Attribution coverage: the number that decides whether any of your other numbers are true
Before you compare channels, ask what share of orders you can attribute at all. Everything else is a ratio of that.
Every attribution report is a division. The numerator is orders you can credit; the denominator is orders you have. If the numerator is 60% of the denominator, every channel comparison you make is a statement about a partial and non-random sample.
Why the missing 40% is not random
Orders lose attribution for specific reasons, and those reasons correlate with channel.
- Safari and in-app browsers lose identifiers faster, and they skew toward paid social.
- Mobile app orders often carry no web tracking context at all.
- Long consideration cycles lose the first touch to cookie expiry, which penalises upper-funnel channels specifically.
So the orders you cannot attribute are disproportionately the ones that would have credited the channels you are trying to evaluate. The gap does not just add noise, it bends the answer in a predictable direction.
What to measure it against
Coverage should be measured against orders that could have carried a signal in the first place. Draft orders, point-of-sale orders and manual orders never had a browser session, so counting them in the denominator manufactures a problem that does not exist. Report them separately rather than folding them in.
Never state a gap against a denominator that could not have held the thing you are looking for. It is the fastest way to lose a client's trust in a report that was otherwise correct.
What to do next
Find your coverage number, then split it by platform - web, app, offline. Most stores discover the aggregate was hiding one healthy channel and one broken one.
Read next
Match quality: why the same sale is worth more to your ad platform with better data