Why Full Funnel Measurement Breaks in Retail.

Q: Why Full Funnel Measurement Breaks in Retail.

Why full-funnel measurement breaks down in retail, and how last-click attribution, walled gardens, MMM and incrementality testing fit together to close the gap.

The short answer

Full-funnel measurement breaks down in retail because the methods brands rely on day to day, last-click attribution, walled-garden platform reporting, and pixel-based tracking, were each built to answer a narrow question, and each tends to overcredit the channels that capture existing demand while undercounting the channels that created it. Privacy changes have weakened the signal these methods depend on, and retail's multi-device, multi-marketplace shopping journeys widen the gap further. None of these methods are wrong on their own; they were simply never designed to see the whole picture, which is why leading teams now combine them with media mix modeling and incrementality testing rather than relying on any single source.

Why platform-reported conversions rarely add up to actual revenue

If you've ever added up what Meta, Google, and TikTok each claim in conversions and landed on a number well above your actual Shopify revenue, you've already seen this gap firsthand. Each platform measures conversions within its own attribution window and its own walls, which is genuinely useful for optimizing within that channel. What it isn't built to do is account for everything else a shopper saw before converting. A shopper who sees a TikTok ad, searches the brand on Google, then converts through a retargeting ad on Meta gets counted by all three, not because any platform is wrong, but because none of them were designed to see outside their own walls.

Summed platform-reported conversions in retail routinely land well above real sales once these overlapping claims are added together. This isn't new information to performance marketers, most of whom already treat in-platform numbers as directional rather than absolute. The gap is exactly why a number of teams are now pairing platform reporting with measurement that can see across the whole channel mix, rather than replacing platform data outright.

Why does last-click attribution undercount the channels that build demand?

Last-click attribution is a measurement method that assigns all the credit for a sale to the final touchpoint before purchase. It's simple to implement and easy to explain, which is exactly why it became the default starting point for most teams. Its gap is structural rather than a flaw in execution: branded search, retargeting, and email, the channels closest to the point of purchase, look highly efficient, while the channel that created the demand, a TikTok video, a YouTube prospecting ad, an influencer post seen two weeks earlier, isn't designed to be credited at all under this method.

This matters because budget tends to follow whatever looks most efficient in the dashboard. When demand-capture channels are consistently credited with more of the outcome than they independently drove, the channels building a brand's pipeline of future buyers can get underfunded, even when they're doing real work further up the funnel. In initial analysis across the brands Fospha has measured, this undervaluation of upper-funnel, demand-generation channels has averaged over 90% relative to their measured contribution, a gap large enough to meaningfully shape a year's media plan if left unaddressed.

Privacy changes removed the signal these systems depended on

Pixel-based and cookie-based tracking only work if you can follow a shopper across sessions and devices, and that ability has been steadily dismantled. On iOS, Apple's App Tracking Transparency framework requires opt-in for cross-app tracking, and a large share of users decline. Safari's Intelligent Tracking Prevention caps first-party cookies at seven days, or just 24 hours when a click identifier is present, so any shopper who takes more than a day to come back and buy shows up as "direct," with the channel that originally reached them erased from the record.

Server-side tracking helps recover some of this signal, but typically only within each platform's own walls. It improves how reliably a platform sends its own data; it doesn't extend visibility across platforms, which is a separate problem that requires a different kind of measurement to solve.

Retail's structure makes the gaps worse, not smaller

Retail eCommerce brands face a version of this problem that's harder than most categories deal with. Shoppers browse on one device and buy on another. They discover a brand on TikTok and complete the purchase on Amazon, where UTM parameters are stripped and DTC measurement tools record nothing once a shopper lands there. Promotional spikes like Black Friday compress weeks of consideration into days, distorting attribution models tuned on ordinary traffic patterns. Long consideration categories like beauty and fashion lose their earliest touchpoints to cookie expiry before the shopper ever converts.

Cross-channel halo effects add another layer. A streaming TV ad or influencer post can drive a direct search, a TikTok Shop purchase, or an Amazon sale that standard attribution credits to whichever channel happened to be closest to checkout, rather than the one that created the demand. In Fospha's own measurement work, a meaningful share of Amazon sales, in some retail categories over 40%, has shown measurable influence from media running entirely outside Amazon, representing demand that single-channel metrics structurally can't see.

How is MMM different from attribution if both claim to measure the funnel?

Media mix modeling (MMM) estimates how much each channel contributes to overall sales by analyzing aggregate spend and revenue patterns over time, rather than tracking individual user journeys. Where attribution follows a single shopper's path and assigns credit at the touchpoint level, MMM works from the top down, measuring a channel's total contribution to the business regardless of whether any single conversion event was ever observed.

This is precisely why MMM and attribution often disagree about the same channel, and why that disagreement is useful rather than a problem to be solved away. When attribution says a channel is underperforming but MMM shows it carrying real weight in the business, that gap is a signal worth testing, not proof that one method is simply wrong. MMM itself isn't a single definitive number either: like any statistical model, it carries its own margin of uncertainty, which is why the most reliable approach treats MMM, attribution, and incrementality testing as three complementary signals rather than searching for one model to settle the question outright.

What does triangulated measurement look like in practice?

Triangulation means treating MMM, attribution, and incrementality as three different instruments reading the same system, not three competitors producing one right answer. MMM shows the strategic weight a channel carries across the business. Attribution shows the tactical path a shopper took. Incrementality confirms what would or wouldn't have happened without the spend. Used together, a disagreement between methods becomes useful information rather than a contradiction to explain away: if attribution undercounts a channel that MMM shows carrying real weight, that's the specific gap worth testing next.

The practical difficulty has always been speed. Incrementality tests are rigorous but periodic, and traditional MMMs are strategically useful but typically run on a quarterly cycle, too slow to inform the budget decisions marketers are making every week. Closing that gap means having a measurement layer that updates daily, stays impression-led so demand-generation channels keep their credit between test cycles, and feeds incrementality results back in to keep sharpening the model over time.

Fospha's Core (Daily MMM - ad-level measurement, every channel) and Halo (Total Commerce Measurement - DTC, Amazon, and TikTok Shop) are built around that gap specifically, running daily rather than quarterly, and designed to work alongside incrementality testing rather than in place of it.

Common questions

Q: Should I stop using attribution data entirely?

No. Attribution remains useful for fast, tactical, day-to-day decisions like creative testing or bid adjustments within a single channel. The gap is treating it as a complete picture of channel value, when its design around the last touchpoint means it was never meant to answer that broader question on its own.

Q: Why do Meta, Google, and TikTok all show different numbers for the same campaign?

Each platform's reporting tracks conversions it can observe within its own walls and attribution window, which is useful for in-platform optimization but doesn't account for what happened on other channels beforehand. A shopper exposed to several channels before buying can get counted by more than one platform, which is why summed platform totals routinely run ahead of actual sales.

Q: Does this mean I should stop running incrementality tests?

No, the opposite. Incrementality tests remain the most reliable way to prove a channel's causal impact, and they're most valuable when paired with always-on measurement that fills the gaps between test periods. Fospha is built to support this combination, not replace it.

Q: What's the difference between impression-led and click-based measurement?

Click-based measurement only credits channels when a shopper actively clicks an ad, which structurally excludes channels like video and social that drive awareness without always generating a click. Impression-led measurement accounts for the impact of being seen, not just clicked, giving demand-generation channels credit for the role they actually play.