The Griotta Desire Index

How the number is built

Every denominator comes from one document: the World Bank's commodity price sheet, the September 2026 edition, published on 2 September 2026. Using a single free, dated PDF for every commodity means the whole bottom half of this index can be checked against one source.

Those quotes are upstream commodities, and that distinction does real work. Cocoa beans are not chocolate; green coffee is not roasted coffee; wheat is not flour. Each object therefore carries a conversion with a source attached: 70 g of cocoa mass needs about 87 g of beans once shell and moisture come off, and 250 g of roasted coffee starts as roughly 298 g of green beans. Two of the seven ranked objects compare like with like on material, differing only in grade and origin. The other five need a processing yield, and are labelled accordingly.

Every numerator comes from observed French shop prices recorded in the Open Food Facts Open Prices database: real prices, each with a date and a shop, not list prices or estimates. Each distinct product counts once, however often it was recorded, so a heavily photographed barcode cannot set the price of a whole category. The published count is distinct products, not sightings.

Dollars become euros at a single European Central Bank reference rate, dated 22 September 2026. One rate, one date, applied to everything.

This is a snapshot, not a live index. The V1 snapshot is 22 September 2026. Commodity prices are the August 2026 monthly averages from the sheet published on 2 September 2026, with low and high bands from the five most recent periods it prints. Retail prices are French shop prices observed in the 180 days to 22 September 2026. The exchange rate is the ECB reference rate for 22 September 2026. Nothing here updates on its own, and cocoa in particular has moved a long way in two years, so the date matters as much as the number.

What the Material Multiple does not mean

It is not markup. Markup compares a selling price to what the seller paid for the finished goods. This compares a selling price to a commodity benchmark several transformations earlier in the chain. A shop buying that chocolate bar at wholesale is nowhere near the cocoa bean price, and the two numbers should never be quoted as if they were comparable.

It is not margin or profit. Everything between a commodity and a finished object has to come out of that distance: manufacturing, labour, tooling, formulation, design, certification, packaging, logistics, tariffs, distribution, retail rent, returns, warranty, marketing, and the brand and convenience that made someone pick this object rather than another.

It does not mean an object is overpriced. A high multiple means an object has travelled a long way from its inputs. Travelling a long way is what manufacturing is. Coffee at 2.4× and a baguette at 25× are not a bargain and a scandal; they are two different amounts of processing, packaging, distribution and shop.

It is not the cost of the materials a manufacturer buys. These are benchmark quotes for bulk lots at a named delivery point, and they exclude processing, refining, freight into Europe, grade premiums and contract terms. They generally sit below what a maker pays for a usable processed ingredient, because they are priced earlier in the chain, but we do not claim a fixed direction for every object: grade, variety, origin and yield pull different ways. Where an object has a bias we can name, the per-object notes say so.

It is not every material in the product either. Packaging is excluded from the denominator for every object, uniformly, even when it is a large part of what you carry home. Commodity-input value is not the same thing as all the physical materials required to put an object on a shelf.

Three objects, worked through

Ground coffee, 250 g — 2.5×. Among the lowest multiples in the index, and the reason is mass. Roasting burns off water, so a 250 g pack starts as roughly 300 g of green beans, and arabica is expensive. The coffee in the bag is most of what you are paying for. Median observed price €5.08 across 70 shop prices; estimated beans €2.07.

Dark chocolate, 100 g — 6×. A 70% bar carries about 70 g of cocoa mass, which needs roughly 87 g of beans once shell and moisture are removed. Cocoa has been historically expensive lately, which pulls this multiple down compared to where it would have sat two years ago. The same bar at 2024 cocoa prices would look considerably more dramatic, which is a good argument for printing the date on everything.

Black tea, 100 g — 21×, and not in the table. Tea would top the ranking, and it is the best illustration of what the number leaves out: a box of tea bags is leaf plus envelopes, tags, string, staples, a carton and a film, and the packaging can outweigh the tea. But it rests on only 10 distinct products, below our threshold of fifteen, so it is reported here and kept out of the ranking. The highest number in the project is the one we are least sure of, which is usually how it goes.

What sits between the material and the price

Reading down the table, the pattern is not about greed. It is about how much has to happen to the commodity. Coffee and sugar are close to their inputs: roast, grind, pack. Pasta and bread involve milling, mixing, shaping, drying or baking, and a distribution chain that must handle a perishable product. Chocolate adds fermentation, roasting, conching, tempering and a cold chain.

