There is a familiar assumption embedded in most marketing decision-making: uncertainty is a symptom of insufficient data, and the remedy is more of it. HMG has examined elsewhere what happens when the data genuinely runs out before a decision must be made. This piece concerns a different condition — one where the data has not run out at all. It has arrived from multiple credible sources, in sufficient volume to matter, and it disagrees with itself.

Marketing measurement is unusually prone to this condition because it rarely comes from a single instrument. A single campaign decision might be informed by platform-reported performance, an independent attribution model, a customer survey, and a sales team’s qualitative read of deal quality — four sources, each methodologically legitimate, each measuring something adjacent but not identical to what the others measure. It is entirely possible, and not even unusual, for platform data to show strong performance while attribution modeling shows the campaign contributing little incremental value, while customers report that the messaging didn’t register at all.

None of these signals is necessarily wrong. Platforms measure engagement within their own ecosystems and may apply attribution rules that assign more credit than an independent method would. Attribution models measure incremental contribution against a counterfactual that is itself an estimate, not an observation. Customer research may capture conscious recall or stated perception while missing other forms of influence that are not directly reported. Each source is measuring a real thing. The three real things simply do not agree.

The instinct in this situation is often to seek a tie-breaker — more research, a longer time horizon, a more sophisticated model that will finally reconcile the conflicting signals into one clean answer. Sometimes this works. In other cases, it does not, because the disagreement may not be a symptom of insufficient rigor in any one measurement — it is a structural consequence of measuring a complex system through several instruments that were never built to agree with each other in the first place. A more expensive attribution model will not make platform data stop measuring what platform data measures. Additional customer research will not make an attribution model’s counterfactual assumption disappear.

This distinguishes evidence conflict from evidence scarcity in a way that matters for how leadership should respond. Scarcity is addressed by acquiring more signal. Conflict is not reliably addressed by acquiring more signal, because additional signal frequently joins one side of an existing disagreement rather than resolving it — a new data source rarely proves that platform reporting, attribution modeling, and customer research were all secretly measuring the same thing all along.

What this leaves for leadership is a judgment that has nothing to do with data volume: a decision about which signal to weight more heavily given what the decision is actually for. A campaign decision oriented around immediate revenue attribution might reasonably weight the attribution model more heavily, accepting that it may understate soft brand contribution. A campaign decision oriented around long-term category presence might reasonably weight qualitative recall more heavily, accepting that it understates precise incremental lift. Neither weighting is dictated by the data. Both are legitimate. The evidence does not choose between them — the purpose of the decision does.

This is a narrower and more specific claim than “trust your gut when the numbers disagree.” The claim is that when credible sources conflict, the resolution may not be a hidden fifth data source waiting to be found, but a judgment about which kind of evidence is actually relevant to the decision being made — a judgment that has to be made deliberately, because the data, however abundant, will not make it by default.