Run A/B tests and personalize your store without code. 14-day free trial, no credit card.
Start free trial →In-depth guides and case studies where this concept is put to work.
Attribution is the process of assigning credit for a conversion to the marketing touchpoints that preceded it. A customer might see an Instagram ad, read a blog post two days later, click a Google search result the following week, and finally buy after a WhatsApp reminder. Attribution decides how much of that sale each of those four moments earned.
There is no objectively correct answer, which is the source of most confusion around the topic. Attribution models are conventions for splitting credit, not measurements of causation. Different models applied to the same data will tell you different stories about which channels are working.
Budget decisions depend on it. If your reporting gives all credit to the last click, paid search and retargeting will always look excellent — they sit closest to the purchase — while the Instagram content that created demand in the first place looks like a cost centre. Cut that spend on the strength of last-click data and search performance mysteriously declines a month later, because the demand feeding it has stopped being created.
The problem has become harder rather than easier. iOS privacy changes, cookie restrictions, and cross-device journeys mean a substantial share of paths are now partially invisible. Platform-reported numbers make this worse: Meta and Google both count conversions they had any hand in, so adding up their dashboards routinely produces more sales than the business actually made.
For Indian D2C brands there is an additional wrinkle. A meaningful portion of the journey happens where no pixel is watching — WhatsApp forwards, family recommendations, Instagram DMs. Any model will under-credit these, which is worth remembering before treating a report as ground truth.
A skincare brand reviewing last-click data concluded that their content programme was not paying for itself: blog traffic showed a 0.4% conversion rate against 2.8% from branded search, and the blog was consuming a substantial monthly budget.
Before cutting it, they ran a holdout: content distribution was paused in three states for six weeks while continuing everywhere else. Branded search volume in the paused states fell 22% over that period, and total revenue fell 14% — considerably more than the blog's own attributed revenue.
The content was creating the demand that branded search was harvesting. Last-click attribution had been reporting the harvest and ignoring the planting.
A/B testing sidesteps attribution's central difficulty. Because visitors are randomly assigned and both groups experience the same marketing mix, any difference in outcome is caused by the variant. There is no credit to divide.
This is why experimentation and holdout testing are the strongest tools available for measuring genuine incremental impact. Attribution models are useful for allocating budget between channels day to day; controlled tests are what you use when you need to know whether something actually worked. CustomFit.ai reports conversion and revenue per variant directly, so on-site changes are measured causally rather than modelled.
Run smarter A/B tests with CustomFit.ai — 14-day free trial, no credit card required.