Using Attribution Windows To Segment Audiences

How To Use Meta's Ad Set-Level Attribution Settings To Adjust Your Audience Segmentation

7-day click, 1-day view. 1-day click, 1-day view. All click, no view. What should you use? It helps to change the framing of this from a reporting question to a targeting question. That sounds weird, I know, but bear with me.

Until recently, at the risk of somewhat embarrassing myself, I was under the impression that selecting an attribution window at the ad-set level affected reporting exclusively. During a recent conversation with some friends, it came up that by changing your attribution window at the ad set level, you could actually inform the algo's learnings and therefore unlock slightly different segments of your audiences. Again, this was news to me, so I went to confirm.

Cody Plofker posted on Twitter recently with a remark about attribution windows, where Yoni, who heads up Marketing Solutions at Meta, stepped in with an interesting tidbit of information – changing your attribution window can actually adjust the way the algo learns and therefore optimizes.

This was surprising to me but exciting. Uncovering new mechanics in Ads Manager means a new tool under the belt to strategize with. So, how can we use this information in any meaningful way? Well, we can draw some conclusions by looking at a hypothetical example.

THE USE CASE

Let's say, for example, you use 1-day click as your attribution window on an ad set. This would mean that you want the ad set to consider purchases only within a 1-day window of clicking an ad as a successful conversion. This successful conversion is fed back to the algo as a signal for who should be targeted, but anyone outside that 1-day window is now removed. Interesting, right?

What is the implication of this? Well, let's say you're running a flash sale. Using 1-day click as your attribution window can, hypothetically, tell the algo that you only want to target the impulse purchasers within your audience segment determined in the ad set. Essentially, it would only see someone who purchased within 1-day as a successful conversion and therefore optimizes primarily to people with that same behaviour.

This can be quite powerful, as it may help you refine your spend towards the lowest-hanging fruit within your audience for a shorter time-frame promotion.

THE QUESTION

The above example does raise a good question, though: what about everyone else? Meta is constantly preaching to 'go broad' because it needs more data. Why would you intentionally leave out a group of similar customers just because their journey was a bit different, effectively limiting the amount of data the algo has to lean on? This is a good question, and unfortunately, I don't have a clear answer to that.

Even though you may have selected 1-day click in the above example, the algo still considers other signals as usual, such as all purchases on your pixel, but balances these differently. In which way it's balanced, exactly, I'm not sure, but the point remains that the data is being limited in some way.

THE TAKEAWAY

With the above in mind, I suppose the rest is up to you. It seems like you effectively have two options: limit the data your ad set has while targeting super high-intent customers while excluding everyone else, or open your targeting up to a slightly lower-intent audience while feeding more data back to your ad set.

It's clear that to decide which to use, you need to have some sort of strategy in place. After speaking to some friends in the industry utilizing this tactic, it seems like one of two paths is best depending on your situation:

• 1-day click: This is best if you're running a promotion and want to unlock impulse purchasers quickly over a short period.

• 7-day click: This is best for most other cases – foregoing impulse purchasers exclusively while opting for more data and quicker learning.

So why is 7-day click best in all cases except for those short promotional periods? My best guess is data – the less you restrict positive signals, the more data the algo has to optimize from, and the faster your ad set can learn.

So, there you have it. Hopefully, micro-segmenting your audiences using ad set-level attribution windows wasn't just news to me, and you were able to learn something from this! Also, if you have more info on this, I'd love if you shared it – I am still learning how exactly this mechanic works, so the more info we have, the better!