Master plan · supersedes nothing, sequences everything · August 2026
Ship less, ask
first, count
two numbers
Five documents now exist: the brief, the business plan, the growth plan, the video playbook and the tagging plan — plus 53 designed screens, a screen index, 16 turns of design argument and 31 logged decisions. This page is the one that says in what order, and what you're allowed to skip.
The whole strategy in one sentence: go deep with fifty collectors before you go wide with anyone — because a collection is the unit of value, and fifty real collections tell you more than fifty thousand installs.
Everything below assumes what's true: one person, evenings and weekends, no funding, no team, and a product that is already designed further than most shipped apps. Your bottleneck is not ideas. It is your own hours — so this plan is mostly about which hours not to spend.
01 · The shape of it
Fifteen months,
six phases, two gates
MONTH 0–1
Cut & join
Scope v1 to 30 screens. Join 5 communities as a person.
MONTH 1–2
Ask
Schema posts. 20 conversations. Collect export files.
MONTH 2–6
Build v1
Importers first, tagging pass 1, then the 30 screens.
MONTH 6–8
TestFlight
50 Founding Cataloguers. Gate 1 lives here.
MONTH 8–9
Launch
5 custom product pages, 2 creators, 8 clips.
MONTH 9–15
Read & decide
Gate 2: build v2, half of it, or neither.
Two hard gates, and they're the only two places in this plan where you're allowed to change direction. Gate 1 (month 8): do fifty invested experts reach 60 objects each? If not, launching is pouring water into a bucket with a hole. Gate 2 (month 15): latent supply ≥8% and latent demand ≥30%? That answers v2 four ways, and you cannot backfill the events that answer it — which is why tagging ships in phase 3, not phase 6.
02 · Phase by phase
What each phase is
allowed to contain
P0
Cut v1 in half
2 WEEKS
NO CODE
HARDEST PHASE
v1 SHIPS WITH
+
Add by photo, barcode, hand — and import.
+
5 types at launch, not 14. The ones your five communities are.
+
Item detail with Overview / Details / History.
+
Lending with handshakes and due dates.
+
Wishlist + the sellable flag (Gate 2 evidence).
+
Search, share links, iCloud, export, review queue.
CUT TO v1.1 OR LATER
—
Nine of the fourteen types. Add one a month, each announced in its community — that's twelve free launch moments.
—
3D capture. Keep the rendered 3D object; cut the scanning flow. It's the most expensive feature per user who'll touch it.
—
Type proposals, mixed collections, no-type-fits, co-shelves, offline queue, playlists, tab-bar editing.
—
The onboarding detail-bar question. Ship one sensible default; the question is a v1.1 refinement.
WHY THIS IS THE HARD PART
You've designed 53 screens and they're all good, which makes every cut feel like a loss. It isn't: an unshipped screen has a value of exactly zero, and each one you cut is roughly a week of evenings back. 30 screens is still more product than most apps launch with — and the cut ones are already designed, so they become your 2026 roadmap rather than deleted work.
DONE WHEN
There is a written list of what is NOT in v1, and you've stopped arguing with it. Put it at the top of decisions.md as D32 so future-you can't quietly re-add things.
P1
Ask before you build
WEEKS 1–8
PARALLEL TO P0
HIGHEST RETURN PER HOUR
THE WORK
1
Join five communities as a person. Answer other people's questions for two weeks. Say nothing about your app.
2
Message the mods, then post the schema question in each: "here are the 11 fields I think describe X — what did I get wrong?"
3
Reply to everyone. Note who wrote paragraphs.
4
Ask twenty of them two questions: what would you pay, and can I have your export file?
WHAT IT PRODUCES
Five expert-validated schemas. A ranked list of importers to build, with real files to test against. A price point that isn't a guess. And 30–50 named people who have already contributed something before being asked for anything — your Founding Cataloguers, and your entire TestFlight cohort.
THE TRAP
Skipping this because building feels more productive. It isn't: a schema corrected now costs an evening, corrected after launch it costs a migration and a bad review. This phase also can't be compressed — communities need weeks of your presence before they'll take you seriously, so it must run in parallel with P0 and P2.
DONE WHEN
You have five real export files on disk, twenty price answers, and thirty names who'd say yes to a TestFlight invite.
