Methodology

How the benchmark is built

Last updated 31 October 2026. An independent, anonymous benchmark of what roasters pay for green coffee, normalised to US dollars per pound.

What we ask

The survey collects one recorded transaction per coffee. Section 1 captures the roastery once — where it roasts, its local currency, its annual purchasing-volume band and how it primarily buys. Sections 2–6 capture each coffee: which of the benchmarked coffees it is (a fixed set — Brazil Santos, Brazil Cerrado, Colombia Supremo, Kenya AA, Costa Rica SHB, Guatemala SHB, Ethiopia Gr 1 and Indonesia Gr 1 — from which the origin, region and grade follow); coffee type, processing method and cupping band; the purchase date, how the coffee was priced (including the C-market contract month where the price tracks the C market), the price, the currency and the unit; a single fixed pricing basis — the price ex-seller's-warehouse at origin, with ocean freight, duty and inland transport excluded; the quantity band and lot structure; and, optionally, commercial details such as whether the coffee was bought direct, whether the price was negotiated, contract length and hedging.

Normalisation to USD per pound

Every quoted price is converted to a single comparable figure — US dollars per pound of green coffee — in three steps, all priced as of the recorded purchase date rather than the date the survey was entered. First, the quoted unit is converted to kilograms using its defined size (a pound is 0.45359237 kg, a 60 kg bag is 60 kg, and so on; a custom unit supplies its own size). Second, the local price is converted to US dollars using the exchange rate for that purchase date. Third, the result is expressed per pound, the convention coffee differentials are quoted in. The rate, its date and its source are recorded with every transaction, so no figure is presented as more authoritative than the data behind it.

Exchange rates

To price a transaction on its purchase date, dated rates are sourced from a keyless public provider of historical daily rates (by default Frankfurter, which publishes the European Central Bank reference series); rates for a past date never change, so they are cached indefinitely. Live previews use a second keyless provider (by default open.er-api.com). If a dated rate cannot be retrieved, the latest available rate is used and the affected transaction is labelled as such — the substitution is never hidden.

Confidentiality and what is published

Individual transactions are never published. Aggregates are only published for a comparison group once it holds at least a minimum number of transactions (one hundred by default). Groups smaller than the threshold are withheld entirely, and the number withheld is reported rather than silently dropped. This k-anonymity rule is what stops any single roaster's purchase from being identifiable from the published figures. Supplier identities and company names are never collected with a transaction.

Medians, averages and ranges

Because coffee prices are skewed by a long tail of expensive micro-lots, each groups' headline figure is the median USD price per pound, alongside the average and the observed minimum–maximum range. Medians are more robust to a single outlier than a mean.

Differentials and the C market

Prices are normalised to USD per pound, the convention in which coffee differentials are quoted. Where a roaster reports a C-linked price, the survey records the C-market contract month, and the implied differential (cents over the C contract) is derived as of the purchase date: the ICE Coffee C settlement for that contract month, on or immediately before the purchase date, is subtracted from the reported price to give a figure in US cents per pound. Where the exact contract settlement is not retrievable for an older purchase, the front-month C for that date is used instead, and the transaction records which basis was applied. The settlement used, its trading date and its source are stored with every such transaction.

Data quality

Every field is validated on the server before it is stored: origins, currencies, units, dates and amounts are checked against controlled vocabularies and ranges, and implausible values are rejected. Each submission records a one-way hash of the submitting IP address for abuse prevention; the hash cannot be reversed to recover the address.

The founding offer

The first 100 verified contributors receive twelve months of Contributor Access at no cost. The seat is claimed atomically at the moment a first transaction is recorded, so the offer cannot be over-granted under concurrency, and it is applied automatically — no card, no trial timer.