Marketing attribution you can defend in a budget meeting
First-touch, last-touch and multi-touch all lie in different ways. A practical attribution setup — captured sources, self-reported attribution, holdouts — that tells you where to put the next dollar without pretending to precision it does not have.
28 Oct 2025 · 2 min read · by HeiSyn
Attribution arguments happen because every model is a story, and the stories favour whoever built them. Last-touch credits the demo form; first-touch credits a blog post from a year ago; multi-touch spreads credit by a formula nobody can explain. None is true. What a budget meeting needs is a defensible, consistent view of which channels produce customers, and the honesty to say where the view is fuzzy.
Capture the sources you can
UTM parameters on every link you control, stored on the lead at first touch and never overwritten. Referrer and landing page. The campaign, ad and keyword where the platform provides it. The source on every CRM record, set automatically at creation, with 'unknown' as an honest value rather than a default of 'direct'.
Ask the customer
One optional question on the form or in the first call: 'how did you hear about us?' Self-reported attribution captures what tracking cannot — the podcast, the recommendation, the post they saw weeks ago. Store it beside the tracked source. Where the two disagree, the customer is usually right about influence and the tracking is right about the final click.
Report both, and by pipeline not just leads
Show tracked and self-reported side by side, and show them at pipeline and revenue, not only at lead count. A channel that produces many leads and no revenue is visible immediately. Keep the model simple enough to explain in one sentence, and use the same one every month.
Use holdouts for the real questions
When the question is 'does this channel work', the answer is a controlled test: pause it in a region or a segment for a month and compare. Attribution models estimate; holdouts measure. Budget the big decisions on measurements and use the model for the small ones.
Questions people ask next
- Is multi-touch attribution ever worth it?
- For companies with many channels, long cycles and the data infrastructure to support it, a simple linear or position-based model can be informative. For most growing companies, first-touch plus last-touch plus self-reported, presented together, gives the same decisions with far less machinery.
- How do we attribute offline conversions?
- Capture the online identity at the offline moment — a booking reference, a form filled at the event, a phone number matched to a lead — and connect it to the record. Where that is impossible, self-reported attribution and holdout tests are the honest tools.
- How do we attribute WhatsApp and phone leads?
- Capture the source at the first message or call — a click-to-chat link with a tracking parameter, a tracking number per campaign — and store it on the lead. Ask 'how did you hear about us' in the conversation as the self-reported backup.
Want this built for your team?
A free scoping call: we map your motion, name the number that should move, and quote a fixed-scope build.
Related
- Reporting go-to-market to a board or investors: the numbers, the cadence and the honesty
- Data quality monitoring for GTM systems: catching the silent failures
- Sales activity metrics without the micromanagement: what to track, what to show, what to leave alone
- Alerts on revenue metrics: knowing on Tuesday what you would otherwise learn at quarter end