P2
Build, importers first
MONTH 2–6
~16 WEEKS OF EVENINGS
BUILD ORDER — NOT NEGOTIABLE
1
Import + the review queue. The path from a stranger's CSV to a full shelf. This decides your north star metric more than anything else you'll write.
2
Tagging pass 1 — nine events. Before TestFlight, so TestFlight produces data and not just opinions.
3
The object model, 5 types, item detail, list, search.
4
Lending and share links.
5
Wishlist + sellable flag. Tagging pass 2 — the Gate 2 events.
6
Photo and barcode add, metered. Unlock and packs.
WHILE YOU BUILD, FOR FREE
Two clips a week from the
video playbook — one sleeve-turn loop (ten minutes) and one story clip. Ninety clips by launch, posted as you go, building an audience that already knows what this is. Filming your own build progress is the cheapest content you will ever make, and it makes the launch a continuation rather than an announcement.
DELIBERATELY NOT BUILT
Accounts. A backend of any kind beyond the share-link server and your one analytics endpoint. A web app. Anything for institutions. Any pricing surface — no valuations, no totals, no market. Every one of these is a month you don't have and a promise you'd have to keep.
DONE WHEN
Someone else's export file becomes a browsable 300-object shelf on your phone, and the nine events land on your endpoint.
P3
TestFlight — Gate 1
MONTH 6–8
50 PEOPLE MAX
STOP-OR-GO
HOW TO RUN IT
Invite the thirty from P1 plus twenty from the clips. Each gets the schema they helped write, credited by name, and a lifetime unlock. Weekly build, weekly one-paragraph note on what changed and what you're unsure about. Reply to everything within a day — this is the month that turns fifty testers into fifty advocates.
WHAT YOU'RE MEASURING
1
Median objects per tester at day 30.
3
Import completion rate, by format.
4
Sellable-flag and wishlist rates — early read on Gate 2.
NOT MEASURING
Bug counts, crash-free rate, feature requests. All useful, none decisive. A cohort of invested experts is the friendliest audience this app will ever have — so their behaviour is a ceiling, not an average. Read it that way.
GATE 1 — MONTH 8
Median 60+ objects per tester → launch. Median 20–59 → the adding path is too slow; fix import and re-run for a month before launching. Under 20 with fifty hand-picked enthusiasts → stop and find out why, because no amount of marketing survives that number.
THE MECHANICS
+
Free to 100 objects, €24.99 unlock, metered lookups with a generous free allowance.
+
Five custom product pages, one per community — their own screenshots, their own link.
+
Subtitle and keywords per storefront: "vinyl record collection tracker", not "collection manager".
+
Two mid-size creators, paid in lifetime codes. Pitch Apple editorial the same week.
+
Launch storefronts: profile A only (DE FR NL UK US JP Nordics). No ad surfaces compiled in.
WHO POSTS WHAT
Not you, mostly. Each Founding Cataloguer announces their own type in their own community, in their own words, with their own page link. That's five posts that aren't advertising because they aren't yours. You reply in the threads; you don't start them.
WHAT NOT TO TOUCH FOR 60 DAYS
Product page optimization. You won't have the impressions to reach confidence, and a running test would split your community and creator traffic across pages you didn't choose. Custom pages for warm traffic, PPO for organic search, never both at once.
DONE WHEN
Five doors live, two videos out, and downloads logged by country and by source page from the first hour.
P5
Read the numbers — Gate 2
MONTHLY RHYTHM
One new type per month, announced in its community. One fix to your weakest funnel stage — the one at your lowest benchmark percentile, never the one you find most interesting. Two clips a week, continuing. And from month 11, one PPO test at a time on the default page: first screenshot, one object versus a full shelf.
PHASE-TWO MARKETS
Around month 12, if profile-A conversion is healthy, open profile B (BR IN ID MX PH TR VN NG) with local price tiers and rewarded video for lookups. This is also when v3 becomes cheap: roles + a queue + three emails, filmed at your recyclerie contact's place as clip 08.
THE INSURER CONVERSATION
Two emails, month 10, before any v2 code exists: what would an itemised export need to contain for you to accept it, and what do you pay for an introduction? Their answer designs the v2 export screen and may be the highest revenue-per-user line in the whole business.
