How we produce a study

A public contract leaves two traces: the notice that announces it, and the award that says who won it and at what price. Depending on the country, those traces live in a national register, in the Official Journal of the European Union (TED), or in both.

TED only publishes what exceeds the European thresholds. Below those thresholds — the vast majority of contracts by count — only the national register exists. We read both whenever both exist.

Three reasons, and none of them is a collection failure on our side.

The register is recent. A country that opened its portal in 2022 has no history before 2022. Nobody can manufacture it.

The country only publishes one side. Romania, Portugal and Lithuania publish awarded contracts, not calls for competition. The Netherlands, Bulgaria and Slovakia do the opposite. That is not a hole in our data, it is the source.

The first years are sparse. Many registers carry a few isolated records long before they were really in service — Lithuania has one dated 1990. So we distinguish the first record found from the year coverage becomes continuous, and we only announce the second.

See the depth of history, country by country →

A register is not a clean database. The same buyer appears under five spellings, an amount is sometimes entered in cents, a lot is republished three times, a classification code is missing.

Before any measurement: buyer and contractor names are normalised, the lots of a single procedure are grouped, outliers are set aside, CPV codes are matched to their label. That work is thankless, and it is what decides whether the result is right.

This is the point most often misunderstood. History can be read at once: the data is already there. But measuring today's demand — what your market is worth right now, not three years ago — means letting an observation period run from the moment the study starts.

How long depends on the sector and on the company. A trade where buyers publish every week can be measured over a few weeks. A trade where three contracts come out a year needs several months: over a shorter window, a single tender shifted by a fortnight makes the result double or halve.

We set that window with you when we frame the study, and we tell you what it allows you to conclude — and what it does not.

When a piece of data does not exist, it is marked absent, never estimated. An unknown volume shows as « — ». Partial coverage is announced as partial. A figure we publish must be traceable back to the source register.