Business

The federal contract register counts every amendment as a contract

Found 3 September 2026, published 4 September 2026, figures restated 5 September 2026 · Danilo Dzyuban, Ottawa · source: Proactive Disclosure — Contracts over $10,000, Government of Canada open data, 1,313,540 rows as downloaded 2 September 2026

The Government of Canada publishes every contract over $10,000 as open data. Anyone who adds up the published rows arrives at a market twice its real size, because the register records each amendment as a separate row carrying the whole value of the contract after amendment.

What the register publishes

A contract enters the register once when it is signed. Each time it is amended, a new row is published. The field contract_value in an amendment row holds the total value of the contract after the amendment, not the amount added by it. The field instrument_type marks such rows with A. Nothing in the dataset documentation says that these rows should be excluded from a sum, and no field carries the increment alone.

One contract, four rows

A contract signed by ADGA Group Consultants on 5 October 2011 appears in the register four times:

Added together, the four rows come to $1,088,662,391. The contract is worth $348,625,938, the value in its last published row. Every intermediate row is a snapshot of the same contract at an earlier date. A third defect sits in the same four rows: the 2011 row names the supplier as ADGA Group Consultants Inc., the three later rows as ADGA Group Counsultants Inc. The rule below collapses the three later rows to one; the 2011 row stands apart on its spelling, so the collapsed register still carries $451,101,569 for this one contract.

The scale

The sum of the rows overstates the real total by 102 per cent. Put the other way, 51 per cent of the summed value is the same money counted more than once. Under the most generous reading of the data, described under Method, the overstatement is 97 per cent. The effect is not uniform: a contract amended many times is counted many times, so the distortion is largest exactly where the money is largest.

A second defect underneath the first

Collapsing the rows requires a key that identifies one contract. The obvious key is the contract date. It fails, because for part of the register the field contract_date holds the date of the amendment rather than the date of the original award. One contract between Maverin Inc. and Shared Services Canada shows it:

Same original value, same department, same first day of work. One contract, and its current value is $21,470,000. A key built on the contract date counts it twice. The key that holds is the start of the period of work together with the original value and the contracting department.

Method

Rows are grouped by supplier name, contracting department, start of the period of work, with the contract date where the start is blank, and original value. A row that carries no original value joins the group whose original value equals its own contract value when such a group exists, and otherwise stands as published; 310,287 rows stand that way. Within each group the last published row, by reference number, is kept. The figures above are reproducible from the public CSV with that rule and nothing else. Two choices inside the rule move the total, both toward a smaller overstatement, and both are bounded. Keeping the highest published value instead of the last raises the collapsed total by 1.2 per cent, to $616,132,651,052. Where a department signs several contracts with one supplier on one start date at one original value, the rule cannot tell them apart and keeps one; that happens in 13,414 groups and hides at most $8,261,219,058. With both applied the collapsed total is $624,393,870,110 and the overstatement is 97 per cent.

What follows from it

The fix belongs to the publisher. Either of two fields would end the ambiguity: a stable contract identifier shared by an award and every amendment to it, or a field carrying the value added by the amendment alone. Until one of them exists, every figure computed from this dataset should state whether amendments were collapsed, and figures that do not say so should be read as inflated.

Every figure that Assay Systems returns to a client from this register is computed after amendments are collapsed to one contract per award, and the report says so on its last page.

Citation: Danilo Dzyuban, “The federal contract register counts every amendment as a contract” (4 September 2026).