The obsession economy

Four data layers, all public:

1. Keyword scan. ~485 niche search terms (crafts, collections, plants, pets, wellness, trades, spiritual, weird) against the iTunes Search API, 200 results each. 110 queries completed before Apple rate-limited the IP; 12,712 unique apps captured. The remaining categories were covered partially through apprank's category charts instead.

2. Enrichment. The 162 highest-signal apps got: subscription price points scraped from App Store pages, recent review velocity and sentiment from the iTunes RSS feed, and the developer's full app portfolio.

3. Revenue layer. apprank.io publishes per-app monthly-revenue estimates (rounded, third-party). 94 of 162 shortlisted apps matched; 2,264 category-chart rows were also harvested. Where apprank is missing, a transparent model gives bands: ratings/mo ÷ 1.5% (0.7–3%) × RevenueCat D60 revenue-per-install benchmarks × 1.6.

4. Benchmarks. RevenueCat State of Subscription Apps 2026 (115k apps, $16B revenue): NA median download→paid 2.6%, D60 RPI median $0.55 (top quartile $1.39), only 4.6% of new subscription apps reach $10k/mo within two years, and new launches went from ~2,000/mo (2022) to ~14,700/mo (2026).

What this is not. Not audited revenue. Not a market map. Apple blocked the rest of the scan, apprank's estimates are rounded and vendor-generated, and the model is directional. Everything here is for choosing where to dig — not for a business case.

P1 — Craft utilities are a proven subscription category. Crochet/knitting/sewing apps alone produced a $500k/mo leader plus a second tier at $100k and a long tail of $1k–$10k apps, all released in the last ~2 years. The niche supports at least 15 live subscription apps; the audience pays.

P2 — Camera → value → collection is the new scanner wave. Trading-card scanners are the biggest cluster we found outside crafts: HoloDex $300k, Foloy $300k, Spoly $100k, Kardo $40k, plus Pulio, CardSnap, CardX, Double Holo. ToyWorth ($30k) does it for toys; Brickify models ~$42k for LEGO collections. The pattern: point camera → identify object → show market value → become the collection system of record.

P3 — App factories beat solo artisans. PIXELCELL (36 apps) runs one playbook — pick a hobby, ship a polished subscription utility, cross-promote, repeat: YarnPal $500k, Sewpal $100k, StitchPal, ArtCoach, GradeUp $90k, Hablo $60k, Deko $30k. Springtown Labs (Loopsy $100k) and Scaling Engine (Foloy $300k) run the same model. The winning unit isn't one app; it's the pipeline.

P4 — Silent earners exist. Twin Tales: $70k/mo on 82 ratings, 2 months old. ToyWorth: $30k on 903 ratings. TCG Foloy: $300k on 2,968 ratings. Low review counts no longer mean low revenue — ratings velocity is a lagging signal when paid UA and hard paywalls are doing the work.

P5 — The leaders are hated by their own customers. 50.4% of YarnPal's recent sampled reviews are 1–2★; Foloy 59%, Spoly 45%, Kardo 74%, CardX 71%, Sewie 75%. Most complaints are trial/subscription anger ("charges after 7 days", refunds) and accuracy gaps. High revenue + high anger = the cleanest wedge on the App Store: keep the demand, fix the trust.

P6 — Games already farm these niches at 10× the revenue. Yarn Loop: Knit Puzzle $3M/mo, Wool Craze $200k, Food Hunt $3M, Coin Master $3M, Dicero $2M. Proof the audiences are huge — but casual-game economics (ad + IAP whale) are a different business than a $40/yr subscription utility.

P7 — The adjacent categories are validated too. From apprank's category charts: Plant Parent $1M/mo, Planta $300k, GrowIt $200k (plants); Woofz $400k (dog training); I Am Sober $200k; Co-Star $400k and CHANI $500k (astrology); Paired/Candle/Couple Joy $200k each (couples). Obsession + recurring job + subscription works across all of them.

P8 — Pricing is standardized and aggressive. Weekly $3.99–$9.99, annual $29.99–$99.99, 3–7 day trials, hard paywalls. YarnPal sells $3.99 weekly and $39.99–$49.99 annual. RevenueCat: hard paywalls convert ~5× freemium (10.7% vs 2.1% by day 35), and annual-plan apps realize ~2× the revenue per install.

Table 1 — third-party revenue estimates (apprank, collected 7 Sep 2026). Non-game apps with an estimate; the vendor's $1k floor means “under ~$5k”. “Angry users” = ≥25% of recent sampled reviews at 1–2★. “Silent earner” = ≥$10k/mo on ≤5k lifetime ratings. “Factory” = developer runs 5+ apps.

Ranked by evidence × buildability × distribution. Evidence links are apps from the shortlist above.

Filters that found the outliers. (1) Niche keyword breadth: apps matching 2+ specific hobby terms. (2) Ratings velocity 3–40/mo: enough traction to prove demand, small enough to beat. (3) Age < 36 months: new winners, not monuments. (4) Revenue ≥ $10k/mo (apprank) with ≤ 5k ratings = silent earner. (5) 1–2★ share ≥ 25% = complaint goldmine. (6) Developer portfolio size: factories signal a repeatable playbook to study (or avoid).

The 30-day validation plan. Week 1: pick two wedges, mine 200 recent reviews of each incumbent for the exact complaint taxonomy, and run an ASO/keyword audit. Week 2: build the thinnest demo that fixes the top complaint (concierge version is fine). Week 3: test distribution — 5 short-form videos in the niche's native format; measure saves/shares, not likes. Week 4: hard-paywall pricing test ($29.99–$49.99 annual, 3–7 day trial) against a cheap build; kill or commit.

Pricing norms to copy. Weekly $3.99–$9.99 for impulse niches; annual $29.99–$49.99 for hobbyists; hard paywall or 3–7 day trial; annual upsell ~2× RPI. Don't ship freemium by default — RevenueCat's data says hard paywalls convert ~5× and retain the same after a year.

What not to build. Generic AI wrappers (41% more revenue per payer but 30% faster churn), another plant-ID or calorie scanner, or a game unless you are a game studio — games took the niche revenue here, but they're a different machine.

Sources. iTunes Search API and iTunes RSS customer reviews (Apple, public); App Store product pages (Apple, IAP price points); apprank.io per-app and category revenue/download estimates (collected 7 Sep 2026, vendor methodology, rounded); RevenueCat State of Subscription Apps 2026 (115k apps, $16B revenue); AppScout niche analysis (YarnPal $500k, MoneyPilot $20k, PadLab $20k, ReStyle $30k claims); Codex-assisted web research (Fishbox ~$100k claim via paywallscreens; LoopCraft $15–80k conflicting via screensdesign — treat both as unverified).

Limits. Apple rate-limited the keyword scan at query 110/485 (403); wellness, food, pets, trades and spiritual categories are therefore covered via apprank charts and targeted checks, not the full keyword sweep. 94/162 shortlist apps have apprank estimates; the rest are modeled. Review sentiment samples ≤ 400 written reviews per app — ratings counts are larger. The model (ratings rate 0.7–3%, RPI × 1.6) is directional; do not underwrite against it.

Local files. data/dataset.csv — all 162 enriched apps with every field. data/apprank_categories.json — 2,264 category-chart rows. data/apprank_apps.json — per-app dossiers incl. developer portfolios. scripts/ — the full pipeline (scan, enrich, apprank harvest, merge).