Choose your data sources

Pick one anchor source that defines who exists in your market, then add others only where the anchor is blind. Teams that buy a large database and treat it as the market inherit its blind spots without knowing where they are.

The source column of your criteria sheet tells you what you need. This chapter compares the options, tests them on accounts you know, and ends with a one-page source register. We stay with our invented payroll software vendor for French multi-site employers.

What each type of source is good for

Source	Gives you	Strong on	Blind on	Cost	Freshness
Company registries	Legal entity, activity, headcount band, sites, status	Near-complete population of legal entities	Website, stack, buyers	Free or low for bulk files	Regular official updates, headcount lags
LinkedIn and Sales Navigator	Company page, size, industry, people	White-collar sectors, tech, services	Trades, small local firms, inactive pages	Subscription for advanced filters	Depends on what companies maintain
Directories, federations, exhibitor lists	Members of a precise niche	Precision on a vertical	Scale, format, updates	Free or membership, plus cleanup time	Uneven
Maps	Physical places with phone, website, rating	Local businesses	Groups, head offices, online-only firms	Low per place	Fairly current, uneven by country
Technology data	Tools detected on websites	Markets defined by stack	Internal software, companies with no site	Low to high	Depends on crawl frequency
Commercial B2B databases	Ready-made records with filters	Speed, large companies	Local firms, small companies, some countries	Licence, often annual	Varies, ask for the update policy
Your own data	Customers, lost deals, event lists	Accuracy on what you’ve seen	Everything you haven’t seen	Already paid for	You control it

In France, the registry is the free SIRENE base. Other countries have their equivalents, such as Companies House in the UK, with very different levels of detail. Don’t assume what SIRENE gives you exists elsewhere.

Choose the anchor from the shape of the market

Your market	Anchor	Complements
Companies with a clear registered activity	Registry	Website matching, LinkedIn for size
Tech and services, active online	LinkedIn or Sales Navigator	Registry for legal identifiers
Local businesses, retail, trades	Maps	Registry, directories
Defined by a technology	Technology data	LinkedIn for size and people
Narrow vertical with a professional body	Directories	Registry to fill in the rest

For our vendor, activity codes and site counts were the filters that held up in the back-test, so SIRENE is the anchor. Maps can catch sites run under an unexpected legal structure, and federation member lists add groups that a holding code hides. When LinkedIn can’t help at all, the niche and local market chapter goes further.

Test coverage on 30 accounts before you pay

Take 20 customers and 10 accounts you would love to sign. Look for each one in each source and record what you find. Vendors’ overall record counts tell you nothing about your segment, and most will give you a sample if you ask.

Source (invented results, of 30)	Found	Correct activity	Size present	Website present
SIRENE	30	26	24	0
LinkedIn	24	18	24	22
Maps	21	19	0	17
Federation lists	9	9	0	6
Commercial database A	16	12	15	15

With one column per test per source, a formula gives the rate:

=COUNTIF(C2:C31,"yes")/COUNTA(A2:A31)

The sample is biased toward companies you already reached, so real coverage of the market will be lower. Use the test to compare sources with each other. Here the commercial database misses nearly half the known accounts, which is a poor basis for an annual contract in this niche, even if the same database may do well for mid-market software companies.

Compare on cost per usable account

The price per row misleads, because rows differ in how many survive your filters. Divide the total cost, including people’s time, by the accounts left at the end.

cost per usable account = (licence + extraction + time × day rate) / usable accounts
usable accounts = rows × share passing hard filters × share with a website after matching

Route (invented)	Rows	Pass filters	Have a website	Usable	Total cost	Per usable account
Registry file, loaded by hand	6,000	30%	70%	1,260	€1,000 (2 days at €500)	€0.79
Commercial list	2,500	60%	95%	1,425	€2,000	€1.40
Maps searches	1,500	40%	80%	480	€650 (€150 + 1 day)	€1.35

The ranking can flip with your context. If nobody on the team can load a large file, the registry’s real cost rises fast.

What the same routes look like in Derrick

I co-founded Derrick, so read this table knowing that. It shows how each source can be reached from one tool, billed in credits.

Source	Feature	Input	Output	Credits
French registry	Import Companies by NAF Code	NAF codes, optional département	Legal name, SIREN/SIRET, open establishments, activity, headcount band, creation date, address, director	1 per company
LinkedIn filters	Import LinkedIn Companies from Sales Navigator	URL of a Sales Navigator search	Company data	1 per company
LinkedIn by description	Import Companies from a Prompt	One sentence describing the companies	Name, industry, country, city, website, LinkedIn URL	1 per company with your own Sales Navigator connected, 10 otherwise
Maps	Google Maps Scraper	What and where, in plain words	Up to 200 places per search: name, phone, website, rating, address	1 per place, paid plans
Technology	Find Companies by Tech Stack	Technologies or site type	Domain, site type, country, technologies	1 per company found

Some limits worth stating. The NAF route is for France only. The Sales Navigator import needs your own Sales Navigator account. Anything reached through Maps is capped per search, so large areas must be cut into smaller ones. And for a nationwide pull of SIRENE, if someone can run SQL on the free file, the file costs less than any tool, mine included. The tool earns its credits when you want a few codes or départements quickly, in a sheet, without handling a multi-gigabyte download.

Check terms and data protection, source by source

Each source has its own licence or terms of use, and they change. Read them for the use you plan, including automated collection, which platforms such as LinkedIn address in their user agreements. Registries can also carry restrictions: SIRENE includes a dissemination status, and some people have asked that their data not be shown or used for prospecting, so keep that field and respect it.

As soon as a record names a person, data protection rules apply. In France, start with the CNIL’s guidance on commercial prospecting, and have counsel confirm your setup, since this guide isn’t legal advice. Opt-outs and retention are covered in keeping the TAM up to date.

Fill in the source register

Source	Role	Fields used	Coverage test	Cost per usable account	Terms checked	Update plan
SIRENE	Anchor	Activity, band, sites, status	30 of 30	€0.79	Date, reader	Quarterly
Maps	Complement	Website, phone, site name	21 of 30	€1.35	Date, reader	Twice a year
Federation lists	Complement	Member names	9 of 30	To estimate	Date, reader	Once a year

Your next step: run the 30-account test on your two or three candidate sources this week, choose an anchor and fill in the register. Then move to building the company list.