Total installations describe volume. To understand acquisition, the team also needs to know which sources were identified, which were unknown, and whether the numbers use comparable criteria.
The same total can hide different questions
Imagine two weeks with the same number of installations observed. In one, the acquisition effort was concentrated in one channel. In the other, it was distributed among several initiatives. The similar final number does not demonstrate that the initiatives had the same performance.
Before recommending a change, ask which source signals arrived, what time period was used, and how each source defines installation. A difference in definition may seem like a difference in outcome.
Three groups that need to appear
Separate records whose origin was observed, those whose origin was inferred by a rule, and those for which there is no sufficient answer. The classification needs to be explainable. Filling in a blank with an assumption does not create new evidence.
The unknown group also tells you something about coverage. Hiding it can lead to a comparison in which only the journeys that are easiest to identify are included. The team must be able to reproduce the calculation and understand the denominator used.
How this guides Bridgee
Bridgee’s app attribution proposal is to work on the connection between eligible source signals and installations, preserving the role of Firebase and GA4 in the app's internal events. The evaluation needs to happen per supported scenario.
An identified source helps understand acquisition. The investment decision still depends on the business objective, cost and subsequent results. It is not an automatic recommendation to increase the budget for a channel.
See how to distinguish the events in the paper download, installation and first_open.