»» insights

Closing the value-price gap in PPC advertising Analyzing advertiser data is the key to objective, dynamic value-based pricing

Number 2 — April 2008 Leading companies in many industries, including financial services, telecommunications and retail, are flocking to search engine marketing and other pay-per-click (PPC) advertising opportunities. But are they getting their money’s worth?

»» More efficient online advertising markets

As online advertising, and in particular PPC advertising, continues to expand rapidly, improvements are on the way to create a more efficient market. Advertisers, now shifting an increasing percentage of their spend to the web, are bringing this progress about by demanding greater accountability for results. Greater accountability than traditional television, print and radio advertising is, of course, a key reason for the rising popularity of online alternatives. On the web, users can immediately respond to ads by clicking on them, and advertisers can measure this incoming click traffic, and track what users do on their sites.

In a more efficient network, every click is paid for at the right price, based on how well a publisher’s PPC traffic has historically converted—and how well it is converting right now (e.g., what results have advertisers gotten over the past 15 minutes).



Still there’s a big hole in PPC accountability. Advertisers generally pay a uniform keyword price per click to every web publisher that drives traffic to their site. It’s the same price regardless of how often click traffic from the publisher converts into a sale or other desirable result (requesting or downloading information, signing up for a mailing list, etc.). Thus in many instances, pricing is out of alignment with value.

www.fico.com

Make every decision countTM

Closing the value-price gap in PPC advertising

»» insights To improve the alignment of price and value, we need an objective, data-driven means of comparatively evaluating what happens after the click—on the advertiser’s side. FICO has developed a patent-pending technology we call “click conversion scoring,” capable of analyzing advertiser data in near real time to differentiate and compensate publishers by the conversion effectiveness of their click traffic. The FICO® Click Conversion Score™ is currently deployed in the ValidClick AdExchange™, an open PPC network launched in November 2007 and operated by Kowabunga!®. This paper includes early results showing benefits for not only advertisers, but publishers and consumers too.

The percentage of US annual advertising spend going to online media is expected to double by 2011 Both Forrester Research and Yankee Group expect revenues in the US to reach $50B by 2011. Growth rates in Europe, spearheaded by the UK, are expected to be even higher.

Figure 1: US Interactive Marketing Spend, 2007 to 2012

Figure 2: US User & Online Advertising Revenue Growth, 2007 to 2011

$60,000 $50,000

50 REVENUE (IN BILLIONS)

$70,000

60 Marketing Method Online video Emerging channels Email Online display Search engine

$40,000

250 Online Advertising Revenue (in Billions) Internet Users 6+ (in Millions)

225 40 30 20

$30,000

10

$20,000

0

190.5 16.9

$0 Search engine Online display Email Emerging channels Online video Total % of total advertising

197.2

21.7

27.6

34.3

202.9

41.9

205.8

50.3

200

175

150 2006

$10,000

195.0

200.0

USERS (IN MILLIONS)

INTERACTIVE MARKETING SPEND (US$ MILLIONS)

$80,000

2007

2008

2009

2010

2011

YEAR Source: Yankee Group & Interactive Advertising Bureau

2007

2008

2009

2010

2011

2012

CAGR

$8,056

$10,432

$13,310

$16,775

$20,993

$25,323

26%

$6,126

$7,656

$9,354

$11,137

$12,745

$13,988

18%

$2,710

$3,189

$3,588

$3,892

$4,130

$4,256

9%

$1,040

$2,130

$3,534

$5,301

$7,519

$10,610

59%

$471

$989

$1,859

$3,198

$4,875

$7,153

72%

$31,645

$40,304

$50,261

$61,330

27%

12%

14%

16%

18%

$18,402 8%

$24,396 10%

(numbers have been rounded)

Source: Forrester Research, Inc.



www.fico.com

page 2

Closing the value-price gap in PPC advertising

»» insights »» Like credit scoring,

conversion scoring enables performancebased differentiation

In the financial services, telecommunications and retail industries, credit scoring has vastly improved performance over the days of “one size fits all” account management by enabling data-driven differentiation. Companies today routinely segment their portfolios and assign treatments based on analytic assessments of the risk and reward potential of individual accounts. Similarly, click conversion scoring will enable the PPC marketplace to move beyond uniform pricing (while retaining the simplicity of current auction-based methods) to data-driven differentiation. Scores provide an analytic assessment of the conversion potential of click traffic coming from individual publishers, and a means of dynamically reducing keyword bid prices for traffic that does not measure up to market expectations. Through its participation in the ValidClick AdExchange, FICO is acting in a similar capacity in the PPC industry as it does in other industries. Our role is that of an objective, neutral provider of data-driven performance assessment, and the scores our analytics deliver benefit all market participants.

