datagrit › Guides › French Company Finder - Sirene Financials

How to search French companies by profit and revenue with the Sirene API

Build lists of French companies from the official Sirene register, filtered by net result, revenue, NAF code and location, using Python or cURL.

Published 2026-10-01 · uses the French Company Finder - Sirene Financials Actor

Say you sell accounting software to French SMEs in the Lyon area, or you are an investor looking for loss-making bakery chains that might be open to a deal, or a procurement team checking whether a supplier's SIREN still belongs to an active company. All three questions have the same shape: give me French companies that match an activity, a place, a size and a financial profile, as a table.

France publishes the raw material for this as open data, including revenue and net result for companies that file public accounts. What it does not give you is a ready-made margin, or a location filter you can trust without checking every row. That gap is what this post walks through: where the data lives, the traps in it, and how to get a clean, filtered company list into Python.

Where the data comes from

The Sirene register identifies French companies by a 9-digit SIREN number and each of their establishments by a 14-digit SIRET number. The state exposes it through a free public search API at recherche-entreprises.api.gouv.fr, and every company has a public page on annuaire-entreprises.data.gouv.fr. For companies that file their accounts publicly, the API also carries the latest revenue and net result, and it accepts revenue and net result ranges as search parameters.

Working with that API directly, you run into these issues:

• Location matches dead establishments. A department or postal code filter matches any establishment the company ever had there, including ones closed decades ago. In a sample of 450 companies returned for eight departments on 30 September 2026, 128 had only closed establishments in the requested department.
• The head office is not always in the matching list. You have to check it separately.
• The matching list is capped. By default the API returns only 10 matching establishments per company unless you ask for more.
• No employer filter. The API has no switch for "has employees"; you need to derive it from the head office employer flag and the employee band.
• One year of accounts. The register exposes only the latest published financial year, so growth rates are not available.
• Hard cap of 10,000 results per search, plus rate limiting with Retry-After.

The French Company Finder Actor on Apify wraps the API and handles these. It requests up to 100 matching establishments per company, checks the head office first, and re-checks every filter on every record. With Only active companies on (the default), a company is returned only if it has an open establishment in your location. It derives isEmployer, exposes the revenue and net result ranges as input fields and computes the net margin on each row.

Step by step

1. Define the input

Software companies with an open establishment in the Rhône department, revenue above one million euros, profitable only:

{
  "queries": ["logiciel"],
  "departments": ["69"],
  "activityCodes": ["62.01Z", "62.02A"],
  "minRevenue": 1000000,
  "minNetResult": 1,
  "onlyActive": true,
  "employersOnly": true,
  "includeDirectors": false,
  "maxItems": 100
}

Notes on the fields:

• queries are searched separately; leave them out to search on filters alone.
• sirens accepts SIREN or SIRET numbers. Number lookups ignore the filters and also return closed companies, which is what you want for KYB checks.
• minNetResult: 1 keeps profitable companies; maxNetResult: 0 finds loss-making ones.
• companySizes takes PME, ETI or GE; employeeBands and legalForms take INSEE codes (for example 5710 for SAS).
• headOfficeOnly: true requires the head office itself to be in the location.
• includeDirectors is off by default because names of directors are personal data.

2. Run it from Python

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")

run_input = {
    "queries": ["logiciel"],
    "departments": ["69"],
    "minRevenue": 1000000,
    "minNetResult": 1,
    "maxItems": 100,
}

run = client.actor("datagrit/french-company-finder").call(run_input=run_input)

for c in client.dataset(run["defaultDatasetId"]).iterate_items():
    if c.get("found"):
        print(c["siren"], c["name"], c["revenue"], c["netMarginPct"], c["matchedEstablishmentCity"])

3. Or with cURL

A KYB check on one company by SIREN:

curl -X POST \
  "https://api.apify.com/v2/acts/datagrit~french-company-finder/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"sirens": ["380059113"]}'

If a number is not in the public register, the run adds a status row for it (not charged) and names it in the status message. If the input has only numbers and none of them exists, the run fails, so a scheduled check cannot pass silently on a typo.

What you get

One row per company (or per establishment for SIRET lookups). The fields you will use most:

The example shows why the matched establishment matters: the company is headquartered in Paris but was returned for a Rhône search because it runs an open establishment in Lyon. Rows also include vatNumber, coordinates of the head office, establishmentsOpen, creationDate, isActive, isSocialEconomy, isEmployer and a sourceUrl to the public company page.

Recipes

Prospect list into a CRM, refreshed monthly

Schedule a monthly run per target region and NAF code, export as CSV and upsert into your CRM using siren as the key. Because every run reads the live register, companies that cease activity drop out of the next list when onlyActive is on, and new ones appear.

Supplier check before onboarding

Pass the supplier's SIREN through the cURL call above in your onboarding flow. Check isActive, closureDate, legalForm and vatNumber against what the supplier gave you. A SIRET lookup returns exactly that establishment, including whether it is still open.

Profitability screen in pandas

import pandas as pd

rows = [c for c in client.dataset(run["defaultDatasetId"]).iterate_items() if c.get("found")]
df = pd.DataFrame(rows)

with_accounts = df.dropna(subset=["revenue", "netResult"])
print(f"{len(with_accounts)} of {len(df)} companies publish accounts")

by_band = (
    with_accounts.groupby("employeeBand")
    .agg(companies=("siren", "count"), median_margin=("netMarginPct", "median"))
    .sort_values("companies", ascending=False)
)
print(by_band)

Limitations

• Financials only for companies that publish accounts. Many companies do not publish their accounts, so revenue and netResult are empty for them, and revenue or net result filters exclude them. The status message reports how many returned companies have revenue.
• One financial year. The register exposes only the latest published year, so there are no growth fields.
• 10,000 results per search. Split very broad searches by department or activity code.
• Establishment count cap. matchedEstablishmentsListed of 100 means 100 or more.
• Opted-out companies are absent. Companies that chose not to appear in the public listing are not in the register the API serves.
• Directors are personal data. They are returned only when you switch includeDirectors on, and their use is your responsibility under data protection law.

The Actor reads only the open data the French state publishes through its public API; it does not log in or bypass access controls.

Wrapping up

Profit filters and location checks you can trust turn Sirene into a usable prospecting and KYB source; pricing is pay per result, and the Actor is on Apify with the full field list in the documentation.

More guides