What this food price data is a price for

A price is never just a price. It is a price for a particular transaction, between particular parties, for a particular quantity. Two figures for the same vegetable on the same day can differ by a factor of three and both be correct, because they describe different transactions. That is the single most useful thing to know before reading anything on this site, so it is worth being exact about which transactions are represented here.

Three economic levels, and none of them is retail

Each stored observation carries an economic level, taken from the provider rather than inferred. On the reading below there were three levels in the data, and retail was not one of them.

Wholesale is the level a distributor sells at, and it is the one closest to what a kitchen pays a supplier, though it is not the same number. Producer is the farm gate: what the grower received, before packing, freight, storage or anyone's margin. Index is not a price at all, which the next section takes up. The site keeps the three apart everywhere, and the methodology page states the rule plainly: producer, wholesale, retail, CPI and index signals are not blended without an explicit conversion model, and there is no such model here.

The index rows are not prices

An index is a ratio against a base period. It moves like a price and it is not one: it has no currency, no unit and no amount you could pay. The Australian series is the only index source configured, and its rows are stored with no currency for exactly that reason. A percentage change on an index is meaningful. A dollar figure derived from one is not, and the site does not offer to compute you one.

Five currencies, and no conversion between them

Currency totals are never converted between currencies here, because the data carries no exchange rate. A converted figure would need a rate, a rate date and a rate source, none of which these providers publish alongside their prices, and inventing one would set an unattributed number beside a sourced one. So currency is a dimension a comparison has to match on, not something the site normalizes away. For a cross-currency read, use the percentage movements, which are dimensionless, rather than the levels.

A movement needs two comparable observations

The site groups observations into series before it calculates anything, and two observations are in the same series only when all seven of these match: country, provider, item name as the provider writes it, economic level, market, unit and currency. Change is then the latest value against the one before it, inside that series and nowhere else.

The last row is the interesting one. For those series there is a current value and nothing to compare it against, so the card shows the value and no percentage. That is a deliberate empty state rather than a gap waiting to be filled: the alternative is comparing across a boundary the provider did not intend, which produces a number that looks like information and is not.

Questions this data can settle

• Has the wholesale quote for a named item, in a named pack, in a named country, moved since the last time that source published?
• How old is the newest observation for an item, and how old is the source's publishing cycle willing to let it get?
• Which official body published a given figure, and at what URL?
• Is a movement large relative to the recent history of that same series?

Questions it cannot

• What a consumer pays in a shop. There are no retail rows.
• What your supplier will quote you tomorrow. A published wholesale figure is an official observation of a past period, not an offer.
• Whether food is cheaper in one country than another in absolute terms. That needs a currency conversion and a unit conversion, and this site performs neither.
• What a movement means for your own costs, unless you match the item, level, unit and currency to something you actually buy.

None of this is a complaint about the sources. Wholesale and producer series are what national agencies publish for free, on a schedule, with provenance, and that is a genuinely useful thing to have. It is only worth being precise about which question a figure answers, because the cost of the confusion falls on whoever acts on it. The understanding the analytics page covers what the change, confidence, impact and priority figures each mean once you have the right series in view.