Benefits for all market participants Advertisers: •  Lower costs since the prices advertisers pay are aligned with value derived from conversions, not click-throughs •  Consistent return on investment—every click paid for at the right price, based on historical and current conversion performance of each publisher •  Protection from financial impact of click fraud •  Most fraud detected on the front-end, so advertisers are not charged for suspicious traffic •  Performance pricing acts as second layer of fraud protection, with fraudulent clicks producing zero conversions, and publisher scores and advertiser payments dropping accordingly

Publishers: •  Better prices for better publishers, who are currently being penalized by uniform pricing •  Publisher / ad network visibility into the relative quality of its traffic compared to peer performance •  Pricing incentives for poor-performing publishers to raise their click conversion scores—with near real-time scoring adjustments as conversion rates improve

Consumers: •  Better  web sites and ad targeting, as publishers work to attract traffic likely to not only click through to advertiser sites, but convert once there •  A healthy, continuously improving web environment for buyers and sellers



www.fico.com

page 3

Closing the value-price gap in PPC advertising

»» insights Background: PPC advertising basics Pay-per-click (PPC) advertising started with search engine marketing (“sponsored links” delivered up by search engines above or alongside regular search results), which still generates the bulk of the traffic. Over the past decade, this type of online advertising has expanded exponentially as leading search engines, portals, comparison shopping services and other online aggregators/syndicators have launched ad networks comprising large numbers of affiliated sites and blogs. The PPC marketplace works through an auction-influenced mechanism. Advertisers target search terms (“keywords”) they believe web users in their intended market are likely to type into search fields. Based on the advertiser’s historical experience of how well a particular keyword drives click-through traffic to its site and how well that traffic has converted into sales, the advertiser will then place a bid setting what it is willing to pay for future clicks on that keyword. In a pure auction system, the highest bidders would appear first in the list of sponsored links, but leading networks also consider other factors, such as click-through rate (the number of users clicking on an ad divided by the number of times the ad appeared on the web site) to determine listing order. Advertisers generally pay a uniform price for any traffic generated by clicks on that keyword within the ad network, despite the fact that the value of the traffic may vary widely according to its source. There are mechanisms, of course, for eliminating really bad traffic—the best ad networks include filters for click fraud, and advertisers can “blacklist” publishers on whose sites they don’t want their links to appear. Still, there are myriad shades of gray in publisher performance, and advertisers, as a result, are paying more than they should for some of the traffic they’re receiving. While the auction-influenced pricing mechanism offers many advantages, it does not efficiently address this question of variable traffic value. Advertisers who obtain poor results from PPC traffic will naturally lower their bids in the future. That re-aligns the price they pay to the overall value of the traffic they’re receiving from all publishers in the network. This process penalizes publishers generating high-value traffic, however, since all network affiliates are being paid at the same now-lower rate. The result is a disincentive for higher performance, which has a depressing effect on creative energies, competitive momentum and market growth. Ad network

Advertiser

Bob’s Lexus Bids 40¢ per click keyword lexus



» Disperses ads to publishers

» » »

Participating publishers

» www.pub2.com » www.pub3.com » www.pub1.com

lexus

search

lexus

search

lexus

search

www.Lexus.com www.BobsLexus.com www.LexusofMarin.com

» www.LexusofMarin.com » www.BobsLexus.com www.LexusLovers.com »

www.Lexus.com www.LexusLovers.com www.BobsLexus.com

Publisher gets paid when user clicks on advertiser’s link

» » »

$$ ($0.40) ($0.40) ($0.40) All publishers paid 40¢

www.fico.com

page 4

Closing the value-price gap in PPC advertising

»» insights »» Two firsts for the PPC industry

The FICO® Click Conversion Score™ breaks ground in the PPC industry in two important ways:

1 2

 F  irst scoring of overall performance of an individual publisher based on continuous behavioral feedback loop of advertisers’ data. This closed-loop mechanism allows rapid updating of publisher conversion scores (presently running at ValidClick AdExchange every 15 minutes, with ability to go much faster if necessary). The bid prices publishers receive thus accurately track their historical and current conversion performance.  F  irst scoring of an individual publisher’s performance for a particular advertiser. This capability supports very granular performance metrics, down to individual publisher / ad campaign combinations, at very high transaction volumes.

