Hacker News Reader

Comment by hn_throwaway_99 | Hacker News Reader

> This single comment contains many of the contrarian view against HN. It is nice we have these real world stories on HN to balance the ideological fight against ads, where All Ads are evil.

I still don't understand your point. The only reason Facebook's ability to properly attribute went down drastically is specifically because of the iOS privacy changes that require user opt in to track across different sites.

Fine, I don't doubt there were other HN commenters who were arguing different meanings of "tracking", but that's all a moot point. In this instance, we know exactly why Facebook's ad effectiveness went down - it went down because they could no longer track all your interactions across the majority of the web. Tough shit.

View on HN

Why does ability to attribute an ad depend on the ability to track individuals?

IIUC, you can do attribution with a very basic ad: <a href="website.com/ad-34" ping="fb.com/my-product/ad-34/ping"><img src="fb.com/my-product/ad-34/img.jpeg" /> Come buy my product, you'll love it!</a>

Then you sum up all the "/ad-34" hits, and figure out how many of those user sessions (which are now on your own site -- single domain, but perhaps leveraging a script supplied by the ad network) went and actually bought something. The ad network can correlate those sessions with the "ping" it receives to determine clicks vs. conversions.

Not all purchases come immediately after clicking through an ad. They may come days later, even. Knowing that "this user who purchased this product saw this version of the ad" and "this user saw the other version" is what attribution is.

You do not understand correctly.

Facebook (and other ad networks) want credit if someone clicks on your ad, spends time on your site, but then doesn't actually buy anything until a couple days when they enter the URL directly (i.e. not clicking on an ad). This is the problem generally referred to as attribution. Facebook (and others) do this by tracking everyone all around the web. They track the ad click and then they track you going back to the website the next day, and they know it's the same person.

Ad attribution directly depends on pervasive tracking.

You can do that with first-party cookies though.

Allow some facebook supplied javascript to set a first party cookie when the user hits website.com/ad-34, and then reference that 1st party cookie the next day when the user returns. Phone home when they buy something. Facebook can now correlate (1) the ping they received to begin with, (2) the user session cookie they initiated on the first visit, and (3) the user session they observed when a purchase was made.

I'm sure things get easier with 3rd party tracking, but fundamentally you can do it without cross-site tracking.

Out of curiosity, does attribution also work if the ad is not clicked?

For example, I bought a product the other day after seeing a Facebook ad, but I didn't click on the ad itself and rather went searching for it on DuckDuckGo, then after reading a bit about it online, bought the product.

I'm curious as to whether that got attributed to the ad impression or not.

Modern models do try to take views into account. (Also cross-device views and DOOH - which again rely extensively on tracking.)

Reply on news.ycombinator.com