Postmates & ubereats tv commercials comparison

COM 5401 Advertising Production and

Management Advertising Theories: Computational advertising

Semester A 2018

© Fei Shen, City University of Hong Kong

Recap of Last Lecture

• Four empirical theories • Balance theory (celebrity endorsement)

• Theory of planned behavior (norm pressure)

• EPPM (fear appeal)

• Elaboration Likelihood Model (detailed information vs. shortcut)

• A Hybrid Model – FCB Grid (involvement – thinking/feeling)

• Three practical theories: • Unique selling proposition (USP)

• Emotional selling proposition (ESP)

• Credo selling proposition (SCP)

Which theories are applied here?

Which theories are applied here?

Which theories are applied here?

The advertising industry has been changing… The core competitive skill sets are changing…

Our jobs are taken away bit by bit…

Changing media technology and waning traditional media

Changing media technology and waning traditional media

You are losing your job: manual based editing vs. algorithm based editing

You are losing your job: Computational advertising vs. traditional ads

You are losing your job: Computational advertising vs. traditional ads

From promoting need arousal to matching

Advertising is information

• “I do not regard advertising as entertainment or an art form, but as a medium of information….” [David Ogilvy, 1985]

• The transformation of information incurs a cost. Advertisers purchase “attention” from audience.

Information needs matching! From mass media to new media

• Traditional advertising paradigm: branding image, spread the message, foster likeness.

• New media paradigm: finding the most matching information for an individual.

Quantitative perspective of advertising research Not all attention has value.

“Half the money I spend on advertising is wasted; the trouble is I don‘t know which half.”

John Wanamaker, 1875

• “The time has come when advertising has in some hands reached the status of a science. It is based on fixed principles and is reasonably exact. The causes and effects have been analyzed until they are well understood. The correct method of procedure have been proved and established. We know what is most effective, and we act on basic law.”

Claude Hopkins, 1923

advertiser

audience

media

Mediator

Third party matcher

Matcher optimizer

Data and advertising

New skill sets needed for ad industry • information retrieval

• large scale textual analysis

• statistical modeling

• machine learning

• microeconomics

• game theory

• optimization

• recommendation systems

Computational Advertising

• An approach of advertising to serve the right ad in a right context to the right consumers on the basis of the users’ response that is enabled by computational methods.

Computational Advertising

• Computational creative • For example, the use of an automated design system based on genetic algorithm to

design banner advertising (Gatarski, 2002)

• Computational Placement • Deals with innovative applications of embedding ads in various kinds of digital

content on the real-time user response or use information • Contextual and behavioral placement • User profiling through data mining

• Computational Evaluation • Focuses on new metrics of measuring the effectiveness of digital media advertising

• It is possible to enable real-time response, maximize the precision, and speed up the process.

How many online advertising types you have seen?

Display ads

History: the oldest banner ads

• More than 20 years ago, on 27 October 27, 1994, the first banner went live on hotwired.com. For over four months, 44% of those who saw it clicked on it.

From static banners to dynamic banners – ad server

Basic terms

• Impression • Every occurrence of an ad within the page

• CPM= cost per mille = cost per thousand impressions • Banner ads • Guaranteed delivery

• CPC = cost per click • Search ads

• CPT/CPA= cost per transaction/action (advertiser perspective) • purchasing, join membership • Driving traffic, business model similar to ebates.com

Display ad purchasing

• Direct purchasing • Buying ad space from media owner directly • There is no need of programmed displaying, fast speed of showing the ad. • This is exactly traditional media space purchasing, no accurate targeting.

• Guaranteed delivery • From buying a particular space, to buying individual audience. Targeting individual

with demographics. • e.g., male, 20-40, interested in sports, living in California • Usually order within a specific time range; if the target audience size is not reached,

then could ask for compensation

• Non-Guaranteed delivery • Usually buying from Ad networks (think about real estate brokers) • Optimization

Contextual matching

• Main product: Google Adsense, Microsoft: ContentAds

• Main problem: relatively low CTR (Click-Through Rate), between 0.001-0.1%, unclear motivation of visiting the page

Display ad purchasing

Demand side platform

Ad network and Ad Exchange

• Ad network: Companies that aggregate supply from multiple publishers or other intermediaries and matches it with advertiser demand.

• Ad Exchange: Marketplace for trading impressions between ad networks and some large advertisers or agencies. Think about stock exchange.

DSP and Data Supplier

• Demand Side Platforms (DSP): technology driven optimization for the demand providers. Help advertiser to solve the bidding problem.

