If your eyes are glazing over at “FLoC origin trials”, here’s why this isn’t as boring as it sounds…
FLoC focuses on how advertisers can target ads to the right people without using privacy-invasive third-party cookies.
Advertisers want to know they’re reaching people who are likely to be interested in their ads. FLoC puts people into specific groups or “cohorts” that let advertisers target those groups without having info about the individuals within.
“Origin trials” are how Chrome experiments with new features, gets feedback, and iterates. Right now, Chrome is testing FLoC in a series of origin trials.
And we’ve been participating for about two months now!
As testers, we have early access to FLoC data — and we analyzed millions of data points to help our advertising partners better understand FLoC and what it means for the future of buying advertising.
You can read the full and in-depth report on CafeMedia.com (with lots more data and examples), but we wanted to pull out some of the most interesting takeaways to share with you here on the AdThrive blog.
Here are ten things you should know about FLoC:
1. FLoC is in testing for a small % of Chrome users right now
FLoC is currently in testing in beta versions of the Chrome browser, so only a small percentage of Chrome users are in FLoCs at this point.
2. Someone can only be in one cohort, but their cohort can change over time
Chrome puts each person into one cohort based on similarities between browsing histories. Each cohort can have thousands of people, but each person can only be in one cohort.
However, as their browsing behavior and the type of content they visit change over time, their cohort can change too.
3. There are currently about 34,000 cohorts!
Chrome identifies cohorts with a long string of numbers called a “SimHash” and groups similar SimHashes together into a cohort.
There are about 34,000 different FLoC IDs, although not every cohort is in use because Google removes any that could be linked to sensitive information.
4. We found that it’s most helpful to look at groupings of cohorts
With so many cohorts but not a ton of data yet — our team decided to put cohorts into more cohorts to do some helpful analysis. (Does the word “cohort” sound weird to you now, too?)
FLoC IDs that are close to each other numerically (for example, 17000 and 17001) include people with lots of overlapping interests. So to make up for the fact that there’s not that much FLoC data available so far, we split the 34,000 FLoCs into groups of 1,000.
We coined the term “kFLoCs” to refer to groups of 1,000 FLoCs.
- FLoC = 1 ID out of 34,000
- kFLoC = a group of 1,000 similar FLoCs
That gave our data team a group of 34 kFLoCs to play around with and look for patterns.
They cross-sectioned the 34 kFLoCs with AdThrive publisher content and created a fascinating map of the types of content each kFLoC visited more than others.
5. kFLoC patterns can help us understand audiences
Zooming out to look at kFLoCs lets us spot some patterns right away!
We can make assumptions like:
- 20000s are into crafting and learning with their family — keywords like “pattern,” “crochet,” “science”
- 14000s focus on career stuff — “business,” “work,” “model”
- 00000s look at pages about “pictures,” “music,” and “support”
- So they might be buddies with one of the 22000s who like “build,” “apps,” and “online”
6. Different content attracts different kFLoCs — pretty cool!
When we look at the types of content different kFLoCs are visiting, there’s definitely a lot of overlap.
But we can also see if different kFLoCs tend to really like (or really avoid) certain topics.
For example, kFLoC 14000 is a niche audience that’s pretty enthusiastic about finance and technology — while kFLoCs 15000–26000 are a lot less interested in that type of content.
And we can see that kFLoC 15000 is Home and Garden’s biggest fan, but some kFLoCs (like 13000, 18000, and 19000) don’t visit as much of this type of content.
7. kFLoC behavior can help us help advertisers
Why do we care which types of content which kFLoCs are visiting?
Because these insights can help us help advertisers reach the right people!
For example, food content appeals pretty equally to many kFLoCs — we’ve all got to eat! But there are a few groups that are less interested in food content (like 7000 and 14000) — and they might be an ideal audience for ads from a food delivery app like DoorDash.
And kFloCs that over-index for a particular type of content can represent the ideal market for advertisers. So kFLoCs that LOVE all things home improvements and gardening could be perfect for ad campaigns from Lowes or The Home Depot.
8. Advertisers aren’t using FLoC yet
It’s important to know that advertisers aren’t using FLoC to target and buy ads — yet. So we all need to do a lot more testing still, but the early data shows some promising ways to help advertisers with that!
9. Expect FLoCs to change as the trials expand
Remember, most of the FLoC data is coming from people using a beta version of Chrome right now. When you think about it, using the beta version of a browser puts you in a unique group of your own anyway, so FLoC isn’t a good representation of the general population yet.
As more browser versions join the trial, we’ll get much more data giving us better insights. So our FLoC and kFLoC analyses will likely change over time — this was just a fun preview of the types of insights we’ll be able to pull from FLoC and share with our ad partners!
10. We’ll keep you in the loop with the latest FLoC news
All these Privacy Sandbox proposals and tests can be a bit hard to follow and esoteric, so count on us to give you the scoop and keep you up to date.
We promise to make it as interesting (and fun) as possible!