You ran a YouTube campaign last quarter. The click-through rates were modest and when finance pulled the ROAS number from GA4, it looked like one of the weakest performers in the mix. The team cut budget on renewal.
Three months later, branded search volume was down. Amazon sales softened. New customer acquisition costs crept up. Standard attribution had no way to connect them, the signal existed, but not in the same place as the decision.
This is the structural problem with measuring upper funnel activity through attribution. And in omnichannel retail - where a customer might see your Meta ad, search your brand on Google, and then buy on Amazon or TikTok Shop - it is not a small rounding error. It is a structural gap in what standard measurement was built to capture.
Why do upper funnel ads tend to look underperforming in standard attribution?
Standard attribution tools assign credit to the touchpoints they can observe. For most brands, that means clicks that happen on or near the website. Upper funnel channels like paid social video, YouTube, and display are impression-led: they plant the intent, they do not capture it. In a last-click model, a customer who saw your TikTok ad twice, clicked a Google Shopping result, and converted gets the conversion logged under Google. TikTok gets nothing. Even data-driven attribution models, which distribute some credit across the click path, remain blind to impression-only exposure and to any conversion that happens outside your own website.
The deeper problem is that impression-led channels have lag effects. Demand generated by a brand awareness campaign in week one might convert in week three, after the customer has done research, compared options, and decided. Depending on purchase cycle and category, the lag between an awareness exposure and a conversion can extend well beyond standard attribution windows - meaning the channel that generated intent may not appear in the conversion path at all. The channel that did the heavy lifting gets erased from the record.
Standard attribution tools also have a visibility ceiling: they only see what happens on your website. They cannot see organic search lifts, branded keyword volume increases, or purchases that happen on Amazon or TikTok Shop - even when your paid media drove them there.
What does "omnichannel" do to the measurement gap?
It makes it significantly larger.
In a pure DTC model, attribution is imperfect but at least confined to one surface. In omnichannel retail - where customers buy across your website, Amazon, TikTok Shop, and increasingly through CTV and retail media - your measurement gap is not just about channel credit. It is about sales destinations that pixel-based tools were never designed to reach.
Consider the cross-channel halo effect. A Meta campaign drives brand awareness. Some of that demand converts on your website. But a meaningful portion - the portion that cannot be measured by your pixel - goes to Amazon, where the customer searches your brand and buys. Your Meta ROAS calculation only sees the DTC conversions. The Amazon lift is invisible.
Nécessaire experienced exactly this. As a brand selling across DTC, Amazon, and Sephora, their team had a strong intuition that TikTok and Meta were driving demand well beyond their own site, but click-based attribution could not quantify it. By adopting Fospha's Total Commerce ROAS to capture performance across every sales destination in a single metric, they scaled TikTok investment with conviction and increased upper-funnel spend ahead of BFCM 2025. The results: +61% YoY growth in paid media revenue, +25% uplift in TikTok Total Commerce ROAS, and +58% uplift in Amazon Sponsored Products Total Commerce ROAS - with TikTok's share of revenue reaching 16% against a cohort average of 9%. The demand was already being created. The measurement just had not been able to see it.
When measurement is scoped to your website, it captures only part of the picture - the demand that flows through other sales destinations remains unquantified, not absent.
Why does this bias your budget toward the bottom of the funnel?
When no one trusts the numbers for upper funnel, teams default to what looks safest. Demand capture channels have short attribution windows, high click volume, and strong reported ROAS. They look efficient because they are measured by tools built around click-path data - which systematically captures demand capture activity and leaves demand creation activity underrepresented.
Demand creation channels have diffuse attribution, impression-based delivery, and lag effects that standard tools cannot account for. They look expensive and inefficient for the same structural reason: the tool cannot see what they do.
Every well-run performance team we have spoken to knows intuitively that their upper funnel is doing more than the numbers suggest. The problem is they cannot prove it, not in a way that survives a finance meeting, anyway.
The practical consequence is systematic under-investment in the channels that create new demand. CAC rises. The brand becomes increasingly dependent on retargeting the same pool of warm users. Growth plateaus. And at some point, someone turns off a brand awareness campaign to save budget, and branded search volume quietly declines three months later.
How do you measure what upper funnel ads are doing?
The missing layer is measurement that captures what happens before the click and beyond your website - the impression-driven demand that standard tools leave unquantified.