• Data Supply and aggregation: BlueKai, eXelate, Experian, Comscore, Nielsen

Audience targeting: Where do the data come from? • Types of data

• Meta data: operation system

• Behavioral data: purchasing, searching, browsing, clicking

• Social network data: social media

• Targeting audience: • Demographics

• Geographical

• Behavior

• Retargeting

• Social network

Demographics

• If we want to promote a 50k USD sports car, who do we want to target?

• Age, gender, income, location, interest

• Challenge: where do the data come from? • Data provided by users: privacy concern, data error (unless you have credit

card and logistics data)

• Model prediction based on behavioral data

Bipartite graph: help filling the blank information when demo data is not available

Retargeting

Retargeting and cookies

• Realized through third party cookie.

• More accurate information from: • Searched product

• Browsed product

• Product in your basket but not checked out.

Privacy issue

• algorithm privacy

• Algorithm could release personal information and personal habit

Social network targeting

• Two types of data • Profile attributes

• Network graph data

• Social network graph data could be used to predict user behavior or identity. This is based on the concept of homophily.

Sponsored search advertising/paid search

Search engine’s job

• Organic search results: (information driven – main products) • Mechanism/algorithm: PageRank, Randex

• Advertising platform: (money driven) • Mechanism: matching content, matching users, profit maximum

Search engine ad flow

• Advertiser: • Define keyword

• Bid keyword

• Pay

• Audience: • Search keyword in search engine

• Search engine: • Matching keyword with webpages

• Matching keyword with advertising

• Display webpage and advertising information

Google “travel agent”

Yahoo “travel agent”

Display URL (destination URL)

headline

Description

Ad extentions (location, call, etc.)

Bid phrase: Travel agent

How search engine make decisions?

• Keyword matching

• Behavioral data matching

(bipartite graph)

Pricing

• Early period (1994): CPM based, high price, individual negotiation, no keyword targeting, slow update speed

• Later (1997): generalized first-price auctions, e.g., goto.com, unstable price

• Now: generalized second-price auctions, Google & Yahoo, using multiple factors for ranking ads

How much does it cost?

How much does it cost?

How much does it cost?

Intro to Bidding

Second price auction (Vickrey)

• The highest bidder wins but the price paid is the second-highest bid.

• For example • Auctioning an ad space

• Bids are $50, $20, $40, and $90

• $90 bidder wins but pays $50

• This game has equilibrium: all players bid their value for the good!

• WHY?

Proof

• Imagine you are a loser: • Decrease your bid? Still lose and receiving nothing (no need to change)

• Increase your bid? • Still lose and receive nothing (no need to change) • Win … but pay more than you value the good (no need to change)

• Imagine your you are a winner:

• Increase bid? Still win and pay the same price (no need to change)

• Decrease bid? • Still win and pay second price (no need to change) • Lose and receive nothing (no need to change)

Second price auction features

• Use dominant strategies: your honest price

• Work with incomplete information

• Compared with first price auction • The highest bidder pays the price they submitted.

• First price auction has no pure strategy equilibrium. Bidders need to revise bidding price multiple times to get best output.

Do all these mean that creative is useless and you are going to

lose your job? Media technologies change;

the basic rules of persuasion do not.

Paradigm Shift

• Broader Integration • Advertising, sales promotion, direct marketing, and public relations all

integrated in online promotion

• Shifted Orientation • Consumers prefer to initiate a brand communication process whenever and

wherever needed – a pull approach

• Brand communications have shifted from a largely “push” approach to a mixed “push” and “pull” approach

• Deeper Engagement • More than involvement; mixing cognitive, affective and conative elements

• Consumer insights

• Sobbing marketing to happy marketing

Throw us a bone campaign

Volkswagen – the fun theory

Platform integration: Let it ring campaign

首先用戶在一個網站（上圖），介紹駕駛安全

的知識，然後提醒你可以轉發給好友。如果你

選擇給好友轉發，那麼你必須留下朋友的手機

以及Email信息。

好友會收到一封如上圖的一封EMAIL，

點擊鏈接，會跳轉到一個視頻網站的

頁面。 可以看到一段駕車行進的視頻

隨著視頻進行，好友

的手機會響

當好友接起電話的時候，視頻畫面（駕車）就會撞上一個

人！緊接著畫面提醒你開車不要接手機，let it ring. 如果始終沒有接起手機，畫面結束時會恭

喜你避免了一場車禍。

Reminder

• Decide on which category for your group to develop a new brand

• Select your two ads for the critique