There's a meeting that happens at almost every Series A and B company, usually about six weeks after a board deck lands badly. Someone pulls up the CAC trend. It's going the wrong way. And within twenty minutes, the conversation has become a debate about the attribution model.
Last-touch is undercounting the top of funnel. The data-driven model is over-crediting branded search. Nobody trusts the self-reported field because half the reps skip it. So the team commits to rebuilding attribution, again, and the quarter gets spent on plumbing.
This is a rational thing to do. It's also, very often, the wrong thing to do, because it treats a measurement problem as the cause of a performance problem. Attribution is exceptionally good at telling you how to distribute a message. It cannot tell you whether the message is worth distributing.
Attribution answers a narrower question than you think
Strip away the vendor language and every attribution model is doing the same thing: dividing credit for outcomes that already happened across the touchpoints that preceded them. That's genuinely useful. It's how you decide whether to move $30K from paid social to paid search, whether to keep the review-site listing, whether the retargeting window should be 14 days or 30.
What it does not do, and what no model does, is tell you why a buyer chose you, or why the ones who didn't, didn't. It measures the allocation of demand. It is silent on the creation of it.
That distinction stops being academic the moment your growth curve flattens. If your message reliably creates preference, better allocation compounds: you find the efficient channels, you push budget into them, and CAC improves. If your message doesn't create preference, better allocation just buys you more expensive indifference. You get very precise data about a set of channels that are all, quietly, doing the same disappointing thing. That is usually the real reason customer acquisition gets more expensive rather than less as a company scales.
And the tell is right there in the data, if you're looking for it.
Five signs the leak is upstream of your channels
1. Every channel underperforms by roughly the same amount. This is the big one. If paid search, paid social, outbound, and organic all degraded within two quarters of each other, the constant across those channels isn't the targeting or the bidding. It isn't even the creative, which is usually the biggest lever nobody is pulling. It's what you're saying. Channels fail for channel-specific reasons. Simultaneous, proportional decay across unrelated channels points to something they share.
2. Your best-performing campaign is branded search. Every growth lead knows this line item is flattering the blended number. The uncomfortable version of the question isn't "should we cut brand?" It's "why is the only campaign that works the one where the buyer already knows who we are and what we do?" This is worth pairing with a check on what AI assistants say about your business, since that is increasingly where the "already knows who we are" part gets decided.
3. Sales calls open with "so, what exactly do you do?" If prospects arrive at a demo needing the category explained, your marketing delivered a click, not an understanding. That's a messaging output, and no amount of touchpoint credit will surface it. It also tends to mean the go-to-market strategy was never written down in language a buyer would use.
4. Win rates swing hard by segment, but CAC doesn't. This means you're acquiring across segments at similar cost and converting them at wildly different rates. You have a positioning problem wearing an audience-targeting costume: the message resonates in one pocket of the market and you're paying full price to reach everyone else. This is what happens when the ideal customer profile was defined loosely or defined once and never revisited.
5. Your closed-lost reasons cluster around "no decision." Losing to a named competitor is a competitive problem. Losing to nothing (a deal that stalls, gets deprioritized, dies in procurement) is almost always a problem of urgency and framing. The buyer never understood why this, why you, why now. It's the same gap between lead generation and revenue growth that shows up on every pipeline review.
None of those five show up in an attribution dashboard. Four of them are actively obscured by one.
What this looks like in practice
A B2B workflow SaaS company, roughly $6M ARR, was spending about $85K a month across paid search, paid social, and retargeting. Over eighteen months they'd rebuilt attribution twice, moving from last-touch to a data-driven model, then layering in a self-reported field on the demo form. The instrumentation was legitimately good.
CAC payback still went from 14 months to 19.
The pattern that mattered wasn't in the dashboard. It was that every channel had softened together, at roughly the same rate. So instead of a third attribution project, we ran twelve interviews with recently closed-lost accounts. Nine of them described the product in terms of a category they'd already allocated budget to something else in. They weren't choosing a competitor over us. They were comparing us to a line item that already existed, and losing that comparison before a rep ever got on the phone.
The fix was a reframe: away from a feature-category description and toward the job the buyer's boss was already being measured on. Same channels. Same budget. Same attribution model, which, it turned out, had been reporting the situation accurately the whole time. Payback dropped to 11 months over the following two quarters.
The point of that story isn't that the interviews were magic. It's that the answer was never going to be in the model, because the model was only ever measuring which door people walked through, not whether they wanted to be in the building.
The diagnostic that costs less than a quarter of engineering
Before you approve another measurement project, run this. It takes about three weeks and no new tooling.
Interview ten recent closed-lost accounts and five closed-won. One question does most of the work: what were you going to do instead? Not "who else did you evaluate," which only surfaces competitors you already know about. "What else was on the table" surfaces the do-nothing option, the internal build, the adjacent budget line. That's where positioning gets decided, and it's the core of any useful market intelligence work.
Hold the channel constant and vary the claim. Same audience, same budget, same placement, three genuinely different framings of what you do, not three headline rewrites of the same idea. If one framing beats the others by a wide margin, you've learned something no attribution model could have told you. If they all perform identically and badly, you've learned something more important. This is also the cleanest argument for concentrating on one channel long enough to read the result instead of spreading thin across five.
Audit for "why now." Walk your website, your top three ads, and your first sales deck. Count how many make an explicit case for urgency that isn't a discount. In our experience the number is usually zero, and it correlates almost perfectly with a no-decision-heavy pipeline.
Then, and only then, go back to the model. Sometimes the instrumentation really is the problem. But it's a much cheaper problem to diagnose once you've ruled out the expensive one. If a stretch of underperforming spend is what triggered this, it's worth reading a paid ads setback as a growth opportunity rather than a media-buying failure.
Measurement is a mirror, not a lever
The reason attribution debates are so seductive is that they feel like progress and they're safe. Rebuilding a model is a bounded, technical project with a clear finish line, and nobody has to say out loud that the story the company has been telling the market for two years might not be landing.
Positioning work doesn't offer that. It's ambiguous, it's political, it usually implicates decisions made by people still in the room, and it can't be handed to a vendor with a scope of work. Often, the marketing problem isn't a marketing problem at all. That's also why this kind of diagnosis tends to sit with someone carrying senior marketing judgment rather than channel execution, whether that's an internal leader or an outside one.
If your channels are all sinking together, the dashboard isn't broken. It's working. It's telling you that the thing being distributed is the problem, and no model, however well built, will ever fix what it was designed to measure.
Looking for hands-on help? Explore our go-to-market strategy services.