These firsts are achieved using FICO’s analytic technologies that are already well-proven in the financial services, telecom and retail industries. For instance, the Click Conversion Score incorporates our patented dynamic profiling, which enables key variables from massive quantities of transactional data to be mathematically summarized into an extremely efficient form for realtime analytics. In this case, transactions are the originating clicks as well as the advertiser-side user actions resulting from PPC traffic. The basic Publisher scoring uses dynamic profiling to compare a publisher’s actual measured conversion rate against an “expected” conversion rate based on the historical performance of peers (other publishers displaying an advertiser’s keyword). The analysis takes into account the variable value of each keyword to each advertiser, but rolls up performance across all keywords and advertisers into a single publisher score. Publisher-Advertiser scoring uses a similar approach, but computes the score at the level of a particular click-through pairing. As a result, a publisher can have multiple scores (one for the value of its PPC traffic to auto dealer A, another for auto dealer B, another for auto insurance company Y), greatly increasing accuracy in performance-based pricing. In practice, the two scores are used in a fluid manner to produce the most accurate statistically valid assessment of publisher conversion performance possible in each instance. Where an advertiser is receiving sufficient click traffic from a particular publisher, the Publisher-Advertiser score is dominant. Where traffic is insufficient, scoring defaults to the Publisher score, with the Publisher-Advertiser score refining the result. These scoring techniques are adaptive. New scores are generated with every PPC click generated, and the dynamic profiles are constantly absorbing advertiser-side data on the outcomes of click traffic, including delayed user responses, and using it to adjust scoring. (The feedback loop from the advertiser side can occur in real time with every conversion event, or the advertiser can, if preferred, submit batch data.) This adaptive mechanism enables the analytics to take into account changes in both expected and measured rates of conversion over time while maintaining statistical significance.



www.fico.com

page 5

Closing the value-price gap in PPC advertising

»» insights »» The FICO® Click Conversion Score™ in action: the ValidClick AdExchange

ValidClick AdExchange is an open PPC marketplace operated by Kowabunga! (AMEX:THK). ValidClick AdExchange provides advertisers with safe, consistent, profitable access to the diverse array of PPC traffic—search engines, comparison shopping, contextual implementations, widgets, domains, etc.—while offering web publishers displaying these advertisements top dollar for high-quality click traffic. Kowabunga! helps to ensure strong return on investment for advertisers participating in ValidClick AdExchange by detecting clicks likely to be erroneous or fraudulent (those made, whether by real users or software bots/scripts, for the sole purpose of making the advertiser pay). Its proprietary FeedPatrol technology filters all traffic for click fraud and prevents advertisers from being charged for most suspicious clicks. To address variations in quality within legitimate traffic and ensure fair pricing for advertisers and publishers alike, ValidClick AdExchange incorporates the Click Conversion Score into its standard pricing mechanism. Advertisers place bids in the normal way for what they are willing to pay per click on each keyword. When a publisher makes a request to ValidClick AdExchange for the ads, however, the advertiser’s bid price is automatically adjusted downward for publishers with low conversion scores. Over time, if publishers improve their scores, their adjusted bid price will move upward, moving closer to the ceiling of the advertiser’s original bid price. Such data-driven selective bid price adjustments are a far more efficient mechanism for performance-based pricing than simply allowing poor-quality traffic to drive down prices for everyone. The Click Conversion Score reduces bid prices for only those publishers generating low-value clicks, without affecting publishers generating high-value clicks. It thereby reverses the disincentive and negative dynamic caused by uniform pricing, creating incentives for all publishers to improve the quality of the click traffic coming from their sites. A side benefit is a second layer of click-fraud protection: fraudulent clicks, which produce no conversions, will rapidly cause prices for the publisher generating the traffic to drop to zero. These efficiencies are available to all advertisers, publishers and ad networks participating in the ValidClick AdExchange. Kowabunga! also offers software that enables other PPC networks to incorporate these capabilities into their own operations.

