Lecture Digital Advertising Display Ads
Lecture Digital Advertising Display Ads
Lecture Digital Advertising Display Ads
Display Ads
Yijun Chen
June 2024
Module Structure
2
Previous Recap
Advertising Strategy Workflow
4
Consumer Decision-making Process
3
Action
5
Traditional Ad Spending Decision
6
7
Is the TV Ad Viewed?
8
Attribution Failure
Different marketing campaigns work at different stages of DMP
Attention
TV/outdoor ad increases brand awareness, instagram/search ads
increases brand preferences/interests, complementarity across ad
channels
But only the final purchase decision is observed
Interest
3
Action
9
Challenges in Traditional Advertising
10
Digital & Traditional Advertising
Comparison
In-Class Discussion: TV vs YouTube
12
Digital Advertising Workflow
Targeting, Automation, Personalisation, Measurement: target the right person, in the right
place and at the right time with accurate measurement of effectiveness
13
Target the Right Person
Demographics
Publisher uses data that a customer has volunteered
Or publisher uses data that a customer has provided elsewhere
Contextual targeting
Ad is matched to content it is displayed beside
Geographic targeting
User entered, inferred from IP, `pinging’ computer, shared by mobile
Behavioral Targeting / Retargeting
14
Cross Platform Tracking and Profiling
15
The Power of Data Aggregators
16
Data on Customer Preference
Traditional advertising may rely on customer survey -> reported preference
Digital advertising relies on rich behavioural data -> revealed preference
What’s the difference between revealed preference and reported preference? Which is more
trust-worthy?
17
TikTok For You Page & Recommendation Algorithm
18
Target at the Right Place
Where do your consumers go online, why do they go there?
Business can infer consumer preferences from online behaviors
19
Target at the Right Time
Not only day-of-the-week, time-of-the-week, but also across devices and channels
20
Digital Ads Effectiveness Measurement
Instead of ROI, digital marketers rely on different key performance indicators (KPIs)
KPI is a customised measure that depends on
What stages the campaign influence (AIDA)
Awareness? Preferences? Information search? Point of purchase?
What advertising channel to use
Social media? Search? Display ads?
Who do you target
Loyal customers? General public? New customers?
Modern web tools enable advertisers to determine the current stage of the consumer and use
the right KPI to measure effectiveness
21
Common KPI Discussion
Early Stages: awareness & Middle stage: desire Last stage: action
interest
Social media interaction Add to cart
Display ads
Facebook likes&shares, Retargeting email open rate
frequency and reach Twitter retweets counts
Ad engagement Search ad
mouse-over-time , click search count, click
through rate through rate, bounce rate
New product email YouTube
open rate product review videos
watch time
22
Display Ads
Display Ads
24
How Does this Ad Get Filled?
25
Matching in a Two-sided Platform
26
Two Ways To Sell an Ad
User Generates
Impression Opportunity
Guaranteed Contract
Publisher
Chooses Sales Channel & Advertiser • Bulk ad purchase, pre-specified price &
quantity, time-frame, targeting criteria
Guaranteed Contract Ad Exchange Ad Exchange (Programmatic Advertising)
• Auction-based, real-time, run by a platform
Advertisers
Buy ads targeted at users
27
Guaranteed Contract
28
Guaranteed Contract +/-
Advantages:
Commitment : guarantees ad message gets out
Fixed pricing: stability in terms of budgeting
For branding advertisers, ensures brand-safe content
Disadvantages
Lack of flexibility: need to forecast ad supply and demand and prices
High contracting costs
Potential overpaying: the market condition changes, paying more per impression
29
Use Cases for Guaranteed Contract
30
Ad Exchanges (Programmatic Buying)
A digital marketplace where advertisers and publishers come together to buy and sell ad
inventory in an automated and real-time manner
Intermediary platform that facilitates the auction-based buying and selling of ad impressions
Ad exchanges are operated by Google (AdX), OpenX, AppNexus
Reach
Ad Formats
Targeting capabilities
Auction Mechanisms
Analytics
31
User Generates
Impression Opportunity
Publisher
Chooses Sales Channel & Advertiser Real Time Bidders bid on
individual impressions using
automated bidding
Guaranteed Contract Ad Exchange algorithm (user data,
- Advertiser G1 $2
- Advertiser G2 $1.50 publisher quality, targeting
- Advertiser G3 $4
parameter) -> optimal
bidding amount
Real Time Bidders
Offline Bidders Offline Bidders submit their
- Advertiser OB1 $0.10
- Advertiser OB2 $1
bids price, budget, targeting
RTB1 RTB1 RTB1 - Advertiser OB3 $0.25 information ahead of time
$1 $1.10 No Bid
RTB2 RTB1
No Bid $0.10
32
Advantages of Ad Exchanges
33
Display Ad Buying Business Model Summary
34
35
What are the challenges for display ads?
One impression served = one
impression viewed?
Display Ads Viewability Issue
One impression served = one impression viewed? Same old problem as traditional tv ad
Ad viewability is an industry concern, e.g. intrusive pop up ads
New viewability definition by the Media Rating Council
Ad viewability depends on the location (above the fold/below the fold), size, media, product
category
A display ad is counted as viewable when at least 50% of its area is visible on the screen for
at least one second
A video ad is counted as viewable when at least 50% of its area is visible on the screen while
the video is playing for at least two seconds
The Internet&web technology is evolving with the view ability metrics
38
How to measure effectiveness of
display ad?
41
Apparel Retailer’s Problem: Does My Advertising Work?
42
Experimental Set-Up
The experimenters created a list 3M users to track during the experiment
Individuals who according to the information available would of commercial interest (target
audience of the retailer)
Assign users from the list to the treatment and control groups at random
43
Medicine: Randomized Control Trials
44
Marketing: A/B Tests or Experiments
45
The Importance of Randomization
If we have randomized correctly, treatment and control groups should look the same before
the experiment!
Check that the groups look the “same”
Pre-experiment sales levels
Demographics (e.g. age, gender)
Plot distribution of variables (e.g. histogram)
Difference in average of these variables by group should be close to 0
46
Treatment & Control Group
47
Treatment Group: Users Exposed to Ads
48
4 Kinds of People Exist
49
Observational Estimator of Sales Lift
53
Summary of Ad Profitability
54
Today’s market demands ad viewabiliy
measurement & tomorrow’s market
demands to know incremental lift
Next: Search Advertising