Marketing Attribution Models, Explained With One Customer Journey

Why do two reports credit different channels for the same sale?

Your SEO team says search introduced a buyer. Your email team says a newsletter brought them back. Your ad team points to the final click. All three interactions might be recorded in the same customer journey. The disagreement starts when each report answers a different question about who should receive credit.

A touch is a recorded interaction in the journey, not a complete account of everything someone saw or thought. A model gives those touches weights. A weight of one means all the credit; a weight of one-third means a third. The calculation can be simple while the business decision remains uncertain.

The three rules below match Attriboo's First-Touch, Last-Touch, and Linear controls. Other tools can define eligible interactions differently. For example, Google's current Analytics attribution documentation describes different available models and special treatment of Direct. Check the actual product's rules before comparing labels across dashboards.

What happens to one $300 sale under three models?

Use one example journey with three distinct recorded touches. On Monday, someone finds a product guide through organic search. On Wednesday, they return through a newsletter. On Friday, they click a Google ad and make one $300 purchase. Assume all three touches belong to the same matched journey and precede the payment.

First-Touch highlights recorded discovery. Last-Touch highlights the final recorded interaction. Linear spreads the credit evenly across all three touches. Equal credit is a convention, not a finding that each touch was equally persuasive. Nothing about the purchase itself changed when you switched columns.

A channel can also appear more than once in a real path. Linear gives equal weight to recorded touches, then a source report can aggregate their credit. It does not necessarily give every unique channel an equal share. Keep that distinction in mind when a frequent returning source has several touches.

Try changing the model—and then remove a missing touch

Use the explorer to change the sale amount and credit rule. Then select the journey where organic search was missed. The sale still happened, but the model can only allocate credit across the remaining recorded touches. This is why more decimal places cannot solve missing measurement.

With the first touch missing, First-Touch now credits the newsletter. Last-Touch still credits Google Ads. Linear splits $300 into $150 for each remaining touch. The newsletter did not become the person's true first encounter just because the report lacks the earlier one. It became the first recorded touch.

For amounts that do not divide evenly, this educational tool distributes whole cents in journey order. A live report may keep fractional credit internally and round at display time. The point of the exercise is the allocation rule and unchanged total, rather than an accounting reconciliation.

Which attribution model should your team use?

Start with the decision you need to make. If you are reviewing how people first discover the product within recorded journeys, First-Touch is a useful view. If you are reviewing the last interaction before purchase, Last-Touch helps. If you want a simple view that shares credit across the observed path, inspect Linear.

1. Choose one primary reporting convention so this month's report can be compared with next month's.
2. Use a second model as a sensitivity check. Ask which channels change substantially and what recorded journeys explain the difference.
3. Keep the website, dates, revenue definition, currency, and tracking coverage consistent while comparing.
4. Record the model beside the result. Write revenue credited under First-Touch rather than revenue generated by SEO.
5. Investigate uncertainty before changing budget. A small sample or missing checkout reference can matter more than the model choice.

There is no universal winner. A discovery campaign can look weak under Last-Touch because another source appears closer to payment. First-Touch can hide useful later interactions. Linear is easy to explain but assumes equal touch weights. Each view describes the observed path through a different rule.

Avoid selecting whichever model makes your department look best each month. If you change the reporting convention, explain why, show the old and new view where possible, and mark the change in the report. Consistency makes a model useful for discussion even when it cannot provide a complete explanation.

How do you compare models in Attriboo?

Start in the public demo to learn the controls without connecting customer data. For your own website, install tracking and connect the supported payment provider first. A model switch does not create a visit record, import a missing payment, or repair a broken checkout handoff.

1. Choose one website and a reporting period. Read the date-range behavior before interpreting a journey that spans several periods.
2. Select First-Touch and review the recorded source credit. Note the model beside any exported figures.
3. Switch to Last-Touch, then Linear, keeping the other controls consistent. Compare source allocations rather than adding model totals together.
4. Review the recorded customer journey behind a surprising source result. Check for missing history or unmatched payments.
5. Use one finding to choose a practical investigation or test. Keep product costs and other expenses separate when evaluating profit.

Attriboo puts these model controls beside revenue-focused reporting, so you can ask a business question without building the first comparison from scratch. The answer still depends on recorded activity and supported matching. Check provider history, refunds, selected dates, and currency before treating a difference as a marketing insight.

What should a model comparison leave you cautious about?

A person might hear a podcast, ask a colleague, return on another device, or decline tracking. Some of those influences will be missing from the recorded journey. A bot crawling your page is also separate from a human referral or purchase. Do not insert every observed signal into a sales path just because it is available.

A spreadsheet can help explain a few known journeys: list the touches in order, apply the weights, and check each allocation sums to the sale total. An existing analytics model-comparison report may also answer your immediate question. A CRM can add context for offline or long sales processes that a website payment report does not cover.

If your real question is whether spending on a channel increased sales, you need a different kind of evidence. A well-designed holdout or incrementality test compares outcomes under different exposure conditions. It needs appropriate planning, sample size, and attention to confounding factors. Rule-based revenue credit alone cannot supply that counterfactual.

Choose a model your team can explain, document the coverage, and compare a second view when the decision matters. That turns attribution into a useful reporting conversation without asking it to reveal every reason someone bought.

Common questions