It also explains why coffee sits at the bottom: green coffee is expensive, roasting only removes water, and the bean in the bag is most of what you are buying. Wheat is cheap, and a baguette is milling, mixing, proving, baking and a delivery round before seven in the morning.

Then there is the part that is not manufacturing at all: the design, the name, the wrapper, the shelf position, and the reason a shopper reaches for one bar rather than the one beside it. That is the part this project finds interesting, and it is precisely the part the arithmetic cannot separate out. The multiple bundles the factory and the story together and refuses to say which is which.

It is also worth saying plainly that none of this is an argument against buying things. Paying for a well-made object, or for one you simply like, is not a mistake that needs correcting.

What v1 leaves out, and why

Sixteen objects were specified; seven are ranked. A plain cotton T-shirt and a plain silver ring have clean denominators — cotton and silver are both commodity rows — but their retail side could not be sampled reproducibly, because mainstream French clothing and jewellery sites block automated access and what does load is marketplace listings with varying specification and seller.

Four more were dropped on the denominator, after checking what each commodity quote physically is. White sugar: the benchmark is raw sugar imported into Europe, not the refined sugar in the bag. Dried pasta: durum wheat has no benchmark row and bread wheat is the wrong species. Milk chocolate: a fifth of the bar is milk powder, which has no row at all. Orange juice: the benchmark is navel table fruit, not the varieties pressed for juice. Pasta and milk chocolate had the two largest retail samples in the whole project and were rejected anyway.

Corn flakes and bar soap were dropped for a duller reason: after filtering, one and three usable prices survived against a floor of ten distinct products. Black tea sits just outside the ranking on the same principle: the method is sound but ten products is thin, so it is shown separately rather than ranked.

Which leaves a first version that is mostly groceries. That is not a claim about consumer products in general; it is simply where a reproducible France-and-euros pipeline could be built. Every candidate and every rejection is listed in the working, so nothing quietly disappeared for having a boring ratio.

Why we are counting this at all

Griotta is a pause you put in front of your own shopping apps, so a purchase gets decided rather than completed. That work keeps running into the same question from a different angle: an object arrives in front of you already carrying a price, and the price is the one thing about it you are not invited to examine.

This index is one attempt to examine it, with a deliberately narrow question and a deliberately boring method. It sits beside our work on how buying became faster than deciding and on velocity bias. Knowing that a bar of chocolate is 6× its cocoa will not stop anyone buying chocolate, and it should not. It is just a fact about the object that nobody was going to hand you at the till.

Method and sources

• World Bank Commodities Price Data (the “Pink Sheet”) — September 2026 edition, published 2 September 2026. Every commodity quote, with low and high bands taken from the five most recent periods in the same sheet, and with what each row physically represents recorded per object.
• Open Food Facts — Open Prices — observed French retail prices, filtered to France and euros, within 180 days, accessed 22 September 2026, aggregated one product at a time.
• European Central Bank euro reference rate — 22 September 2026, 1 EUR = 1.1463 USD.
• Conversion assumptions. Cocoa bean to usable nib, 0.78–0.82: shell is 14–15% of bean weight (peer-reviewed analysis of cocoa bean shells), with roasting removing a further 4–5% as moisture. Coffee roast loss, 12–20% of green mass (Royal Coffee, Sweet Maria's). Soluble coffee, about 2.4 kg of green coffee per kg (stated in the background section of a published process patent) . Wheat to white flour, 72–78% extraction, with 75% the FAO convention. Baguette, dough at roughly 1.68× the flour weight from 65–68% hydration, less 10–13% oven loss. Rice and tea use no conversion: the benchmark and the product are the same material, differing in grade and origin.
• The data. Download the V1 data (CSV) — every published figure, including the unranked object, with its retail sample sizes, commodity-input range, multiple range and confidence. It is generated from the same dataset this page renders, so the two cannot disagree.
• Composition, quantities, conversion yields, the confidence rules, the candidate manifest and every rejected object are documented per object. The calculation is a script over four files: given the same inputs, it reproduces every number on this page exactly.

Found something wrong? That is the point of publishing the working. Tell us and we will correct it in v2.