By Ryan Richardson · Published 8 October 2026
A new event gets added to the page: a button click, a new step in checkout. Reports for that event stay empty, or a chart that used to track smoothly suddenly changes shape on a date when nothing visibly changed on the page itself.
A strict event collector only accepts a fixed, short list of event names. Anything outside that list should be rejected loudly, but a collector built to fail quietly instead drops it with no error and no log entry, which means the gap is invisible until someone happens to notice a report that looks wrong. The same silence shows up as schema drift: a field renamed or a value format changed upstream, with the chart simply shifting shape on a date nothing changed.
A second, separate version of the same problem runs through third-party analytics suites. A large free analytics tool running alongside your own event log sets its own thresholds for how small a count can get before it withholds the row, without publishing what that threshold actually is. A book funnel at modest volume hits those thresholds routinely, and the suite just shows fewer rows with no explanation.
Make the collector strict and loud: a short, fixed list of accepted event names, and an explicit complaint every time it rejects one. Treat your own event log as the record of what happened, and treat any third-party analytics suite as a second opinion only, never the source of truth. Watch for a chart changing shape on a date nothing changed on the page; that's the signature of schema drift, not a real shift in behaviour.
Check whether your collector logs a rejection every time it drops an event, or whether it just silently discards anything unrecognised. If it's silent, you have no way of knowing how much data you've already lost. Add logging on rejection before adding any new event type to a page.
| Claim | Value | Source |
|---|---|---|
| What happens to an event added to a page but not the collector's accepted list | dropped silently unless the collector is loud about rejecting it | Measured in Real Money, Field Manual |
| Behaviour of large free analytics suites alongside a book funnel at modest volume | they withhold rows below their own data thresholds, without publishing the exact number, and a modest-volume funnel hits those thresholds routinely | THE BOOK FULL.md Piece 17, line 1872 |