> What does it mean when attribution is a mess? Attribution means being able to "attribute" a user and the conversions (eg signups, purchases, etc) they do bac

> What does it mean when attribution is a mess? Attribution means being able to "attribute" a user and the conversions (eg signups, purchases, etc) they do back to your ad spend. So typically, advertisers want to be able to know that e.g. this ad campaign spent $100, led to 5 clicks, which led to 2 sign-ups and 1 purchase. Before the iOS privacy changes, Facebook could piece all those things together even though the ad might be running either on Facebook or another app, and the conversions are happening in your ad. Aka they could "attribute" the events back to the ad. With the new changes, Facebook can't put all those pieces together as easily, so they can either only do it on an aggregated basis (ie not user-by-user), and/or use probabilistic "modeling" to do attribution. This is good for users, since they are less likely to be tracked, but bad for advertisers who now cannot easily know how effective their ad spend is. > ... efficiency fluctuating Let's say one day you spend $1000 and get 100 purchases, that's $10 / purchase. A purchase nets $15 for you, so that is good ad spend. The next day, you spend $1000 and only get 50 purchases, so it's $20 / purchase and you are losing money on your ad spend. These numbers fluctuate a lot more wildly than they did in the past, and it seems to happen randomly, making it hard to commit to spending a certain budget. This is probably related to attribution, because Facebook uses a lot of Machine Learning to predict who to show an ad to, where to show it, etc based on past conversions. But if they can't really attribute past conversions, it's hard for their Machine Learning model to consistently get results.