K-Anonymity Cohort>Conversion Calculator

[angelinaeng]

Since there has been no final decisions made to cohort/k-anonymity sizes, as well as conversion thresholds, I created a calculator to see how different scenarios would perform by inputting cohort sizes, ad frequency rates, click rates, landing page rates, conversion rates, as well as % of conversions resulting from post view. I used to use this type of math to project how advertisers' campaigns would likely perform and make decisions on where to shift budgets.

Each campaign can vary based on objectives, as well as where it's running (programmatic, social, direct, etc.), and it appears that there could be many scenarios where conversion data would not be reported if the minimum thresholds were to be 100 conversions for an event, and if there was little to no attribution from post-view. This could be especially challenging for small to mid-size marketers.

If most of these scenarios are true, marketers would still find it valuable to see these numbers (including post view) to help calculate CPA and ROI since they are basing their decision on which publishers to invest with and which ones not.

Here's the link to the Google Sheets with my calculations: https://bit.ly/3pS38hq

Note: all fields highlighted can be edited, otherwise the rest of the sheet is locked. Feel free to play around with it.

Angelina Eng VP, Measurement & Attribution @ IAB/IAB Tech Lab

[dmarti]

Have you considered adding a factor for % of cohorts that are usable for a campaign? (Cohorts that reveal sensitive categories will be blocked, so not everyone who is in the market will be reachable by FLoC).

• A general interest group, van intenders, might include several cohorts -- some of which are wheelchair users, which is a cohort based on sensitive information. So the overall van intender market might be reachable by only a subset of the cohorts that make it up.

  • An interest group for people interested in legal services might have a large number of relevant cohorts blocked because of the limits on targeting based on crime victim status or financial distress.

[michaelkleber]

Hi Angelina, thanks for the sheet and especially the range of scenarios.

In the Chrome Privacy Sandbox approach to conversion measurement, you can attribute a conversion event to a specific ad click using the Event Conversion Measurement API, which is already available as an Origin Trial. And a person's flock is knowable at the time of the ad impression.

That means there is no thresholding that gets in the way of a site learning which flocks' clicks converted.

This isn't helpful for some use cases, of course, particularly your scenario C where most conversions come after views without clicks.

It seems like two ideas we've discussed previously would help with understanding flocks and view-through conversions:

• Measurement API issue #98 is specifically about supporting event-level view-through reporting, and what sorts of protections would need to be in place to prevent its abuse to learn someone's cross-site browsing behavior. Perhaps this would solve your problem directly.

• Measurement API issue #85, and before that the SPARROW reporting proposal, talk about conversion reports that flip which side gets event-level versus low-entropy data. Would your use case be adequately addressed by a scenario in which, for each conversion on the advertiser's site, they could learn just a little bit (say three bits) of information about what previous ad campaigns the converting user had seen?

(@csharrison FYI for both of those possibilities.)

[angelinaeng]

@dmarti - thanks for that. I've gone ahead and updated the calculator so that people can play around with the "reach rate". Which means the % of devices you expect to reach from the cohort. This obviously reduces the numbers of impressions, clicks, landing page visits and conversions across the board. Obviously budgets would be a factor in terms of being able to reach the # of max users in the cohorts.

@michaelkleber - thanks for responding. Glad to hear that view-through conversions can be reported, and that there would not be any thresholds for conversion events. Thru the years, I've seen that typically post-view conversions occur within a 24-48 hour timeframe, in some cases 72 hours, even if the latency window at 14 or 30 days. However, this does differ by advertiser. Example auto buying can take up to 60-90 days to convert, or purchasing high-end products (eg. appliances, furniture, etc.) can see conversions at 30 days. However, normally for other advertisers the view-through window can be 2, 7 or 14 (whatever clients want their latency window to be). In some cases, some advertisers would set it at 14, but on the backend look at the data and adjust the latency window depending on the specific product, or for a specific campaign (brand campaigns longer latency vs. direct response), or publisher in the buy (eg. remarketing vendors should have shorter latency windows than sponsorship buys)

It would also be helpful via the event conversion report API and/or Aggregated Reporting API to be able to see assisted & unassisted conversions, as well as duplicated and unduplicated reports. These are definitely helpful reports available thru DV360. Typically we like to see this by publisher, or campaign, or publisher & campaign).

Regarding your question about the 3-bits, in my opinion, advertisers find it more valuable to understand what previous creative ads were seen before converting (not campaign), as well as what publishers served an ad before converting (path to conversion), and see it by each publisher or campaign, if it was a viewed ad or clicked ad.

I'll take a look at the links you referenced again. Appreciate your response.