Scope calibrates its estimates against up-to-date French public data. The sources and methodology are published in the open, with no dependency on an opaque proprietary dataset.
01
Four families of public data feed Scope's medians: IT public procurement, business demographics, compensation benchmarks and sector lexicons.
≈ 200,000 IT/digital records (CPV code 72* filter)
French public contracts above €25,000 net, published under the ODbL licence via data.gouv.fr. Filtered on CPV code 72 family (IT services) and cross-checked with codes 48 (software) and 79 (business services) for project margins.
French business registry (daily updates)
SIRET / NAF reference used to qualify the end customer (company size, industry) and correlate person-day medians with the IT services / scale-up / enterprise segment.
Cigref, Apec, Talent.io (yearly / quarterly publication)
Three public benchmarks cross-checked: Cigref (per-day rates IT services / large-enterprise IT), Apec (FR IT executive salaries), Talent.io (FR tech freelancers). Weighted by seniority (junior / mid / senior / lead) and job family (dev, ops, data, AI, PM, design).
6 sectors: insurance, banking, energy, industry, services, other
Domain vocabulary extracted from public requirements documents per sector, used for automatic sector detection and prompt adaptation to sectoral jargon.
02
The calibration pipeline is deterministic and auditable: no opaque proprietary model, just statistical aggregates on public sources.
Exclusion of non-IT contracts (CPV code outside families 72/48/79), of outlier amounts (>3 standard deviations) and of incomplete records. Deduplication by public-contract identifier.
Calculation of percentiles p25, p50 (median), p75 by CPV code × INSEE headcount bracket × administrative region. Minimum sample of 30 records per cell required for publication.
Person-day medians from public contracts are cross-checked against per-day-rate benchmarks (Cigref / Apec / Talent.io) to validate budget-envelope vs project-duration consistency.
Quarterly update cycle for DECP (open publication) and semi-annual for compensation benchmarks. Reference version displayed in every estimate produced.
03
The UK competitor ScopesLab and generalist US tools do not have the same granularity of access to French public data. Three structural barriers.
The official DECP API on data.gouv.fr applies rate limits dedicated to FR-resident IPs. A non-FR actor must reconstruct the corpus via batched bulk downloads, without real-time incremental updates.
FR sector vocabulary ("assistance à maîtrise d'ouvrage", "PCA / PRA", "COMEX / CODIR", "DGS / DGA") is not translated in US/UK datasets. An anglophone actor captures the PM brief at AXA / BNP / Société Générale poorly.
Cigref, Apec and Talent.io do not publish international grids. A UK tool applying its own medians (Stack Overflow Salary Survey, Glassdoor) is off by 15-30% on FR IT-services rates.
04
Four sources, their weight in the calibration, their licence and their refresh cadence.
| Source | Scope | Licence | Refresh |
|---|---|---|---|
| DECP v3 (data.gouv.fr) | ≈ 200,000 IT public records | ODbL | Quarterly |
| INSEE SIRENE | FR business registry (size, NAF) | Open Licence 2.0 | Daily |
| 6 sector lexicons | FR domain vocabulary per sector | Internal (public corpus) | Semi-annual |
| Cigref / Apec / Talent.io rates grid | FR IT compensation by seniority | Citation allowed | Semi-annual |
05
Detailed methodology note: filters applied, percentiles computed, cross-checks.