Measuring hype: knowing the stealth engine works

you know the stealth engine works when three lines move together: per-account view trajectories bending upward, brand-name search volume climbing, and install spikes that match the days your videos ran. any one of those can lie on its own. tracked weekly, per account, per person, the three together are hard to fake and hard to miss.

Track every account weekly, every person separately

the unit of measurement is one account, one week, one owner. when i ran this for a consumer AI study app, every person was tracked separately from day one, which is what made the numbers actionable instead of decorative: a flat program average can hide one account breaking out and two dying, and the response to those is opposite. the weekly row per account stays small on purpose: videos posted against the daily target, total and median views, best video with its hook, completion signal where the platform shows it, followers, comments worth reading, and anything that mentioned the app by name. thirty minutes on a sunday, twenty rows, and the whole engine becomes legible.

Read the view trajectory, not the view count

a week’s view count tells you almost nothing; the shape across weeks tells you everything. every account sits in one of three stages, and each stage has one correct response.

Dead

a dead account gets near-zero views on content that used to get some, and stays there across weeks. distribution has been withdrawn, usually after behavior that looked automated or a flagged post. the response is covered in running theme pages at scale: stop posting, use the account like a human, and if reach does not return in a few weeks, retire it.

Flat

a flat account holds a low, steady band, the 50 to 200 floor or the 1,000 plateau from six months of testing, while posting daily. flat is not dead, and it is not failure inside the first three months; it is the stage where testing has to accelerate. the response is more format research and sharper hooks, not fewer posts.

Breakout

a breakout account has separated from its own baseline, one or more videos running far past the band. the response is to feed it: double down on the winning format, keep the cadence, and move the proven video toward paid. breakouts also reset the clock on patience for that person’s other accounts, because the skill transfers.

Brand search volume is the compounding metric

the number that proves the engine is building an asset, not renting attention, is people searching the app by name. views reset to zero every morning; searchers accumulate. watch brand-name queries in search console, App Store search traffic, and search-sourced sessions, and expect the line to trail the content by weeks, for the reasons laid out in how burner content feeds SEO and AI search. a program with rising views and flat brand search is entertaining the wrong audience, and that finding should redirect the content before it redirects the budget.

Match install spikes to content days

the crudest check is still the most convincing one in a founder meeting: lay daily installs over the posting calendar and look for spikes that line up with specific videos. attribution here is honest correlation, not a tracked click, because most of this traffic arrives by typing the app’s name. when a 2M-view day and an install spike share a date, you do not need a dashboard to argue causation. keep the overlay chart current weekly and screenshot the matches; they are the program’s receipts.

When to kill an account, and when to kill the program

kill an account when it is dead by trajectory and a rest-and-humanize cycle did not revive it, or when its operator has stopped delivering the daily cadence; the account was only ever that person’s voice with history attached. kill the whole program only on program-level evidence: six months of full cadence with no account reaching breakout, brand search flat the entire time, and no install correlation ever, which together mean the niche, the product or the format library is wrong. what does not justify killing it: two flat months, one dead account, or a nervous week. the engine’s costs are front-loaded and its returns are back-loaded, and measurement exists so that decision gets made on lines, not moods.

this page is part of the stealth growth playbook. the breakdown call is free. Let’s talk Growth.