Every few months, a marketing dashboard gets printed out and brought into a leadership meeting as evidence. Paid search drove this many leads. Organic drove that many. Someone makes a budget decision on the spot. Three months later, revenue hasn't moved the way the model predicted, and nobody can explain why.
The problem isn't that your attribution is set up wrong. It's that attribution was never built to answer the question you're asking it. It can tell you which touchpoints were recorded before a deal closed. It cannot tell you which touchpoints caused the deal. In a B2B or considered-purchase business with a multi-week sales cycle, those are very different things.
Why the Model Breaks in the Middle Market
Attribution tools were built around a clean assumption: a buyer clicks, lands, converts, and the chain is traceable. Most mid-market buying journeys violate that assumption from the first step.
Your buyer hears your name on a podcast, searches it three weeks later, reads a comparison page on their phone, forwards it to a colleague, and that colleague fills out your form on a desktop from a different IP. Your analytics sees one direct visit and one form fill. The podcast gets nothing. Meanwhile, someone asks an AI assistant for vendor recommendations, gets your name, and arrives via a branded search. Branded search takes the credit for demand it didn't create.
Layer on privacy changes, cookie restrictions, ad blockers, and the simple reality that a lot of influence happens in conversations you'll never instrument — and your model isn't measuring the market. It's measuring the slice of the market your tracking can see. That slice is systematically biased toward the last, cheapest, most trackable touch.
The Question Behind the Question
Before you tune anything, decide what decision you're actually trying to make. Attribution questions usually collapse into one of three:
- Should I spend more or less here? This is a causal question. Attribution dashboards are weak at it. Tests are strong at it.
- Where is my pipeline coming from right now? This is a directional question. Attribution plus self-reported data handles it well.
- Which content and pages do buyers actually engage before they convert? This is a behavioral question, and your analytics is genuinely good at it.
Most arguments about attribution happen because two people are asking different questions and both pointing at the same chart.
Three Things That Work Better Than a Better Model
Ask the buyer. Add an open-text "How did you hear about us?" field to your forms and have sales ask it in discovery. Self-reported attribution is messy and it's also the only data source that can see the podcast, the referral, and the conversation at a conference. Don't make it a dropdown — dropdowns force buyers into the channels you already know about. Review the free-text answers monthly and look for patterns your tracking never surfaced.
Run holdouts, not just reports. If you want to know whether a channel is doing work, turn it off in one region or for one segment and watch what happens to pipeline. Geographic holdouts are the most practical version for most companies. The result is uncomfortable to collect and far more trustworthy than any model-assigned credit.
Measure revenue, not leads. A channel that produces a high volume of cheap leads and a channel that produces fewer leads that close at twice the rate will look very different depending on where you stop counting. That means your CRM has to carry source data all the way through to closed-won, and the stage definitions have to be consistent across teams. This is usually an integration and data-hygiene problem long before it's a marketing problem.
The Plumbing Decides the Insight
Here's what most attribution conversations skip: the model is only as good as the handoff between your website, your marketing platform, and your CRM. If form submissions arrive without UTM parameters, if sales edits the lead source field by hand, if your CRM and your ad platform define a "qualified lead" differently, no amount of modeling fixes it. You're not looking at attribution. You're looking at data entry.
The fix is unglamorous and it's where the real returns are. Standardize source capture at the form level. Pass it through to the CRM automatically so no human has to retype it. Lock down the fields that matter. Write down the definition of each pipeline stage and make the system enforce it. When the plumbing is clean, even a simple first-touch and last-touch comparison becomes genuinely useful — and the gap between the two tells you something real about how long your buying cycle is.
Use It as a Compass, Not a Verdict
Treat attribution as one instrument on the dashboard, not the whole dashboard. Use it to generate hypotheses, then test the ones that carry real budget. Pair it with self-reported data to catch what tracking can't see. And make sure the underlying systems — site, forms, marketing platform, CRM — are actually talking to each other before you argue about what the numbers mean.
If your marketing reporting and your CRM tell two different stories about where revenue comes from, that's a systems problem we work on every week. Reach out and we'll walk through what your current setup can and can't tell you.
