Cartograph — iOS MVP Scope

Company: Cartograph (company id retro-games) · Date: 2026-10-01

Stage: concept-stage company — this document is a scope, not a shipped app. Nothing below is built unless tagged [existing].

Status legend: [existing] = demonstrated on the landing-page demo today · [planned] = MVP build target · [later] = explicitly out of MVP scope.

1. Product & audience

Cartograph is a scan-and-collect app for physical retro video games. A photo of a cartridge, disc or case becomes a verified catalog entry — console, region, edition, condition — plus a dated, sourced value estimate, shelved in a collection that feels physical.

Primary audience / job to be done

Audience	Job
Collector with 50–2,000 items	"Tell me what I have, what it's worth, and what I'm missing — without a spreadsheet."
Flea-market / yard-sale buyer	"In 10 seconds in the aisle, is this a $5 cart or a $500 one? Am I about to buy a duplicate?"
Retro game store	"Batch-intake inventory and price checks at the counter."
Haul/shelf content creator	"Show my collection beautifully; make shelf content people can browse."

The v1 promise: photo → verified edition → graded condition → dated estimate → shelf. Deliberately not a marketplace, not a price-guide website, not an emulator front-end. No game downloads of any kind.

2. Screens & golden path

#	Screen	Purpose
1	Onboarding (2 panels + sign-in)	Pick collected consoles; choose anonymous-identifier sign-in (Sign in with Apple)
2	Shelf (home)	Tactile shelf/grid of owned items; totals; wanted-list tab
3	Scan camera	Live camera w/ viewfinder; single-shot capture; bulk "desk-scan" mode `[later]`
4	Confirm match	Identified title + confidence; user picks region & edition chips `[existing: demo]`
5	Condition	Loose / CIB / Sealed + grade chips with guided checklist `[existing: demo]`
6	Estimate card	Dated, sourced value estimate + 12-mo trend `[existing: demo]`
7	Item detail	Full record: photos, edition, condition notes, estimate history, duplicate flag
8	Wanted list	Items marked wanted; estimate drift alerts `[planned]`
9	Settings / export	CSV export, grading defaults, data controls

Golden path (MVP acceptance flow): Open app → tap Scan → photograph a cartridge → confirm edition (≤2 taps) → grade condition → see dated estimate → item lands on shelf → rescanning the same item surfaces duplicate detection.

3. MVP feature set & acceptance criteria

Feature	Acceptance criteria
Photo capture	`[planned]` AVCaptureSession single-shot; works under typical desk lighting; saves HEIC + downscaled copy for upload
Identification	`[planned]` On-device candidate proposal → catalog match with confidence; user must confirm console/region/edition before save. `[existing: demoed]`
Edition confirmation	`[planned]` Region (NTSC-U/PAL/JPN…) + edition tree per title (e.g. PS1 Black Label vs Greatest Hits). Ambiguous matches always ask rather than assume.
Condition grading	`[planned]` Loose/CIB/Sealed + 4-step grade with per-completeness checklists. `[existing: demoed]`
Value estimate	`[planned]` Per edition × completeness × grade; always dated, always shows source label ("sold-listing aggregate"), never a bare number. Sample data labels where applicable.
Collection shelf	`[planned]` Persistent local collection; tactile shelf UI; sort/filter by console, value, recently added. `[existing: demoed]`
Wanted list	`[planned]` Mark items wanted; visible in shelf + item detail. `[existing: demoed]`
Duplicate detection	`[planned]` Scan of an already-owned edition flags it and opens the existing record instead of adding. `[existing: demoed]`
Purchase price field	`[planned]` Optional "paid" amount per item; collection totals show paid vs. estimated.
CSV export	`[planned]` Full collection export from Settings.
Offline	`[planned]` Shelf and records readable offline; scans queue until connectivity returns.

Out of scope for MVP [later]: buying/selling/trading lanes, physical mystery drops (scoped separately — inventory, fulfillment, returns, margins), barcode scanning for newer games, Android, social profiles, grading-service integrations.

