Hi there,

If a shopper saw your ad on Instagram, then Google, then bought the product in a Target store, which platform gets credit for that sale?

I've been in three calls this week where a smart, sophisticated marketer hit the exact same wall.

One asked me, “Are we double counting our conversions across platforms if that purchase falls inside more than one attribution window?” Right now, most brands don't have an answer.

What "closing the attribution loop" actually means

Every brand has the outbound half of this figured out: you know exactly what you spent, where, and against which audience. The half that's usually missing is the return half, the part where an actual purchase gets matched back to a real person and reported to the platform that served the ad.

Without that return half, you're not doing attribution. You're watching total sales move, watching spend move, and telling yourself a story about how the two are connected. Closed-loop attribution replaces the story with a fact: this specific person bought this specific product, and here's the campaign that reached them. One is estimated. The other is confirmed.

Why throwing smarter models at it doesn't fix it

For years, the industry's answer to this gap has been better modeling. Multi-touch attribution. Marketing mix modeling. Incrementality testing. All three are legitimate tools, and I'd never tell a brand to drop them. But to close the loop you have to add to them.

You can't model your way to a fact you don't have. MTA needs a journey to divide credit across, and a shopper walking into a store leaves no trail behind. I talked to a retail-only food brand this week selling in about 21,000 doors. No DTC site, no pixel, nothing to model against. Their words to me: "We haven't had an attribution conversation at all," because there was never the right information to build one around.

MMM can handle offline sales in aggregate, which is genuinely useful for board-level budget calls, but it can't tell an ad platform's bidding algorithm that one specific person converted this morning. Incrementality testing gets closest to real proof, but it's periodic and expensive, not something running underneath every campaign in real time.

That's exactly why the double counting question from this week doesn't have a clean answer yet. Meta, Google, and TikTok are each doing their best modeling on the same incomplete picture. Of course they all think they deserve some of the credit. Nobody's lying. Its just that nobody has the missing piece.

US advertisers will pour $100.86 billion into Meta ads alone this year, enough to put it ahead of Google in the US for the first time ever [1]. That's a massive amount of budget riding on numbers nobody can fully trust. One executive covering this space called offsite attribution "2027's problem" [2]. Which is a polite way of saying: we know this is broken, and we're choosing not to fix it yet.

I get why. Fixing it feels hard. But the brands that wait on this are the ones still making budget calls off vibes next year, while everyone else is making them off receipts.

What changes once you close the loop

Once a real purchase gets matched back to a real campaign, a few things happen quickly.

Reported performance stops undercounting revenue that was already there, so your ROAS numbers finally reflect sales that were always happening.

Slower, but bigger: the platforms themselves get smarter, because their bidding and audience models are learning from your whole buyer base instead of the DTC sliver. And the thing every brand assumes but almost nobody can prove, whether a Meta or TikTok impression actually nudges someone into a store, stops being a matter of faith and starts being something you can see.

When a shopper discovers a product online, 40% of them go visit a physical store before they ever buy it [3]. This is the reality of omnichannel shopping. And for most brands, that entire path from discovery to store to purchase is invisible.

And once you close the loop, the sale isn't the finish line. That verified buyer becomes someone you can retarget for the next purchase, a real seed for a lookalike audience instead of a guess based on site visitors, someone you can suppress from prospecting spend because you already know they're a customer.

Most brands are still spending millions based on metrics that can’t tell them the thing they actually want to know.

The loop is closed when the purchase record completes the circuit rather than being inferred. Get that, and every measurement approach you already run gets better.

Talk soon,

Kait

Connect with me on LinkedIn