Bob’s Lexus Bids 40¢ per click keyword lexus



»

ValidClick AdExchange

FICO® Conversion Scores

Ad network

Advertiser

50 100

Updated every 15 minutes

25

» » »

Participating publishers

» www.pub2.com » www.pub3.com » www.pub1.com

lexus

search

lexus

search

lexus

search

www.Lexus.com

» www.BobsLexus.com www.LexusofMarin.com www.LexusofMarin.com » www.BobsLexus.com www.LexusLovers.com www.LexusLovers.com » www.Lexus.com www.BobsLexus.com

Publisher gets paid when user clicks on advertiser’s link

» » »

$$ ($0.20) ($0.40) ($0.10) publishers paid based on score

www.fico.com

page 6

Closing the value-price gap in PPC advertising

»» insights »» Results: When publishers

ValidClick AdExchange, launched in 2007, has already demonstrated the range of variation in the quality of PPC traffic. Through the analysis of advertiser conversion data, FICO click conversion scoring clearly reveals that some publishers, as well as the ad networks they participate in, are performing at much higher levels than others.

pay a fair price the marketplace wins

The graph below compares the click conversion score distribution by publisher for two ad networks (collections of publishers). Traffic from Network 1 is converting well for advertisers, and as a result, about 85% of the publishers in the network have the top score of 100, which means they are paid the advertisers’ full keyword bid price. Most of the rest have scores in the 50 to 90 range and are therefore being paid for the clicks they generate at a prorated price. Traffic from Network 2, in contrast, is of poorer quality. Fewer than 5% of the publishers in that network have scores of 100. The majority of publishers are scoring below 50 and being paid at a commensurate level. Through this ability to differentiate click conversion performance, ValidClick AdExchange is able to dynamically adjust PPC bid pricing in a manner that creates an incentive for performance improvements, a positive dynamic that ultimately benefits all market participants. The next two pairs of graphs show how this occurs. In both examples we compare the average prices a particular publisher was paid per day over a two-month period by a leading search engine ad network with the average prices paid per day to that publisher by the ValidClick Ad Exchange.

Figure 3: Last Score of Publishers on two Networks (100, 31): THK themselves attest to ROI over a naïve Network Score 100%

PERCENT OF PUBLISHERS

90%

Network 1 Network 2

80% 70% 60% 50% 40% 30% 20% 10% 0% 0–10

10–20

20–30

30–40

40–50

50–60

60–70

70–80

80–90

90–100

LAST SCORE



www.fico.com

page 7

Closing the value-price gap in PPC advertising

»» insights Figure 4: Poor Publisher Tier 1 search engine network bid prices

ValidClick AdExchange bid prices

$1.00

BID PRICE

$0.80

FICO® Click Conversion ScoreTM introduced

$0.60

$0.40

$0.20

$0.00 11/1/2007

12/18/2007

1/31/2008

11/1/2007

12/18/2007

1/31/2008

TIME (RESULTS ARE TRACKED AGAINST 2 MO)

Figure 5: Conversion performance of a publisher across different advertisers

PUBLISHER-ADVERTISER SCORES FOR A PARTICULAR PUBLISHER

70 60 50

A publisher whose traffic is converting poorly overall may still be converting well for some advertisers or even particular campaigns. The graph at right, for example, shows how a publisher with a Publisher click conversion score of 22 is performing for the 13 advertisers receiving the highest volume of traffic from that publisher.

40 30 20 10 0 1

2

3

4

5

6

7

8

9

10 11 12 13

ADVERTISERS RECEIVING TRAFFIC FROM THIS PUBLISHER



In the left-hand graph above, a publisher generating traffic that converts poorly for advertisers continues to receive the same bid prices from the search engine network across the two-month period. The right-hand graph shows what happened on the ValidClick AdExchange, however, when the FICO® Click Conversion Score™ was introduced. Very quickly the FICO analytics recognized the poor conversion performance of clicks coming from this publisher and prorated the bid price to reflect the real value of the traffic.

The efficiency of the marketplace is clear: Advertisers #1 and #7, for whom the publisher’s traffic is performing well, are paying their full bid price, or close to it, for these clicks. Advertisers #2, 4 and 10 are receiving traffic that performs very poorly indeed from this publisher, and so paying the publisher a steeply reduced bid price that accurately reflects this lack of value. The rest of the advertisers are seeing performance somewhere in-between these extremes and paying bid prices prorated accordingly.

www.fico.com

page 8

Closing the value-price gap in PPC advertising

»» insights In the next pair of graphs, a publisher generating traffic that is converting well experiences only a slight increase in the bid prices paid by the search engine ad network. There’s little incentive for the publisher to do better because price increases have no relation to individual performance. As poor-performing publishers receive reduced prices, advertiser ROI improves… As ROI improves, advertisers raise their bid prices… As bid prices rise, the payment gap between poor-performing and highperforming publishers widens… Publishers receiving reduced rates are motivated to close the gap by taking steps to improve their performance. They sharpen their own marketing, site design and content, online services and ad placement strategies to better target and encourage click-through by consumers likely to convert once on advertiser sites… As publishers deliver click traffic that converts at higher rates, advertiser ROI improves even more… And the virtuous circle continues...

