Attribution
Attribution assigns conversion credit across the channels and touchpoints that influenced a customer on the way to a purchase.
What is attribution?
Attribution is the method of assigning credit for a conversion across the marketing channels and touchpoints a customer interacted with before buying. In practice, your attribution model decides which channel gets the credit when a customer clicks a Facebook ad, later searches for the store on Google, and finally buys through a newsletter. Common models are last-click, first-click, linear, time-decay, position-based (U-shaped), and data-driven attribution — and none of them is objectively “correct.”
Why does attribution matter?
Attribution shapes how you read your numbers and split your budget. Choose last-click and the final channels in the journey (often brand search and retargeting) take all the credit, while top-of-funnel channels look undervalued. That directly affects where you scale up or down — and therefore your real ROAS. At the same time, privacy changes like Apple’s ATT, Safari’s ITP, and the deprecation of third-party cookies make browser-based measurement less accurate, which is a key reason stores move to server-side tracking.
Common use cases
- Budget allocation. Judge whether a channel actually drives sales or just harvests credit late in the journey.
- Model choice. Pick last-click for short sales cycles and multi-touch for long, considered purchases.
- Campaign tracking. Tag links with consistent
utm_source,utm_medium, andutm_campaignso sessions are attributed correctly. - Reconciling platforms. Explain why Meta, Google Ads, and Shopify count the same sale differently.
Shopify perspective
Shopify’s marketing reports have historically used last-click (last non-direct click), though newer analytics also offer broader attribution views. Apply UTM parameters consistently to every paid and organic link so traffic lands in the right channels in the reports. Expect Shopify, Meta, and Google Ads to show different figures for the same order — they use different models and attribution windows — and pick one source as your reference rather than adding the numbers together.