A baseline is a reconstruction, not a vendor scorecard.
Before anyone picks the destination SDK, we write down what the live app already emits and which leadership charts depend on it. That document is the thing engineering can actually schedule.
What we mean by baseline
For a team leaving a legacy app, a baseline is the last honest description of the measurement surface: every screen that still opens, every event string that still arrives in the property, the identifier that currently means “the same person”, and the handful of Monday figures that leadership will still request after the store listing changes.
It is not a maturity model. It is not a comparison of analytics vendors. Those conversations eat freeze calendar and do not tell you that OrderConfirmV3 only fires on Android.
-
Print the pack
We ask for the last twelve weekly packs. Every number that comes from inside the binary gets highlighted. Store downloads stay in a separate column; they usually survive a move without our help.
-
Click the live client
With a production build, we walk reachable screens and write two names: the title a person sees, and the string that actually lands in the property. Silence is recorded as silence.
-
Name the identifiers
Returning users, sessions, and “the same device after reinstall” each need a sentence with an owner. If the destination build will mint a fresh ID, that sentence becomes a freeze ticket or a footnote — not a surprise in week one.
-
Write the break-list
The brief states what continues, what must be remapped, and what will be lost unless someone specifies it for the new client. Lost items keep the chart they used to feed, so product can retire a comparison on purpose.
Artefacts you should expect
- Screen and event inventory from the live app
- A gap matrix against the destination plan, if you already have one
- Identifier notes: user, session, day bucket
- Footnotes for the first Monday after cutover
- A walkthrough with engineering and the pack owner
What a baseline will not decide
- Which vendor to sign with
- Whether the rewrite itself is wise
- How to spend media the week of launch
- Code in your repository — we write the list, your team instruments
Begin while tickets can still change.
Three to five weeks is the usual span. If freeze is already inside that window, say so in your note. A cutover continuity review can still name the footnotes; it cannot invent an inventory that was never made.
Read the flagship engagement