GATE 2 — MONTH 15
Supply ≥8% and demand ≥30% → build v2 option B with a seller subscription. Demand only → wishlist hits + affiliate, a fraction of the work. Supply only → the seller's toolkit (comps, price history, insurance export), no index at all. Neither → don't build v2; take the same code to shops, clubs and recycleries.
03 · The decision calendar
Nine things to decide,
each with a date
Open questions are only dangerous when they're undated. These have dates, so they stop being anxiety and become appointments.
NOW
The v1 cut list
Your hours. Write it as D32 and stop negotiating with yourself.
NOW
Which five hobbies
Where you'd enjoy the conversations for a year. Enjoyment is a real selection criterion for solo work.
MONTH 2
Price point of the unlock
Twenty answers from P1. Not your instinct, and not a competitor's page.
MONTH 2
The app's name — one hobby word, or none
Whether you're the vinyl app that also does coins, or the collection app. The first ranks in search; the second doesn't.
MONTH 4
Photos versus renders (brief Q3)
Real user photos in the build. Current answer: photos where an object is named, renders where it's counted.
MONTH 8
Launch or extend TestFlight
Gate 1. Median objects per tester.
MONTH 10
Whether insurance is a real revenue line
Two emails and the export rate among payers.
MONTH 12
Open profile-B markets
Profile-A conversion being healthy, and an ad partner worth signing.
MONTH 15
v2: full, half, toolkit, or none
Gate 2. Latent supply and latent demand.
04 · Honest risks
The five ways
this dies
MOST LIKELY, BY FAR
You run out of evenings at month four
Sixteen weeks is a long time alone with no users. Mitigation is structural, not motivational: P1 runs in parallel so there are fifty people waiting for the build, and two clips a week means you're publishing something the whole time. Visible progress is the fuel; a silent four months is how solo projects actually end.
SECOND MOST LIKELY
People catalogue 15 objects and stop
The app is lovely, the work is boring, the shelf stays thin. Every mitigation is in the import path — which is why it's built first and measured hardest. If import completion is under 50% at TestFlight, nothing else you build matters.
REAL RISK
Scope creep back to 53 screens
The cut screens are designed and beautiful, and each one costs a week. D32 exists to make re-adding them a decision with a name on it rather than a Tuesday evening's whim.
STRUCTURAL
Lookup costs outrun a one-time price
A single €24.99 payment against unlimited API calls is unbounded exposure. The meter and the free allowance must exist at launch, not be retrofitted — retrofitting a limit onto a paid app is the most hated move in software.
QUIET ONE
You build for shops because they answer their email
Institutions reply, give feedback and feel like progress; consumers are silent. Following that warmth pulls the roadmap toward permissions and admin views and away from the person on the sofa. Institutions are distribution until Gate 2 says otherwise.
05 · Monday
The first two weeks,
concretely
1
Write the cut list. One evening.
Open the
screen index, mark each of the 53 screens v1 / v1.1 / later. Target 30 in v1. Write the not-in-v1 list into decisions.md as D32.
2
Pick five hobbies and five communities. One evening.
Where you'd happily read long arguments about grading for a year. Make the accounts. Answer three strangers' questions in each. Say nothing about the app.
3
Film clip 03 — the sleeve turn. Ten minutes.
You can shoot this today from the prototype. It costs nothing, it's the asset you'll reuse in app previews and product pages, and it proves the content pipeline works before you depend on it.
4
Message five sets of moderators. One evening.
Not a pitch — a question about where a schema discussion would belong. Some will pin it. This is also how you learn each community's tolerance before spending your one first impression.
5
Draft the five schema posts. Two evenings.
One field list per hobby, from the types you already designed, written as a genuine question. These are the highest-leverage words you will write this year — they recruit your testers, correct your data model and seed your launch, all in one post.
6
Then, and only then, open the editor.
First file: the CSV importer. Not the shelf, not the 3D object, not the tab bar. The importer.
You have the rarest thing in this category already: a designed product with an argument behind every decision. What you don't have yet is fifty people who've put their collection into it. Everything on this page is in service of getting those fifty — and after that, the app makes its own case.