Field note · August 2025
Instrumentation debt at Series A
Two hundred events and no owners is not a culture of measurement. It is a liability that compounds every sprint, especially after the round when everyone suddenly wants “better dashboards.”
Series A is when App Analytics usually goes from “the founder’s Mixpanel login” to “a hiring plan and a board slide.” The tracker, meanwhile, still contains every experiment from the seed years, including the ones named after jokes. Investors ask for activation. Nobody can reproduce last month’s number. A consultant suggests a CDP. The CDP arrives. Now you have two piles of jokes.
Debt has a shape
In audits we keep seeing the same four piles: (1) events with no owner, (2) events that fire twice, (3) properties that changed type after a refactor, and (4) dashboards that filter the mess until the line looks like the pitch. Only the fourth pile is visible in the board pack. The first three are why the fourth exists.
A consumer lending team we later wrote up as a case study arrived with 214 live events. The useful set was 61. The rest were not “historical richness.” They were unowned triggers that still affected funnel counts when someone forgot a filter.
Triage without a six-month freeze
Stop adding events that are not on the activation, conversion, or expansion contract. Appoint owners for the core twenty in two weeks. Dual-fire only with a kill date. Put deprecation into CI if you can; if you cannot, write the date on the engineering wiki and accept that some zombies will survive.
Do not migrate to a new vendor as the cleanup strategy. Migrations copy debt faster than they delete it. Clean names and owners first; move pipes later if you still need to.
What the round actually bought you
Money buys engineering time, not truth. Truth is a written activation contract and a warehouse query a new PM can run. If the Series A plan includes “stand up a growth analytics function” without hours for deleting events, you have funded a second dashboard, not measurement.
Activation Signal Laboratory is built for this moment: not because we enjoy scolding early teams, but because the cost of an unowned dictionary jumps the day you hire the third PM. Observatory time exists for companies that already know that and still cannot get a kill date through engineering.
If you are earlier than Series A, Desk Trace is enough. Do not romanticise debt while the product still fits in one person’s head. Write the contract anyway. Future-you is a stranger.