← Lab

Fitnit · app store

Six store
pages

Apple lets one app have many App Store pages, and lets you A/B test them. It is the only surface where Fitnit has enough traffic for a test to finish. Here is what the two tools actually are, the arithmetic that says test one thing at a time rather than three, and six alternate listings worth building.

Nothing here is built. The page concepts are written, not designed — the screenshot boxes say what each frame would show, they are not artwork. The numbers are real: App Store figures from 17–26 August 2026, paywall figures from the last 90 days.

The two tools, without the acronyms

Custom Product Page Product Page Optimization
What it is An extra version of your App Store page — different screenshots, different text. Apple's built-in A/B test. Upload alternate art, Apple shows it to a slice of real visitors and reports which won.
How many Up to 70 Up to 3 alternates against the original, at once
Who sees it Since July 2025 you can attach words from your keyword field, and it shows up in ordinary App Store search for those words. Also works as an ad landing page. A percentage of everyone who lands on your page, chosen by Apple.
What you can change Screenshots, preview video, promotional text. Icon, screenshots, preview video. Not the text.
Tells you which won? No — it just serves the page Yes, with a confidence range
Costs Nothing Nothing, but takes up to 90 days

Yes, three at once — and you should not

Apple allows three alternates against the original in a single test. At Fitnit's traffic that is the wrong call, for a reason that is pure arithmetic: a test needs a certain number of people per version, so four versions need roughly four times the traffic of two.

Test shapePage views neededAt 18 a dayRisk
Original vs one alternate ~350 ~3 weeks Clean read
Original vs three alternates ~700 ~6 weeks Three chances for one to look like a winner by luck

Both assume you are trying to detect a large change — a download rate going from 28% to about 42%. A modest 20% improvement needs roughly 2,100 page views per version, which is past Apple's 90-day limit and therefore not measurable here at all.

So: one alternate at a time, and make it a genuinely different idea rather than a tweak. Subtle is untestable at this size. Build all six as Custom Product Pages — those need no traffic to be worth having, because each one just serves a better page to a different search — and feed them through the A/B test one at a time.

Where the people actually go

Ten days of real App Store data. Almost all of it is search: 1,576 of the 1,700 recorded events came from someone typing something into the App Store.

152 / day Saw Fitnit in search results
↓ 12 in 100 tapped
18 / day Opened the App Store page
↓ 28 in 100 downloaded
5.1 / day Downloaded
↓ and eventually…
1.3 / day Saw the paywall
0.1 / day Bought
Read this before designing anything. Those two conversion rates — 12 in 100 tapping, 28 in 100 downloading — are normal-to-good for an iPhone app. The page is not broken. What is small is 152 people a day seeing Fitnit at all, and neither of these tools creates that number; both only squeeze the traffic you already have. More impressions come from ranking for more words, which is the keyword field, the title and the subtitle — or from paying for them.

What the listing says today

Name Fitnit: AI Rep Counter Subtitle Count Reps & Calories with AI Keyword field
workoutcaloriefood trackergym weight losspersonal trainermacrosbarcode squatpushupformdiet

Snapshotted at v1.97.0, so it may have moved since — the app is on v1.150.0. Re-pull with asc metadata pull before acting on it.

Those twelve words are really four different people searching. One wants a rep counter, one wants a calorie app, one wants a trainer, one wants to lose weight. They all get the same page today. That is the gap the six pages below fill.

Six pages

Each is a Custom Product Page: same app, different first impression, attached to the search words it answers. The screenshot boxes describe the frame, they are not designs.

Page 01 · the unique one

Your camera counts the reps.

Point the phone at yourself. That is it.

The only thing here no competitor has. Everything else Fitnit does, five other apps also do; this one is worth leading with for anyone whose search implies form or movement.

For the person searching
squat, pushup, form — movement words, not goal words
Keywords to attach
squatpushupform
Frames
live pose overlay mid-squat, count on screenform feedback calloutset finished, reps logged
Page 02 · the big market

Photograph your plate.

Calories and macros, without the weighing.

Cal AI's territory, and the highest-volume search cluster in the keyword field. A person typing “calorie counter” does not care that Fitnit counts push-ups, and the current page spends its first frame telling them.

For the person searching
calorie, food tracker, macros, barcode, diet
Keywords to attach
caloriefood trackermacrosbarcodediet
Frames
meal photo becoming a macro breakdownthe five nutrient slidersa day's log, totals at top
Page 03 · the expensive alternative

A coach that reads your log.

It knows what you lifted last week.

“Personal trainer” is a search with a price anchor attached — the alternative costs $60 an hour. The pitch is grounding: this one has actually seen your history, which is the thing a generic chatbot cannot say.

For the person searching
personal trainer — and comparing against a real one
Keywords to attach
personal trainergym
Frames
coach citing your own numbers backa session it wrote, as tappable cardslogged from the chat
Page 04 · the outcome

Watch the line go down.

Weight, trend, and an honest projection.

Weight-loss searchers want proof it works, not a feature list. Fitnit's trend line with its capped projection is unusually honest for the category — it says “projecting” on thin data rather than drawing a confident line to zero.

For the person searching
weight loss, diet — outcome words
Keywords to attach
weight lossdiet
Frames
weight chart with trend + projectioncalorie budget vs intakea streak of logged days
Page 05 · the friction one

Log it without opening the app.

Say it to Siri. It lands in your day.

Shipped in v1.149.0 and currently invisible on the store. Logging friction is the number one reason people abandon trackers, and “you do not even have to open it” is a sharper promise than any feature count.

For the person searching
workout, gym — people who log every session and resent it
Keywords to attach
workoutgym+ “siri” (new)
Frames
Siri accepting a spoken setit appearing in today's logthe home widget updating
Page 06 · the untested one

Nobody trains alone.

Friends, leaderboards, challenges.

The only concept with no keyword behind it today, so it needs a keyword-field change before it has an audience. Listed because it is the one angle competitors in the rep-counting niche do not run, and because the social features already exist and are entirely unadvertised.

For the person searching
nothing yet — would need “challenge” or “leaderboard” added
Keywords to attach
challenge (new)leaderboard (new)
Frames
leaderboard with real names + flagsa challenge in progressa friend beating your record

The order I would do this in

  1. Re-pull the metadata first. The snapshot above is from v1.97.0 and the app is on v1.150.0 — act on what the listing actually says today, not on a two-month-old copy of it.
  2. Build pages 01 and 02 as Custom Product Pages. They cover the two biggest and most different search intents, they cost nothing, and they do not need a test to be worth having: a better-matched page is better whether or not you can measure it.
  3. Run ONE A/B test — the camera-first page against the current one. That is the real strategic question: is the rep counter the hook, or the generalist framing? Three weeks to a readable answer, if the difference is large.
  4. Then the rest, one test at a time. Page 06 needs a keyword change first, so it is last.

And the thing worth repeating: none of this makes the paywall testable. That needs about 530 paywall views to detect even a doubling, and at 1.3 a day it is a year away. Store work earns users and revenue on its own merits. It does not buy statistical power downstream.

Board written 8 September 2026 · concepts are copy, not designs
App Store figures: asc_metrics_daily, 17–26 Aug 2026 · paywall figures: PostHog, trailing 90 days
Listing copy: metadata/app-info/en-US.json and metadata/version/1.97.0/en-US.json