Attribution doesn’t fail loudly. It just keeps confidently crediting the wrong channels until budget quietly flows to the ones that look good rather than the ones that work. Here are six signs yours is misleading you — and every one of them costs real money, because attribution decides where spend goes.
1. One channel wins suspiciously often
If a single channel takes credit for nearly everything, check your model before you celebrate. Last-touch attribution systematically over-credits whatever comes right before the form fill — often branded search or direct — while starving the channels that actually started the deals.
2. A big share of revenue is “unknown” or “direct”
When a large chunk of closed pipeline traces back to no known source, your data is broken somewhere in the middle, and the model is dividing credit across a picture that’s already wrong. Confident output on incomplete input is worse than no output.
3. The model and your gut violently disagree
If attribution says a channel does nothing but your sales team keeps hearing prospects mention it, trust the pattern and interrogate the model. Attribution frequently misses influence that doesn’t show up as a clean click — dark social, word of mouth, the podcast nobody fills a form after.
4. You’ve never questioned which model you use
If nobody can say whether you run first-touch, last-touch, or multi-touch — and why — you’re acting on a default someone set once. Each model answers a different question, and using the wrong one for your decision is a silent, standing error.
5. Budget shifts never change results
If you keep reallocating spend based on attribution and the pipeline never responds the way the model predicted, the model isn’t describing reality. That’s the clearest possible sign it’s misleading you.
6. The numbers change when nobody changed the strategy
If channel credit swings month to month without a corresponding change in what you did, your attribution is unstable — usually a data or tracking problem — and unstable attribution is untrustworthy attribution.
What to do about it
None of these mean abandoning attribution. They mean treating it as one input, knowing which question your model answers, and fixing the source data underneath before trusting the chart. Attribution is a useful lens and a terrible oracle.
Related reading
See what marketing attribution in B2B is and what closed-loop reporting is.
Suspect yours is off? Book a health check.