Last year I ran a geo holdout test for a brand that swore Meta prospecting was their best channel. Their dashboard said so. Last-click had Meta responsible for 41% of revenue.
We split the country into two matched groups by past sales volume. Paused all Meta prospecting in one group for three weeks. Left everything else running exactly as normal in both groups.
Revenue in the paused regions dropped 4%. Not 41%. Four.
Most of the sales Meta was "generating" were people who'd already decided to buy — through word of mouth, an Instagram post they saw organically, a friend's recommendation — and Meta's ad happened to be the last thing they clicked before checkout. The platform got full credit for a decision it didn't influence.
That gap between what the dashboard says and what actually happened is the single most expensive blind spot I see in growth marketing. Bigger than any landing page mistake, any wasted ad set, any missed email flow from the last six posts in this series.
Why Last-Click Lies
Last-click attribution gives 100% of the credit to whatever channel someone clicked right before converting. It's the default in nearly every ad platform's own reporting, for one obvious reason: it makes that platform look good.
Meta's own dashboard uses a 7-day click, 1-day view attribution window by default, meaning it claims credit for anyone who so much as saw an ad and bought within a day, even without clicking. Google Ads does something structurally similar. Every platform is grading its own homework, and every platform's homework comes back with an A.
The real customer journey almost never looks like one clean click-to-purchase line. Someone sees a Meta ad, ignores it, searches your brand name on Google two days later, reads a review site, gets a retargeting ad, then finally buys through an email link. Six touchpoints. Last-click gives 100% of the credit to email and 0% to everything that actually built the intent.
The Attribution Methods, Ranked by How Much I Trust Them
What most brands are running their entire budget on, often without realizing it. Fast, free, built into every ad manager. Also the most biased number in your entire marketing stack, because every platform inflates its own contribution using generous attribution windows nobody agreed to.
Spreads credit across every touchpoint in the journey instead of dumping it all on the last click. Linear splits credit evenly, time-decay weights recent touches more, U-shaped weights the first and last touch most heavily. GA4, Triple Whale, and Northbeam all offer some version of this.
Better than last-click, but still a modeled guess about influence, not proof of it. It answers "what did the customer touch" — it doesn't answer "what would have happened if that touch never existed."
The only method that actually answers the real question: what happens if this channel disappears? Geo holdout tests split your market into matched regions, turn a channel off in one, and measure the real revenue difference. Meta and Google both offer built-in conversion lift studies that do a lighter version of the same thing.
This is the closest thing to ground truth in marketing measurement. It's also the method almost nobody runs, because it requires actually turning off a channel you're nervous about turning off, and waiting three to four weeks to see what happens.
You Don't Need a Data Science Team to Start
Full media mix modeling and formal incrementality testing at scale genuinely does need statistical rigor most small teams don't have in-house. But a basic geo or spend-pause test doesn't.
Pick your most-trusted channel by last-click. Pause it entirely for two to three weeks — either in half your regions, or account-wide if your traffic is small enough that regional splitting isn't practical. Watch total revenue, not just that channel's reported conversions. If revenue barely moves, you've just found a channel getting credit it didn't earn. If revenue craters, you've confirmed it's doing real work.
Three weeks of mild discomfort is a small price for knowing where your next lakh in budget actually belongs.
What I Actually Recommend for Most Accounts
Use multi-touch data (GA4 is free and good enough for most accounts under enterprise scale) for weekly decisions — where to shift small amounts of budget, which creative angle is pulling its weight through the journey. Run an incrementality test quarterly on your top-spending channel to sanity-check whether the multi-touch numbers and the platform-reported numbers are anywhere close to reality.
If they consistently diverge by more than 20–30%, trust the incrementality test and adjust your mental model of that channel permanently, not just for that quarter.
✅ You know which attribution window each ad platform is using by default
✅ You're not making budget decisions above a few thousand rupees on platform-reported last-click alone
✅ GA4 or a multi-touch model is set up and checked at least monthly
✅ You've run at least one geo holdout or spend-pause test in the last 12 months
✅ You know the real gap between platform-reported and incremental performance for your top channel
Pick your most-trusted channel this month and run the pause test. Three weeks. Watch total revenue, not the dashboard for that one channel. You'll either confirm what you believed or find out you've been funding the wrong thing.
Next up in this series, I'm going organic — SEO and content distribution, since this series has been entirely paid-channel so far and that's only half the picture for most brands.
See you there.
— Suraj
Frequently Asked Questions
Because both platforms use generous default attribution windows — Meta's default is 7-day click and 1-day view — and both report on their own dashboard using last-click logic. A platform grading its own contribution will almost always grade itself favorably.
A measurement method where you split your market into two matched regions, pause a specific channel in one region while running everything as normal in the other, then compare actual revenue between the two. It measures real incremental impact rather than modeled or platform-reported credit.
GA4 is free and sufficient for most brands under enterprise scale, especially for understanding the general shape of the customer journey. Triple Whale, Northbeam, and similar tools add value mainly at higher spend levels where the cost of a wrong budget call justifies the subscription.
At minimum once a year on your top-spending channel, and again any time you significantly increase budget on a channel or launch a new campaign type like Performance Max or Advantage+ that could be reallocating existing conversions rather than creating new ones.
Comments
Post a Comment