4. Architecture & Apple frameworks

App: SwiftUI, iOS 17+ deployment target, SwiftData (or Core Data) for the local collection store.
Camera: AVFoundation (AVCapturePhotoOutput); Vision framework for on-device rectangle/text detection to crop the cartridge before upload.
Identification pipeline (planned, dependency-flagged):

1. On-device pre-processing: object bounding box + label crop via VNDetectRectanglesRequest / VisionKit.

2. Hosted match service: image embedding similarity against our edition catalog (self-hosted model or a vendor API — a dependency decision, see §6). Deterministic edition is only finalized with user confirmation.

3. Pricing: dated estimate pulled from a pricing-data source (sold-listing aggregates) — external dependency, never invented numbers.

Auth & sync: Sign in with Apple; collection sync via CloudKit private database (no custom auth server needed for MVP; keeps costs near zero and data private-by-default).
Processing split (local vs hosted): camera capture, cropping, shelf UI, collection store, duplicate detection → on-device. Image embedding match and pricing aggregates → hosted. Offline scans queue locally (BackgroundTasks for retry).
Settings: StoreKit 2 later for subscription; Photos add-only permission for saving scan photos to library (optional).

5. Permission UX

Permission	When asked	Fallback
Camera	First tap on Scan, with a one-line explainer ("to identify your games")	In-app photo-picker via `PhotosPicker` (PHPicker requires no permission)
Photo library (add-only)	Only if user enables "Save scan photos"	Feature off by default
Notifications	After first wanted-list mark, contextual	Silent wanted list, no alerts
Sign in with Apple	First sync/save	Local-only mode — collection persists on device

Every permission is asked in context, never on first launch.

6. Data & API dependencies (open, not invented)

Dependency	Need	Status
Edition catalog	Title/console/region/edition tree + label imagery for matching	Dependency — options: licensed catalog DB, community-sourced dataset, or own catalog build for the pilot console set
Sold-listing pricing	Dated, sourced value estimates	Dependency — partner feed or licensed data; MVP shows date+source with every number
Match service	Image→edition embedding match	Dependency — build small (pilot consoles only) or vendor
Backend	Thin hosted API: match endpoint, estimate endpoint, edition tree	Cloudflare Workers + D1/R2 pattern (matches studio infra) or similar low-cost edge stack

7. Monetization (hypotheses, to be validated)

Pilot: free.
Post-pilot hypothesis: free tier (≤100 items, 20 scans/mo) + Cartograph Pro ~$4.99/mo or ~$39.99/yr (unlimited scans, full history, wanted alerts, bulk mode). Numbers are hypotheses pending pilot willingness-to-pay data.
Store tier (batch intake, multi-user) explored only after collector PMF signals.
Later lanes (marketplace take-rate, mystery drops) scoped and modeled separately — inventory, fulfillment, returns, margins.

8. Build sequence

Sprint 0 — catalog data model (title/console/region/edition/condition), seed pilot console set; shelf UI skeleton.
Sprint 1 — camera capture + on-device crop; confirm-edition flow against stubbed catalog; condition grading; local SwiftData collection.
Sprint 2 — hosted match service for pilot console set; estimate endpoint (sourced, dated); duplicate detection; wanted list.
Sprint 3 — Sign in with Apple + CloudKit sync; CSV export; offline queue; TestFlight pilot build → accuracy-validation pilot (see GTM).
Hardening — accessibility audit (VoiceOver, Dynamic Type, Reduce Motion), App Review privacy manifest + nutrition labels, then store submission.

9. Key unresolved decisions

Catalog source: license vs. community vs. build-in-house (cost vs. coverage trade-off).
Match model: self-hosted embeddings vs. vendor vision API — depends on catalog licensing terms.
Pricing feed: which sold-listing source allows app redistribution, at what cost.
Whether condition grading guides toward hobbyist grades (Good/VG/NM) or aligns to WATA-style tiers later.
Store/batch-scan persona timing vs. collector-first focus.

Cartograph · company id retro-games · generated 2026-09-30 · internal planning doc