Marketing Attribution Explained Without the Jargon

Why the numbers never add up

A customer sees an Instagram ad, forgets about it, searches your brand two weeks later, clicks a Google ad, leaves, then arrives via a newsletter and buys. Three channels touched one sale.

Meta claims it, because it showed the first ad. Google claims it, because it delivered the click before purchase. Email claims it, because it was last. If you total the platform reports you will find you sold roughly twice what you actually sold, and every platform is telling the truth from where it stands.

What the common models actually assume

Each model encodes a belief about how buying works:

• Last click — all credit to the final touch. Systematically overvalues search and retargeting, undervalues everything that creates demand.
• First click — all credit to the discovery touch. The mirror image, overvaluing awareness channels.
• Linear — credit split evenly. Simple and defensible, but treats a passing impression as equal to a decisive one.
• Data-driven — the platform's model assigns credit statistically. Better in principle, opaque in practice, and it always seems to favour the platform reporting it.

Pick one model and hold it still

The value of a model is consistency, not accuracy. If you use last non-direct click all year, a rise in a channel's contribution means something real changed, because the measurement did not.

Switching models to find a flattering number destroys the only thing attribution reliably gives you — comparability over time.

Use blended metrics for actual decisions

The most trustworthy number is the least sophisticated one: total marketing spend divided by total new customers. It cannot be gamed by any platform because it does not depend on any platform.

Track blended CAC monthly alongside per-channel figures. When platform-reported performance improves but blended CAC does not move, the channels are competing over credit for the same customers rather than finding new ones.

Test incrementality when the stakes justify it

The only way to know a channel's real contribution is to turn it off. Pause it in one region or for a defined period and watch total revenue rather than that channel's reported revenue.

This is uncomfortable and it is the only genuine answer. Brands that run these tests routinely discover that some of their best-reported spend was buying customers who were going to purchase anyway — usually branded search and heavy retargeting.