Business delta · what the new types change · one page · September 2026
The model holds.
The sequence and
the story don't.
Since the business plan and revenue model were written, five things were designed: Wine, Retro Computer, Artwork, the Thing type with community proposals, and migration. None of them breaks the money model — free fifty, one-time unlock, no subscription on the catalogue, no ads. Four of them change what the unlock is for, who it's sold to first, and what it costs to run.
Before, the pitch was "a catalogue that remembers." Wine and retro computers make it "a catalogue that tells you when to act." That is the first honest reason for anyone to open this app on a Tuesday — and retention, not features, is what the €29.99 was missing.
What each type changes
WINE · CONSUMABLE, ON A CLOCK
The first type people already pay for
Drinking windows, "drink tonight", one notification a year per bottle. Cellar apps charge subscriptions and people pay them — this is a market with a paying habit and a real incumbent.
SOWine is the strongest type for the unlock's value, and the one where "no subscription" is a visible advantage on the store page. Also the first type whose free tier of fifty is too small — a modest cellar is 200 bottles. Keep fifty; wine converts faster, not slower.
RETRO COMPUTER · DEGRADES, NEEDS SERVICE
Zero lookup cost, maximum loyalty
The whole catalogue ships in the bundle; identification needs no network. A service clock that says "open it this week" makes the app an obligation, in the good sense. Small community, no incumbent app.
SOLookup credits earn nothing here and cost nothing either — the metered meter is type-dependent, so it can't be the headline SKU. A cheap community to win completely, and the best possible reference customers for the "types by collectors" story.
ARTWORK · UNIQUE, HIGH VALUE
Pulls insurance forward
Few objects, each worth a lot, each with provenance — which is exactly what History already records. Art collectors already pay for appraisals and insurance riders; they are the v2 insurance-export customer, arriving in v1.
SOThe insurance conversation (plan step 3) moves before launch, not before v2. And a second segment appears: artists cataloguing their own inventory. That is a professional use — the one place a subscription would be honest. Resist it until they ask.
THING + PROPOSALS + MIGRATION
The roadmap becomes public
Nobody is turned away on day one; the popular custom types become official ones; the proposer's objects migrate intact. Each promotion is a free launch moment in its own community.
SOThe "five types at launch, cut nine" argument softens: ship five real types plus Thing. But proposals are user-generated content — moderation is the first ongoing cost that isn't Apple's €99, and the shared type board is shared infrastructure arriving a year earlier than the index was meant to.
Where the plan moves
LAUNCH SCOPE
Five types, nine cut to v1.1+
Five types + Thing. The cut types stay cut; the proposal board tells you which to build next instead of guessing. Retro computer is a candidate for slot five: cheapest to build (offline corpus), loudest community.
WHAT THE UNLOCK BUYS
Unlimited objects, import, exports
Plus the clocks. Drinking windows, service risk, maturity sort. Computed on-device from stored fields, so zero cost — they go in the paid entitlement whatever its container (lifetime in FR/DE, yearly in UK/US), never as a separate SKU. Pricing rule 6 holds: plans differ in how you pay, not what you get. This is what justifies €29.99 over €19.99.
RETENTION
Lending — the reason non-cataloguers open the app
Lending + time. Two reasons to return that aren't "I bought something". Add "objects with an active clock" to the seven numbers; it predicts renewal of attention better than objects-per-user does.
INSURANCE
Step 3 — before building v2
Step 1½ — before launch. Artwork and wine collectors need the export in v1. Two emails, same as before, a year earlier. The referral is revenue with no marketplace.
LOOKUP CREDITS
A meter across the app
A meter for network types only. Vinyl and cards spend credits; wine, retro computers and Thing never do. Fine — but it stops being a story you tell on the store page.
COSTS
€99/year and your evenings
Plus moderation. AI pass is cents; the human pass is your hours until proposals exceed ~20/week, then a trusted-collector council before a paid moderator. Budget it as time, and cap proposals per user per month.
