Why platform ROAS lies, and what to use instead

This one is less a lever than a piece of knowledge, but it creates enormous clarity about how to run a business. Platform-level attribution data is nice to look at. You cannot take it seriously.

It was already murky in 2016

In 2016, spending over a million dollars a year in ad budget across Facebook, Instagram and Google, the question was simple: how does the dollar we spend come back to us? The answer was already murky, and attribution windows were the reason.

Platforms take two views: a view-through window and a click-through window. Back then the default was 30 days for both. That means someone could scroll past a Facebook ad without interacting with it at all, buy 28 days later, and Facebook would claim the sale — pulling reported cost per acquisition down and platform ROAS up.

The same buyer also saw an ad on Google in that window, so Google claimed it too. And Pinterest. Every platform had its pixel on the site, every platform saw the purchase, and every platform took credit — which drives reported acquisition costs to a number that bears no resemblance to reality.

Then iOS 14.5 removed what was left

Around 2020, Apple shipped iOS 14.5 and blocked the pixels. Roughly 80% of website traffic is generally iPhone. That pixel traffic no longer reaches Meta, Google, Pinterest or TikTok, so the platforms mostly do not know the purchase happened.

What lands in your ad account is therefore best-guess data. Use it as a baseline for direction only: if platform ROAS in Meta goes from 1.5 to 2 over a month, read that loosely as a 33% improvement. If it goes from 1.5 to 1, read it as a 33% decline. Do not set ad budget off the absolute number.

The metric to use instead

Total revenue divided by ad spend — global return on ad spend, or what the industry now calls marketing efficiency ratio. Watching that change over time is a much better and far less complex answer to the real question: is the ad budget adding revenue to the business, or taking it away?

It is worth splitting by channel mix. At BottleKeeper, wholesale mattered: people who saw the product on a shelf often already knew what it was because of ad spend on Facebook and Instagram, so a global figure captured value that a direct-to-consumer figure alone would miss.

Alongside it, track direct-to-consumer revenue — including Amazon — divided by ad spend. That is a much tighter loop, because online acquisition spend shows up in direct-to-consumer revenue more immediately. Include Amazon rather than treating it as separate: some percentage of customers will see your ad and simply go and buy it there.

One clarification: click-through rates and impressions are fine to trust. Those happen inside the platform, so it knows exactly what occurred. It is the metrics requiring information to travel back from your website — cost per acquisition, ROAS, customer acquisition cost — that became unreliable.