Methodology

PriceChkd measures prices, not motives. Every result on this site is built from verified observations — submissions an admin has independently reviewed — and every number is labeled with how much data backs it.

Two axes, not one

Every price comparison on this site measures two independent things:

• Others right now (cross-sectional) — what are comparable people seeing at approximately the same moment, in the same store/fulfillment/account context?
• Price history (longitudinal) — how has this same comparable offer moved over time, regardless of who was looking?

We never conflate the two. A price that changes over time is not evidence that different people are being charged differently at the same moment, and different people seeing different prices at one moment is not evidence that the price is generally unstable.

How we decide what counts as "comparable"

Each observation is fingerprinted from its context — merchant, product, store, fulfillment, quantity, account/loyalty/promotion state, and currency — using SHA-256. The cross-sectional fingerprint adds a narrow time bucket (15 minutes for retail/hotels, 5 minutes for delivery/ ticketing, 60 seconds for rideshare); the longitudinal fingerprint omits time entirely. Two observations only get compared if their fingerprints match exactly.

Verification

Every submission starts as Pending Review. An admin marks it Verified, Excluded (doesn't fit this comparison — wrong store, wrong fulfillment, etc.), or Rejected (not usable at all). Nothing is ever deleted — excluded and rejected observations remain in the audit trail. Only Verified observations feed public comparisons and statistics.

Confidence scores

Each observation gets a deterministic 0–100 confidence score built from what's actually known about it: exact product (+20), exact store (+15), fulfillment/account state/promotion state/ quantity (+10 each), whether it's still fresh for its category (+10), whether evidence was attached (+10), and a valid currency (+5). 90+ is Very High, 75+ High, 60+ Moderate, below that Weak. A low score isn't a red flag about the merchant — it usually just means the submission was missing context.

Pricing behavior classification

Where enough data exists (30+ verified observations for a given offer), we classify its pattern as Stable, Dynamic (prices differ across bucket time, not fairness), Cross-Consumer Dispersion (people at the same moment saw different prices), or Dynamic and Dispersed (both). Below that threshold we say Insufficient Evidence rather than guess.

What this classification does not do: it never claims a price difference was caused by who the shopper is. Dynamic pricing does not imply personalization, and cross-consumer dispersion does not by itself establish why prices differed. A claim like that requires a purpose-built controlled study — holding every other variable fixed and comparing only the one thing being tested — which is a different, much higher bar than an observational classification.

Reference prices

When no verified crowd comparison exists yet for an offer, we may show a reference price instead — a price licensed from a third-party data provider (currently Oxylabs), fetched one product at a time by an admin, never automated or run on a schedule. A reference price is never counted as a verified observation, never blended into "others right now" or "price history" statistics, and is always shown with its own label, source, and fetch time. It exists only to give a first-time comparison something to measure against — it is not a person's own report.

What we will not do

• We do not scrape merchants ourselves without authorization — see our published AGENTS.md standing rules.
• We do not fetch arbitrary URLs a user pastes in.
• We do not store merchant credentials, cookies, or payment details.
• We do not silently revise a published research result — corrections are recorded, not hidden.
• We do not present a reference price as a verified observation, or blend the two together.