SHARED INFRASTRUCTURE
The v2 index of offers is the first
The type board is the first. Lower stakes than money, same lesson: once people download types from it, you can't switch it off. Decide now that it's free forever and never carries a shelf — types only, no objects.
Across the markets already planned
Read against Pricing by Market, the Japan brief, the Vinted model and Market B. The types don't change any of those calls; they change the argument for three of them.
DOCUMENT
ITS OPEN PROBLEM
WHAT THE TYPES DO TO IT
PRICING · US / UK
yearly-led
"People finish cataloguing and stop. The renewal must land on a screen that shows what changed in a year, or it dies at month 13."
The clocks are that screen. "Three bottles entered their peak, two machines are due a service, one artwork's insurance value was updated." A yearly container was the weakest part of the plan because a finished catalogue has nothing to renew for; time-aware types are the first honest answer. Yearly-led markets gain more from these types than lifetime-led ones.
PRICING · FR / DE
lifetime-led
"Cash arrives once. No recurring floor in the market we know best."
Unchanged — deliberately. Tempting to make the clocks a subscription here; it would break rules 1 and 6 and the "no abonnement" pitch that sells the app in France. The recurring floor in FR/DE stays the insurance referral, which artwork and wine now bring forward.
JAPAN
blocked on research
Three questions out, nothing back yet. Placeholder prices. "Would a Japanese collector use this at all?"
Retro computer is the Japan door. MSX, PC-88/98, X68000, Famicom — a hobby Japan leads, with a closed corpus that ships in the bundle, so it needs no Japanese-language API and no server. Add a fourth question to the brief: where do retro-computer collectors gather, and do they keep lists? If the answer is yes, that's the launch type for JP and the pricing question becomes answerable.
VINTED MODEL
decisions 2 and 4
"Free tier — wall at the object count, or at the second type?" and "What is our defensible virtue claim?"
Thing settles decision 2: count, not type. A type-gated free tier leaks the moment Thing exists — everyone's second collection is a Thing. Fifty objects, one collection, any type. And decision 4's longevity claim becomes instrumentable: service logs on machines, repairs, intact transfers — the retro computer type is the first that records longevity rather than asserting it. Model C (the provenance rail) gets stronger with artwork and composites; still not a v1 question.
MARKET B
Android, Brazil-first
Chat-group traders and micro-merchants with mixed stock; offline-first; trap 5 — "don't let market B set the product direction."
Thing is mandatory there; wine and artwork are irrelevant there. Mixed merchant stock needs a type that says yes to everything, and the offline corpus pattern from retro computers is exactly what a prepaid-data market wants. The new risk is the proposal board: market-B users will propose "phone lot", "sneaker resale batch" — merchant types, not collector types. Moderation policy needs a market rule before Android ships: proposals are scoped per storefront and promoted per storefront.
What doesn't move, and the one risk
UNCHANGED
=
Free for one collection of fifty; one-time unlock with a rising founding price; the only subscription is v2 market data.
=
Never subscribe the catalogue, never ads, never sell schemas — proposing and downloading types is free, always.
=
Gate 1 (fifty collectors at sixty objects) and Gate 2 (latent supply and demand) stay where they are.
=
Go deep with fifty collectors before going wide. Thing widens the door; it doesn't change who you sit down with first.
THE NEW RISK
Time-aware types make the app responsible. A missed "open it this week" on a leaking battery, or a "drink it" on a bottle that was already gone, is the first moment CLCTN can be wrong about something that matters. The pitch gets stronger and the tone has to get more careful: every clock is an estimate, labelled as one, and the notification is opt-in per type — one push a year for wine, one per machine for computers, none by default for anything else.
LOG ITD33 — clocks are estimates, computed on-device, included in the unlock, never subscribed, opt-in to notify. D34 — the type board is free forever, carries types never objects, and is scoped per storefront. D35 — the free wall is a count (fifty), never a type.