Study customer movement, not just a total.
A customer count compresses many different events into one number. Additions, removals, coexistence, and potential switches can tell different stories about the underlying business.
Public observations are proxies for relationships. They are not audited customer counts, revenue, net retention, or a direct measurement of paid usage.
Reconstruct what was knowable.
A historical research process should use only information observed by the chosen cutoff. A disclosure discovered later must not quietly appear in an earlier view of the market.
This is why observation time, reported event time, revisions, and collection gaps need explicit treatment.
Keep the denominator visible.
An increase in detected customers may reflect better coverage rather than business growth. Compare a stable sample, track sources entering or leaving that sample, and document the matching rules.
A large dataset is not automatically a representative dataset. Smaller companies, private systems, and different geographies leave different public footprints.
Keep the analysis traceable.
A useful research workflow keeps dated observations, source references, and matching decisions together so another researcher can revisit the conclusion.
Our research notes explain how to evaluate vendor-change signals. They do not provide investment recommendations or performance forecasts.
