 36

Dis iscussed Topics of f This Tal alk

Fundamentals ls

  • CTR/CVR Estimation
  • Bid Landscape Forecasting
  • Bidding Strategies

Advances

  • Arbitrage
  • Unbiased Training and Optimisation
  • Co

Conversio ion Att ttrib ributio ion

slide-37
SLIDE 37

Conversion Attribution

  • Assign credit% to each channel according to contribution
  • Current solution: last-touch attribution

[Shao et al. Data-driven multi-touch attribution models. KDD 11]

slide-38
SLIDE 38

Multi-Touch Attribution

  • How to estimate the contribution of each channel?

[Shao et al. Data-driven multi-touch attribution models. KDD 11]

  • A more general formula

[Dalessandro et al. Casually Motivated Attribution for Online Advertising. ADKDD 11]

slide-39
SLIDE 39

[Shao et al. Data-driven multi-touch attribution models. KDD 11]

slide-40
SLIDE 40

Bidding in in Multi-Touch Attribution Mechanism

  • Current bidding strategy

– Driven by last-touch attribution

  • A new bidding strategy

– Driven by multi-touch attribution

[Xu et al. Lift-Based Bidding in Ad Selection. ArXiv 1507.04811. 2015]

slide-41
SLIDE 41

Val alue-based bidding v.s .s. . Lif ift-based bid idding

slide-42
SLIDE 42

Val alue-based bidding v.s .s. . Lif ift-based bid idding

  • Comparison

– Lift-based bidding help brings more conversions to advertisers – but its eCPA is higher than value-based bidding because of last-touch attribution

  • Lift-based bidding with multi-touch attribution could

bring a better eco-system

slide-43
SLIDE 43

Tak aking-home Messages

  • St

Statis istic ical l Arb rbit itrage Mini ining: The internal auction selects the ad with highest arbitrage margin instead of the highest bid price.

  • Unbia

iased Tr Train inin ing: Add the weight to each instance to eliminate the auction-selection bias.

  • Attrib

ibutio ion and Bid iddin ing: Bidding proportional to the CVR lift instead of CVR value.

slide-44
SLIDE 44

Computatio ional l Advertising Research in in Academia

Disa isadvantages

  • Lack of data and online test platform
  • Lack of specific domain knowledge

Advantages

  • Good at mathematic modelling
  • Focus on knowledge collection and communication
  • More research human resource
slide-45
SLIDE 45

OpenBidder Pro roje ject: : www.openbidder.com

  • Online open-source benchmarking project

– Bid optimisation, CTR estimation, Bid landscape etc.

  • Bridge academia and industry research on

computational advertising

45

slide-46
SLIDE 46

Coll llaborations

  • Collaborations are more than welcome!

UK: US: CN:

slide-47
SLIDE 47

Thank You! Questio ions? http://www.computational-advertising.org http://www0.cs.ucl.ac.uk/staff/w.zhang Ad Sci cience WeChat Gro roup