How attribution actually works
Most conversions involve more than one touch: a customer sees a Meta ad, later searches your brand, clicks an email, and finally converts through a Google ad. The attribution model is the accounting rule that decides how to split the credit for that sale across those touches. It does not change what happened; it changes the story the reports tell about what happened. Because budgets follow credit, the model is not a reporting detail. It is a decision about which channels get funded and which get cut.
Last-click vs data-driven
Last-click attribution hands all credit to the final click before conversion. It is simple and stable, but it systematically overvalues bottom-of-funnel channels like brand search and undervalues the prospecting that created the demand in the first place. Data-driven attribution, now the Google Ads default, uses the account's own conversion patterns to distribute credit across touches. It is fairer to the upper funnel but harder to explain and less stable, since the model shifts as data accumulates. Neither is truth; each is a different lens on the same journey.
Why your platforms never agree
Google, Meta, GA4, and Shopify each run their own attribution with their own windows and their own view of the journey, so they will never reconcile. Each platform can only see its own touches and naturally credits itself generously, which is why the sum of platform-reported conversions almost always exceeds the real total. This is not a bug to fix; it is the structural reason iClick steers by blended, top-down numbers like MER for budget decisions and treats each platform's attributed ROAS as a diagnostic, not a source of truth.
How iClick actually uses attribution
The practical stance is to pick a consistent in-platform model for optimising bids, usually data-driven, while never trusting any single platform's attribution for how much to spend overall. Bid decisions use the platform model because that is what the algorithm optimises to. Budget decisions use blended performance and, where it matters, incrementality tests that measure what actually changed when spend moved. Attribution tells you how to allocate within a channel. Only holdouts and blended math tell you whether a channel is truly adding revenue.