The methodological answer is Media Mix Modeling (MMM): a statistical approach that looks at the relationship between marketing inputs and business outcomes at an aggregate level, without relying on cookies, pixels, or click tracking. MMM gives full credit to impression-led channels because it observes their effect on revenue over time, not just on the click path.
The practical limitation of traditional MMM is speed. Quarterly reports describe a media environment that has already changed. By the time the model tells you YouTube was working, you have already reallocated the budget somewhere else.
The better version - daily-updated MMM at the ad level - gives you the strategic coverage of a full model with the operational speed your team needs. Brands running Fospha's always-on MMM can see, every day, how each channel is contributing across their DTC site, Amazon, and TikTok Shop - and act on it before the insight goes stale.
Underoutfit used this approach to scale YouTube spend by 315% and TikTok by 93% in a single month. Guided by Fospha's always-on measurement, the outcome was $3.3M in incremental revenue with blended CAC down 15% - and critically, the data showed exactly how much headroom remained to scale further.
What does this mean for how you run your budget?
Three practical implications:
1. Your upper funnel budget is almost certainly under-allocated relative to actual contribution. If your measurement cannot see halo effects, lag effects, or off-site sales, the credit for upper funnel activity has been silently redistributed to bottom-funnel channels for months. The first question to answer is not "should we invest more in YouTube?" - it is "what is YouTube contributing that our current tools cannot see?"
2. Your finance conversations will keep going badly until the measurement changes. Defending upper funnel spend with impression-based metrics against a finance team holding a GA4 ROAS number is an unwinnable argument because you are using different evidence bases. The only resolution is a measurement system that marketing and finance can both trust, built on methodology that does not systematically under-credit the channels you are trying to defend.
3. Omnichannel expansion without measurement reform is a trap. Every new sales destination you add creates another surface where your media is driving demand that your current tools cannot see. If you are already measurement-blind on your website, adding marketplaces makes the gap bigger, not smaller. The measurement architecture needs to expand as the commerce footprint does.
Frequently asked questions
Why does last-click attribution undervalue upper funnel channels specifically?
Last-click gives 100% of the conversion credit to the final tracked touchpoint before purchase. Upper funnel channels typically generate awareness and intent early in the journey, not the final click. Because these channels do not capture the click, they do not capture the credit, regardless of how much demand they created upstream.
Can I fix this with view-through attribution in Meta or Google?
Platform-reported view-through attribution is useful as a directional signal, but it is self-reported: each platform applies its own model and attribution window, making cross-channel comparisons unreliable. Platform attribution also cannot see conversions that happen outside their own ecosystem - so Amazon or TikTok Shop sales driven by a Meta campaign will not appear in Meta's reported numbers.
What is a cross-channel halo effect and why does it matter in omnichannel?
A halo effect occurs when media spend on one channel drives purchases on another. For example, a Meta campaign generates awareness that leads to branded search on Amazon, which converts to a sale. The Meta campaign gets no credit in standard attribution. In omnichannel retail, where a significant portion of conversions may happen off your website, halo effects can represent a material share of marketing's true contribution - and they are structurally invisible to pixel-based tools.
Is Media Mix Modeling the only solution?
MMM is well-suited to full-funnel, impression-led measurement across a broad channel mix, particularly where cookie or pixel dependency is a constraint. Incrementality testing offers stronger causal proof for specific channels or budget decisions, but is episodic by design. Attribution is fast and granular - valuable for in-platform diagnostics and short-cycle decisions - but it captures a click-path view that under-represents impression-led and off-site contribution. For most omnichannel brands, the approaches work best in combination: always-on MMM for continuous direction, incrementality tests for high-stakes validation, and attribution for day-to-day diagnostics.
What channels are most affected by this measurement gap?
Paid social video (Meta, TikTok), YouTube, CTV, and display are the most systematically undercredited. These are impression-led, upper-funnel channels with long lag effects and no reliable click path to conversion. They are also the channels most likely to drive cross-channel halo effects - particularly on Amazon and TikTok Shop - that pixel-based tools cannot see.
Does this mean I should stop tracking clicks and ROAS?
No. Click-based metrics and channel ROAS remain useful diagnostic signals, particularly for demand capture channels like branded search and Shopping. The problem is treating them as the primary or sole measure of marketing effectiveness. The goal is a measurement layer that integrates impression-led contribution, lag effects, and off-site sales - so click-based data sits within a complete picture rather than pretending to be one.
If your current measurement cannot explain what your upper funnel is doing across every channel and every sales destination, see how Fospha approaches full-funnel omnichannel measurement.
