Open data methodology
How AISubDeal verifies subscription prices
We combine public storefront observations, vendor plan pages and exchange-rate normalization. Each row carries a source label and an observation date; a separate verification date is set only when we have an explicit check. Estimates are never presented as checkout guarantees.
Collection and normalization
Plan names are matched to the closest public vendor tier. Local storefront amounts are stored unchanged. A separate FX snapshot creates the USD comparison value; it never replaces the billing currency.
Sources and verification types
Every price row maps to one of four public evidence levels. Observed means the amount was derived or seen only as a scaled tier, without a product-specific confirmation. Source verified means an independent listing confirmed the local amount: OpenTheRank / AppPriceWatch scrapes on the VPS, Apple IAP / iTunes lookups, or the product-specific verified tier map (those local prices were cross-checked against apps.apple.com). Checkout checked means we reached a payment confirmation screen without completing the purchase. Purchase tested means we completed a real or key purchase path. Source verified is not a completed purchase.
USD equivalents use a public mid-market FX snapshot (fxUpdatedAt) and never replace the billing currency. Tax is included, excluded, or unknown — we do not invent checkout tax. Web list prices and App Store list prices are different channels. Savings vs the US App Store are never computed against a US web price.
Purchase flow testing (separate from price verification)
Completing account setup, redeeming a gift card, and buying a subscription is a different evidence layer from reading a storefront price. When the site owner records an end-to-end or partial test, product×country pages may show a Purchase flow tested badge with the test date. Those records expire (default ~120 days) and never claim a permanent guarantee. Display priority: active owner test → partial step test → official source → price observation → regional template → unknown.
Update frequency and FX
Price collection and site builds are separate. A deployment does not mark every row as newly verified. FX timestamps identify the conversion snapshot; pages use the latest real data or content change for sitemap lastmod.
Tax treatment
Tax is recorded as included, excluded or unknown. We do not infer checkout tax only from a country tax rate. Unknown means the displayed amount must be confirmed at checkout.
Risk and recommendation model
Regional risk combines payment restrictions, account-region friction and setup difficulty. Homepage recommendations weight savings at 55%, risk fit at 25% and payment feasibility at 20%. A cheap region is not recommended when the selected payment method cannot reasonably support it.
Price verdict (not a product review)
The homepage feed and product hubs show a price-worthiness score from 0–100. It answers only: is the current App Store regional gap large enough, and is the floor storefront reasonably buyable? It does not score model quality, brand reputation, or “which AI is better.”
- Price value (65%) —
40 + 1.6 × maxSaving%vs the US App Store row of the base plan, capped at 100. No regional saving scores 35. The US web list is a separate benchmark. - Purchase ease (35%) — starts at 50, then adds or subtracts the floor country’s risk level, gift-card support, and region difficulty.
- Bands — Strong value (score ≥ 75, saving ≥ 18%, and purchase ease ≥ 50) · Fair (score ≥ 58 or saving ≥ 10%) · Marginal (smaller gap) · Skip (no vs-US saving). A 90% discount with a hard-to-buy storefront cannot be Strong value.
This is separate from the homepage decision-tool weights above. Both read the same matrix. We do not publish follower counts, “92% worth buying” polls, or invented product-quality scores. When the daily price pipeline updates a floor, the verdict updates with it.
Optional Worth / Skip votes ask only whether the current regional price is worth switching for. Your choice is stored on this device. Public totals appear only after 10 real votes on that card. Until then we show no count — empty samples stay empty.
Quality and limitations
Schema checks reject unknown products, plans, countries, currencies, duplicate rows and invalid dates. Currency overrides without evidence are quarantined. Store eligibility, payment methods and promotions can still change without notice; AISubDeal does not guarantee checkout.
Corrections and reuse
Report a correction with the product, country, plan, source URL and observed date. Corrections are reviewed before they change indexable pages. The JSON and CSV files may be cited with attribution to AISubDeal and this methodology.