All-In Price Comparison Engine

Why `Math.min` is wrong

Two listings: $80 and $95. The $80 quotes pre-fee and charges 30% at checkout, so it is really $104. Ranking the raw numbers doesn't just misprice them — it puts them in the wrong order, in the direction that costs the buyer money. That is the bug this exists for, and it is the default behaviour of almost every price-comparison feature.

Unknown fees mean unrankable, not zero. A source that quotes pre-fee but publishes no fee model gets excluded with a stated reason. Treating it as all-in ranks it first and is wrong by exactly the fees — a confident, wrong winner. The comparison reports itself as *partial* so the UI can say "cheapest of the 3 we could price in full" rather than "cheapest". That distinction is the whole difference between informing someone and misleading them.

Fees are not one percentage. Percentage (often capped) + per-ticket + per-order + delivery, then tax on the total. A flat "add 25%" is wrong at both ends of the range, and a cap can flip which marketplace wins.

Per-ticket and per-order are different numbers. A big per-order fee makes a source lose at quantity 1 and win at quantity 5. The cheapest source genuinely depends on how many you're buying, so quantity is an input, not an afterthought.

A listing you can't buy is not a price. Split rules (sellers refusing to strand a single seat), minimum quantities, and insufficient inventory are excluded *before* ranking — not ranked and then failed at checkout. Notably, these exclusions do not mark the comparison partial: that listing genuinely can't be bought, so nothing cheaper is being hidden.

Never convert currencies without an explicit, timestamped rate. Missing or stale rate is a refusal, never a conversion at 1.0.

A stale price is not a price. A feed that goes quiet otherwise keeps winning the cheapest slot forever with a number that no longer exists.

Verified by mutation

Every rule had its guard removed one at a time and the suite was confirmed to fail — 11 of 11, including the headline case (ranking raw prices across mixed fee semantics) and both directions of the partiality flag: claiming completeness when something was hidden, and over-warning when nothing was.

36 tests passing · typecheck clean · 6 deliberate defects applied to the real source, all 6 caught. The set ships in mutations.json, including a staleness boundary a sweep found untested.

Shape

Pure functions over plain objects — no network, no database, no clock. Time is always passed in, so results are reproducible. Money is integer minor units throughout; floats do not belong in a checkout total. Built for ticket resale, but nothing in it is ticket-specific: it applies to any market where sources quote fees differently.

Delivery

Private repository invite within 24 hours. Single-product commercial license: use and modify in any number of products; no redistribution or resale of the source.

Interface

What you call, and what comes back. Types and signatures only — the implementation ships with the source.

  export function computeFees(subtotalCents: Cents, quantity: number, fees: FeeModel): Cents;
  export function purchasableAt(q: Quote, quantity: number): ExcludedQuote["reason"] | null;
  export function normalizeQuote( q: Quote, opts: CompareOptions, ):;
  export function compareQuotes(quotes: Quote[], opts: CompareOptions): Comparison;
  export function explainComparison(c: Comparison, opts: CompareOptions): string;

  export type Cents = number;