Prices in ValidClick AdExchange, however, respond with greater sensitivity to improvements in individual performance. Very quickly the publisher begins to earn more. There are two market forces at work here: 1. The upward trend may be influenced by advertisers, who, since they are no longer paying full price for poor-converting traffic and thus are obtaining higher ROI, may start raising their bid prices. 2. Any price reductions the publisher may have been experiencing as a result of a low click conversion score gradually diminish as the publisher works to improve performance and starts scoring better. Likely it’s a combination of these factors, which interact in an upwardly spiraling market dynamic to drive higher and higher efficiencies. Everybody wins from this increasingly efficient market. Advertisers win because they obtain better ROI and because publishers working to improve their performance provide more effective online venues for ad placement. Publishers benefit because they have the opportunity to earn more through their own efforts to achieve excellence instead of having their prospects held down by low-performing peers. Consumers benefit from publishers improving their sites, services and targeting mechanisms.

Figure 6: Good Publisher Tier 1 search engine bid prices

ValidClick AdExchange bid prices

$0.25

BID PRICE

$0.20

$0.15

$0.10

$0.05

$0.00 11/1/2007

12/18/2007

1/31/2008

TIME (RESULTS ARE TRACKED AGAINST 2 MO)



11/1/2007

12/18/2007

1/31/2008

TIME (RESULTS ARE TRACKED AGAINST 2 MO)

www.fico.com

page 9

Closing the value-price gap in PPC advertising

»» insights »» Conclusion—value-

based pricing driven by advanced analytics

The financial services industry is a mature marketplace that has long understood the value of segmenting a portfolio of accounts and recognizes that precision in sorting accounts is critical to profitability and customer satisfaction. The PPC industry is poised for a similar transition from “one size fits all” pricing to differentiated pricing. Driven by demand for transparency and fairness, the PPC industry will adopt an exchange model that segments and pays publishers based on conversion value. The importance of precision in the sorting process —both for equity and efficiency—will become increasingly clear, and so will the need for the advanced analytics and fast, scalable delivery engines needed to achieve precision in high-volume transactional environments. With analytics and automation in place, the market will be positioned to move to the next level of growth, attracting and supporting accelerated migration of advertising spend from old media to the PPC environment.

Dynamic scoring of every click requires advanced analytic technology The FICO® Click Conversion Score™ is based on advanced analytics that have already had a transformative impact on the financial services and telecom industries. Key among these is FICO’s patented profiling, which enables vast quantities of historical transactional data for individual entities (such as a publisher) as well as peer groups (such as multiple publishers) to be mathematically summarized for efficient real-time scoring and updated with every new record. In the case of the Click Conversion Score, these updates occur with each click and conversion. FICO’s profiling technology captures both short-term and long-term behaviors of individual entities or combinations of entities (e.g., a particular publisher and a particular advertiser), so scores reflect the most current trends and how they contrast with established patterns. Initially used for credit card fraud detection in conjunction with neural network models, this technology is today an integral component of advanced systems protecting and improving a wide range of financial and telecommunications transactions. Equally innovative FICO software supports the near real-time deployment of dynamic profiling with models and scoring engines in highvolume transactional environments. This software, which enables multiple profiles to be held in memory for simultaneous analysis, is highly flexible and massively scalable. It is currently being used by a leading UK telecom company to detect and predict network fault and revenue leakage—where it is processing over 1.5 billion transactions a day.

The Insights white paper series provides briefings on research findings and product development directions from FICO. To subscribe, go to www.fico.com/insights. For more information

US toll-free +1 888 342 6336

International +44 (0) 207 940 8718

email [email protected]

web www.fico.com

Fair Isaac, FICO, Click Conversion Score and “Make every decision count” are trademarks or registered trademarks of Fair Isaac Corporation in the United States and in other countries. Other product and company names herein may be trademarks of their respective owners. © 2008-2009 Fair Isaac Corporation. All rights reserved. 2470WP 06/09 PDF