Reading a first attribution report without misreading it

What the rows and columns mean

Every attribution report has the same shape. Rows are where users came from. Columns are what they did and what it cost.

One row will be labelled organic or unattributed. It holds every install the MMP could not match to a click. The guide on telling organic installs from paid ones explains what ends up there.

Read from cost to revenue

The install column is the easiest to read and the least useful on its own. A campaign with many cheap installs and no purchases is losing money. Sort the table by revenue, or by a later event if revenue is still thin, and then look at what each row cost. The ratio of revenue to spend is ROAS, and it is the number most budget decisions rest on.

Check which date the report uses. Some views group events by the day they happened. Others group them by the day the user installed, which is a cohort view. The same purchase lands on different days in each, so two reports of the same week can show different revenue and both be right.

Why the totals differ from the ad platform

Your MMP's install count and the ad platform's install count will not match. This is normal and has a name: discrepancy. Four common causes are:

• The platform counts by its own rules and its own window. The MMP uses the window you set.
• When two channels both reached the same person, each platform claims the install. The MMP gives it to one.
• The dashboards use different time zones, or one reports the click date and the other the install date.
• On iOS, part of the data comes from Apple's SKAdNetwork, whose postbacks are sent after a random delay.^[1] MMPs usually show it in a separate report. See SKAdNetwork postbacks.

A small, steady gap is expected. A gap that appears suddenly, or a channel that shows spend and no installs, points to a broken setup.

Four misreadings to avoid in week one

1. Judging on a handful of installs. Ten installs and one purchase is not a conversion rate. Wait for volume.
2. Comparing a new campaign with an old one on revenue. The older users have had more days to buy. Compare both at the same age.
3. Treating organic as free growth. Some of that row is paid traffic the MMP could not match.
4. Reading a creative-level row as final. The deeper you split, the fewer installs each row has and the noisier it gets.

What to do next

On day two, open the report and run the checks in the table above. On day seven, sort by your main event and note which campaigns lead. Change budgets only after the same campaigns lead two readings in a row. If a number looks wrong, the list of common first-campaign attribution mistakes is the fastest place to find the cause.