28 Jan 2026 · Instrumentation
The unnamed screen is still the cheapest way to ruin a funnel
Product loves a mysterious drop-off. Instrumentation usually knows exactly which view was never baptised.
28 Jan 2026 · Instrumentation
Product loves a mysterious drop-off. Instrumentation usually knows exactly which view was never baptised.
Every few months a team arrives at the High Street table with a funnel that “falls off after onboarding.” The plot looks expensive. The payload is not. Somewhere between welcome and the claimed feature sits a view whose analytics name is empty, inherited, or copied from a previous white-label.
Once that happens, App Analytics becomes theatre. You can still draw a staircase. You cannot say which step a person actually saw. Designers argue about copy. Engineers argue about a race. The missing name sits in the schema like a blank ledger line, and nobody wants to be the person who admits the line was never filled.
In Release Pulse Atelier we stop the week if the claimed screen has no stable name. We do not “approximate” with session length. We do not stitch from screenshot heatmaps. Approximation is how unnamed screens survive another quarter.
A good name is boring: screen_checkout_review, not “CheckoutExperienceV3Final.” It survives a rebrand. It does not include the designer’s first name. Taxonomy Desk exists because some houses need four weeks on this boredom before they are fit for the flagship.
If you must present before the name is shipped, the reservation is the presentation. Write: “We cannot locate the claimed screen in the payload; this funnel is not evidence.” That sentence has ended more bad store experiments than any A/B calculator we have seen.
Related reading: listing tests that never reach first-open. To practise the defect list on a live table, start at the atelier.