The marketing attribution playbook
Perfect attribution is unavailable at any price — tracking limits guarantee it. What is available is an honest picture: clean data, several imperfect methods triangulated, and uncertainty stated instead of hidden.
Attribution went from difficult to structurally incomplete: privacy changes, blocked tracking, and cross-device journeys mean every single method now lies differently. The stores that decide well stopped chasing the perfect model and built triangulation instead — platform numbers, GA4, post-purchase surveys, and the occasional incrementality check, each covering the others’ blind spots. This playbook builds that setup at merchant scale.
Stage 1: Hygiene first — most "attribution problems" are data problems
Before any modelling: UTM discipline (a written convention, applied everywhere, including the email and social links everyone forgets), verified GA4 e-commerce events, and channel definitions that reflect how you actually buy traffic. Half of attribution confusion dissolves at this stage — traffic arriving as direct or unassigned is not mysterious, it is untagged.
A written UTM convention enforced on every paid, email, social, and partner link
GA4 events validated and revenue reconciled before anyone trusts a report
Channel groupings matched to your real buying structure
The untagged-traffic share measured and driven down
Stage 2: Triangulate three views that lie differently
Platform-reported numbers (generous: they claim views), GA4 (conservative: last-click-leaning, consent-gapped), and the post-purchase survey (human: catches podcasts, word-of-mouth, and seen-but-not-clicked social that no pixel sees). Read side by side, per channel. Where all three agree, act with confidence. Where they diverge, the divergence is the finding — a channel whose platform numbers dwarf its survey mentions is probably taking credit for demand it did not create.
Stage 3: Test incrementality where the money is
For the largest spend lines, the periodic honest question: what happens if we stop? Geo splits or clean on/off windows on a single channel, revenue watched against baseline. Crude but unspoofable — and routinely humbling for branded search and retargeting, which excel at claiming conversions that were coming anyway. One such test per quarter, on the biggest line item, disciplines the whole budget.
How it works
Fix the data layer
UTM convention written and enforced, GA4 validated, channel definitions aligned to spend.
Stand up the triangulation
Platform, GA4, and survey views per channel, read as one panel with divergences flagged.
Run quarterly incrementality
The biggest spend line tested against its absence; budget reweighted on the result.
What you get
A data layer clean enough that reports mean something
Three imperfect views triangulated into one honest picture
Channels that over-claim identified and reweighted
Uncertainty stated, so decisions carry the right confidence
Handled by these agents
Frequently asked questions
Which attribution model should I use in GA4?
Use the default data-driven model, but hold it loosely — the model matters less than knowing that GA4 is one conservative view among three. Triangulation beats model selection.
Platform ROAS says one thing, GA4 another. Which is right?
Neither — they measure different things (claimed views versus last-click sessions). The survey and incrementality checks arbitrate. A channel strong in all views is real; strong in only its own reporting is suspect.
Is multi-touch attribution software worth it for a store?
At significant multi-channel spend, possibly — see the dedicated tooling for that. At typical merchant scale, clean hygiene plus triangulation plus an occasional incrementality test answers the decisions you actually face, at a fraction of the cost.
Which agents run this?
Analyst owns the whole loop: hygiene enforcement, the triangulated panel, divergence flags, and incrementality test design. Findings arrive as reweighting recommendations, with the uncertainty stated rather than smoothed over.
Keep reading
Analytics reporting
A weekly report that says what changed, why, and what was done about it — without a dashboard.
Set up GA4 e-commerce tracking
The tracking foundation everything else stands on — events, parameters, and validation.
Paid ads landing pages
Stop sending paid clicks to generic pages. Build and iterate landing pages that match the ad.
Design a loyalty programme
A loyalty programme designed from repeat-purchase economics — not points for their own sake.
Put the team to work on your store
Connect your storefront, analytics, and email stack, set a goal, and let the agents run the work end to end. Start on the free plan with your own model key.