Estimate a software project budget (France)

If the user passed arguments, they are: $ARGUMENTS

Give a range with its sources and assumptions, never a single magic number.

Language: answer in the user's language. Translate the section headings of the output format into that language.

Workflow

1. Clarify the scope in a few questions if missing: project type, users, number of processes/modules, integrations, data migration, platforms (web/iOS/Android), hosting constraint, deadline. Map the need to a project_type: erp, mobile_app, digital_transformation, gpao_mes, digital_twin. AI projects have no reliable public dataset: estimate them from effort (step 3) and say so.
2. Market data — if the prix-logiciel MCP tools are available:
   • price_statistics for the project type (and nature: creation, redesign, maintenance) → median and quartiles of attributed public contracts;
   • private_market_prices → verified private figures (surveys, published grids) with URL and date;
   • daily_rates → reference day rates by profile;
   • search_public_contracts → 3 to 5 comparable contracts to show concrete examples. Public procurement amounts are often multi-year framework ceilings that include maintenance: say so, and prefer the "creation" nature and monthly figures when comparing to a one-off build.
3. Effort-based cross-check: break the scope into blocks (framing, each module/process, integrations, migration, acceptance, training, project management ≈ 15–20 %), estimate days per block (low / likely / high), multiply by a market day rate from daily_rates. Show the table.
4. 5-year total cost: build + recurring (maintenance, hosting) + third-party licences (users × list price × 12 × 5, dated, "may change"). Unknown third-party costs are "to be confirmed", never 0.
5. Present the result.

Output format

## Estimated range
Low – likely – high (excl. VAT), and the main driver of the spread.

## How it was built
Effort table (block, days low/likely/high, rate) and total.

## Market references
Public contracts: n, median, Q1–Q3, period, source (DECP, Etalab licence), with the caveats.
Private market: figure, scope, source URL, date.

## 5-year total cost
Build / recurring / licences / total.

## Assumptions and what would move the price
Numbered.

## Next step
What to write in the specification to get comparable quotes.

Principles

• Cite every figure with its source and date. If a figure is not in the data, do not invent it.
• Respect the dataset's publication floors: when a statistic is withheld because the sample is too small, say "not enough contracts to publish a median" rather than computing one yourself.
• Framing before quoting: on a complex scope, recommend a paid framing phase whose deliverable makes the build priceable.

Maintainer

Maintained by JAIKIN — AI, Odoo and custom software, Strasbourg (France). This section is directory metadata, not part of the answer.