What is marketing attribution?

Every model is a rule about splitting credit, and every rule is a simplification of a decision nobody actually observed.

First touch credits what introduced you. Last touch credits what closed. Linear splits evenly, and time-decay favors the recent.

The point is not to find the true answer. It is to pick one rule, apply it consistently, and compare periods against themselves.

Phone calls are where most small-business attribution simply stops. A call that converted without leaving a trail is invisible to the model.

Say a kitchen remodeler wins an $18,000 job. Tracing back, the homeowner clicked a Facebook ad in March and found a blog post through a Google search in April. In May she clicked a search ad on the company's name and called.

First touch gives all $18,000 to Facebook. Last touch gives all of it to the brand search ad. Linear gives each of the three $6,000. A time-decay rule might give the search ad $9,000, the blog post $6,000 and Facebook $3,000.

Every one of those reports is correct by its own rule. The trouble starts when the Facebook report uses one rule and the Google report another, and both get pasted into the same budget meeting. Add up what each platform claims and this single kitchen shows as $36,000 of revenue.

Missing from all four is the neighbor who recommended the company in February. No model can split credit with a touch it never saw.

Attribution models were built for companies with thousands of conversions. With 12 sales a month, one job moving from one column to another swings a channel by 8%, and the models mostly report noise.

At that size a simpler system works better. Ask every new caller how they heard about you, and log the answer beside the tracked source. Keep a plain tally by month.

The interesting rows are the disagreements. Tracking says brand search and the caller says my neighbor told me? That's a referral, and brand search was only the doorway. Tracking says Facebook and the caller says I saw your truck? Both are probably true.

Review it quarterly, when you have 35 or 40 sales to look at. Make one budget change at a time and give it a full quarter. Changing three things at once leaves you unable to say which one moved the number.

Calls stop being a blind spot once each one produces a record. Name, number, reason and time are enough to join to the rest of your data.

Where the caller mentions the source, the transcript captures it in their own words rather than as a dropdown choice.

Pushing that into an analytics or CRM system depends on the integration, and our matrix says which ones write natively and which route through Zapier.