# Canadian Software Procurement Notice Pilot Methodology

Version: 0.1.0

Generated: 2026-08-28T02:16:03.695Z

## Purpose

This pilot tests whether public CanadaBuys notice data can support a reproducible software-project readiness benchmark. It is not a census of Canadian software projects, a cost study, or a claim about every requirement in an RFP.

## Source frame

The pipeline parsed 16,203 rows from the official CanadaBuys 2024-2025, 2025-2026, and 2026-2027 tender-notice CSV files, then deduplicated amendments by notice reference.

## Selection

Notices were deduplicated by reference number, matched against development-oriented title and description patterns, screened for product-only and equipment exclusions, ranked by classifier confidence, and limited to 100 records. The resulting 100-notice sample is purposive and is not statistically representative.

## Requirement detection

Requirement flags use documented keyword patterns against English title, classification, and notice-description fields. Counts indicate explicit mentions only. Detailed requirements may appear solely in attachments, so a missing flag must not be interpreted as proof that a requirement was absent.

## Exclusions

The pipeline excludes clear physical-equipment notices and product-only licence, subscription, or renewal notices when the title does not indicate development work. Contact names, email addresses, phone numbers, and street addresses are not published.

## Reproducibility

The source manifest records the official URL, fiscal year, byte size, row count, and SHA-256 hash for each input file. The public record file contains only the fields needed to audit selection and requirement flags.