Method Note: Two Corrections to Our SEAO Analyses (July 2026)

Method note, July 22, 2026.

We publish analyses based on SEAO open data. When a figure we have already published turns out to be inaccurate, we correct it and we say so, rather than changing it silently. On July 22, 2026, a revision of our calculation methods led to two corrections. This note explains which ones, why, and what they change, figure by figure.

The overall thrust of our analyses does not change: in Québec, a clear majority of public contracts are awarded without a call for tenders by count, but these contracts remain a minority by value. What these corrections refine is the exact size of each share.

Correction 1: deduplicating export releases

What we were doing. SEAO periodically republishes the history of its contracts. In assembling these successive exports, our corpus contained, for a portion of the contracts, several copies of the same contract, drawn from exports of different dates. Our deduplication did not catch all of these duplicates, because the copies carried slightly different time zone labels.

Why it was a problem. These duplicates inflated the cumulative amounts (the same contract value counted two or three times) and, in our count of municipal contracts, the number of contracts itself. In our other analyses, the counts by contract were not affected (we count each contract only once, by its unique OCID identifier); the dollar totals, however, were overstated.

What we do now. Our deduplication now groups the copies of a single contract regardless of their time zone labels, keeping the most recent version. An automatic check verifies, at each update of the corpus, that no duplicate of this kind remains.

Effect. The value totals go down. For example, the total awarded value in our supplier-concentration analysis goes from $284 billion to $213 billion; the recorded municipal market goes from 160,664 to 132,106 contracts and from $84 billion to $71 billion. The proportions move by at most one percentage point; the top ranks of the rankings hold, but a few positions swap lower down in the municipal ranking, reordered accordingly.

Correction 2: mandated purchases are calls for tenders, not mutual agreement

What we were doing. We placed "mandated purchases" and "groupings of bodies" (joint purchasing) on the side of contracts awarded without a call for tenders.

Why it was an error. A mandated purchase is a body (often the Union des municipalités du Québec, the Fédération québécoise des municipalités, or the Société québécoise des infrastructures) running a call for tenders on behalf of several bodies that buy together. SEAO publishes them as an open procedure, and most receive several bids (78.3% count at least two). It is a competitive process, not a mutual-agreement award. Counting them as mutual agreement therefore overstated the share awarded without a call for tenders.

We do not claim, however, that every mandated purchase guarantees effective competition: a minority receive only a single bid, and some groupings of bodies may cover procurement from a single supplier. We classify them according to the award method published on SEAO (an open procedure), which is the verifiable criterion, without ruling contract by contract.

What we do now. Mandated purchases are counted among the contracts awarded competitively. This definition is now the same across all our analyses.

Effect. The share of contracts awarded without a call for tenders decreases, especially by value. Over 2021-2023, it goes from 60.6% to 58.3% by count (nearly 6 contracts in 10) and from 33.5% to 24.2% by value (barely 1 dollar in 4). In other words, the value that goes through a call for tenders rises from 66.5% to 75.8%, more than 3 dollars in 4.

Summary table of the published figures that changed

This table covers the main statistics of each analysis; the derived values (amounts by supplier, by city or by sector) are recomputed in each article using the same method. All these figures are recomputed against the corpus at runtime.

What does not change

The core finding holds: by count, a clear majority of public contracts are awarded without a call for tenders; by value, most goes through a call for tenders. The breakdown by sector and the annual trends are not affected by these corrections; in the supplier rankings, the top ranks hold, but a few positions lower down swap after the amounts are recomputed. The share awarded by strict mutual agreement (excluding mandated purchases) was already computed correctly and does not move.

Method and scope

Our analyses count by procurement process (one contract equals one OCID identifier, never summed across its lines), over dated windows, from SEAO open data in OCDS format. The classification of award methods relies on the field published by SEAO, and not on a contract-by-contract interpretation. Corpus as of July 22, 2026. We have also made our calculations fully reproducible, identically from one run to the next, which was not guaranteed for a small fraction of contracts carrying several lines on the same date. This note also replaces the July 14, 2026 box, which flagged the move to a display recomputed against the corpus and the correction of one amount in the ranking (a spelling variant of Cardinal Health Canada counted separately); that correction remains built into the figures shown.

If you spot an error in one of our figures, write to us: the reliability of these analyses is what gives them their value.

Sources