The last few weeks were full of news in the debate around online tracking privacy. Not only did the FTC endorse the notion of providing a durable “Do Not Track” option for consumers, Microsoft promptly announced that they are building it into the next version of Internet Explorer, and it includes not only durable opt-out settings, but also opt-in settings as well. Add these developments to the self-regulatory approach already under construction, and a more complete framework for tracking choice is coming into view.
What will be the economic impact of this new framework on the ad business? With my simple model, you can use your own assumptions to approximate how much revenue may be lost — and how much may be gained — by providing enhanced tracking choice across the ad ecosystem. The impact depends on factors like what percentage of users click on ad notices, how many of them make a tracking choice, and to what extent they are invited not simply to opt-out, but instead to engage with and improve their own ad profile.
A simple model
The purpose of this simple model is to estimate the potential revenue impact of providing enhanced tracking choices (including Do Not Track). The model is based on published market forecasts for 2014; the impact will likely grow thereafter as behavioral methods become more effective and prevalent.
The model assumes that tracking choices are implemented through integration of simple and durable browser settings (i.e. all browsers implement an approach like Microsoft’s Tracking Protection Lists); and that choices are presented through in-ad notices anchored to icons in advertisements and website notices, as provided in the Digital Advertising Alliance framework.
The model assumes there will be both affirmative and negative tracking options: “Do Not Track” elections reduce behavioral ad revenue; but theoretically, “Opt-In” interactions increase behavioral ad revenue through more accurate and extensive targeting profiles. To the extent some value-exchange is required for Opt-In, particularly for lesser known or trusted companies, the model includes a cost factor for inducement.
Significantly, the model assumes that a Do Not Track election by any user does not impair the non-behavioral (i.e. contextual) advertising opportunity as to that user. This means tracking choices are implemented in a refined way, with ad and data companies segregating behavioral and non-behavioral targeting methods. See the spreadsheet for some other caveats.
Here are some of the key factors built into in the model including rationales for some initial settings (download the spreadsheet to try with your own assumptions):
- How many users will notice and then click on a behavioral targeting icon in an ad or on a website? (Set at 15%, which is a icon-click rate inferred from recent research published by Better Advertising.)
- What percentage of electing users will choose Do Not Track? (Arbitrarily set at 20% of users who make it to the choices. Comments from Google indicate that of those who encounter Google’s tracking profile manager and opt-out interface, just under 7% elect to opt-out of tracking, 28% edit their profile and the remainder do nothing.)
- What percentage of electing users will choose to Opt-In? (Arbitrarily set at 40% of users viewing choices, or two times the number opting-out; with Google’s ratio being 4x).
- How many clicks will it take to make a choice election and what is the conversion rate at each step? (Arbitrarily the model assumes one interstitial screen with a 50% drop-off, and an 80% completion rate on the browser election process.)
- Is there a multiplier of targeting value if user Opts-In and how much is it? (Arbitrarily set at 50% above average value.)
- What is the cost to induce Opt-In elections? (Arbitrarily set at zero.)
With these starting assumptions, it’s not hard to imagine that Do Not Track elections could reduce behavioral ad revenue by over $100 million in 2014. However, to the extent that Opt-In choices are compelling to consumers and increase targeting value, they can be modeled to defray or even exceed the cost of Do Not Track.
What’s missing or wrong in this model, and how would you change the assumptions? Please post your own version and I will link it here.