Data Vendor Evaluation: A Scorecard for Company Data

The best way to evaluate a data vendor is to score every candidate against your own list of companies, weighted by the decision the data will support. Do not compare headline record counts. Coverage of the companies you care about, the accuracy of each field, and the proof behind each record decide whether the data is usable. When one platform combines firmographics, ownership, financials, risk, trade data, patents and growth signals, you can run this evaluation once instead of stitching several vendors together. That is the approach behind Milda, which covers 420 million companies and 50 million verified contacts.
TL;DR
- Build the scorecard before you talk to vendors. Weight coverage and accuracy highest, then compliance, integration, freshness and price.
- Test on a sample of your own companies, including the awkward ones: private firms, subsidiaries and companies outside the US.
- Ask where each field comes from. A record with a traceable source is worth more than a record with a confidence score you cannot inspect.
- Check data protection for contact data before you sign, not after.
- Fewer vendors usually means less reconciliation work. Prefer one integrated source over several partial ones.
Table of Contents
- What a data vendor evaluation is for
- Step 1: Write down the decision the data supports
- Step 2: Build a weighted scorecard
- Step 3: Test coverage on your own companies
- Step 4: Check accuracy and where the data comes from
- Step 5: Check compliance before you sign
- Step 6: Judge freshness and signals
- Step 7: Test integration and workflow fit
- Step 8: Count how many vendors you really need
- Step 9: Compare price on the right basis
- Putting the scores together
- What Milda offers for this evaluation
- Our view: evaluate the data, not the demo
- FAQ
- Sources
What a data vendor evaluation is for
A company data vendor sells records about businesses: who they are, who owns them, how they are performing, who runs them and how to reach those people. You buy it to make decisions. An investment team sources targets, a business development team finds buyers, an analyst maps a market, an innovation team scouts partners, and a government team measures how competitive an industry is.
That is why a generic supplier scorecard, built around on-time delivery and invoice accuracy, does not fit. Those scorecards rate how well a vendor performs against a contract. Data is different: you have to judge the product itself, the records, before you can judge the service around it. The evaluation below is built for that.
The cost of getting it wrong is mostly hidden. A missing company never shows up as an error. It simply is not in your shortlist, your market map or your risk review. A wrong owner or a stale contact does show up, but only after someone has spent time on it. The evaluation exists to find both problems before the contract does.
Step 1: Write down the decision the data supports
Start with the job, not the vendor. Write one sentence for each use case, for example “find acquisition targets in industrial software with 50 to 500 employees in three countries” or “verify the ownership chain of a new supplier before onboarding”. The sentence tells you what to test.
Then list the fields that decision depends on. A short list works better than a long one:
- Sourcing and market mapping: industry classification, size, location, growth signals, competitors, technology and patent portfolios.
- Deal and investor work: ownership, shareholders, financials, event-driven signals such as new funding or expansion.
- Sales and business development: decision-maker names, roles and verified contact details, plus firmographics for segmentation.
- Due diligence and supplier checks: ownership structure, financial health, risk assessments, trade data. Our guide to supplier due diligence for US teams covers this use case in depth.
If a field is not on your list, do not score it. Vendors will show you impressive extras, and extras are how a shortlist gets decided on the wrong criteria.
Step 2: Build a weighted scorecard
Pick six to eight criteria and fix the weights before the first demo. Fixing them first keeps a polished presentation from rewriting your priorities. A workable starting point looks like this:
| Criterion | Suggested weight | What you are really asking |
|---|---|---|
| Coverage of your target companies | 25% | How many of my sample companies are present, and with which fields filled? |
| Accuracy and source traceability | 25% | Is each field correct, and can I see where it came from? |
| Compliance and data protection | 15% | Can I lawfully use the contact data in my markets? |
| Freshness and signals | 10% | How quickly do changes and events appear? |
| Integration and access | 10% | Web app, browser extension, API: can it fit how my team works? |
| Breadth in one place | 10% | How many separate tools does this replace? |
| Price and scaling | 5% | What does it cost as my team and usage grow? |
These weights are a starting point, not a standard. A government team mapping national competitiveness may put coverage and breadth at 60% combined and contact data at zero. A sales team will do the opposite. What matters is that the weights are written down and agreed by the people who will use the data.
Score each criterion from 1 to 5 using evidence from your own test, not from the vendor’s claims. A score with no evidence behind it is an opinion.
Step 3: Test coverage on your own companies
Coverage is where most evaluations go wrong. Headline numbers, such as hundreds of millions of companies, sound decisive, but they measure the whole database, not your slice of it. Many records in any large database are thin: a name and an address and little else. What you need to know is how many of your companies are present and how complete each record is.
Build a test set of 100 to 200 companies, drawn from your real work:
- 30% companies you already know well, so you can check accuracy.
- 30% small and mid-sized private companies, where data is hardest to find.
- 20% companies outside the US, if you operate across borders. Our guide to cross-border company data explains why registries, languages and filing rules make this the weakest spot for many vendors.
- 20% recently founded or recently changed companies, such as ones that were acquired, renamed or restructured.
Run the same list through every vendor. For each company, record whether it was found, which fields were filled, and whether the corporate family (parent and subsidiaries) was resolved correctly. Then calculate two numbers per vendor: the match rate and the average share of important fields filled.
Then run the inverse test. Describe a market in plain terms, for example “logistics software companies in Germany and Poland”, and compare how many relevant companies each vendor surfaces. This tests discovery, not lookup, and it is the test that matters for sourcing and market mapping. Milda’s positioning is deeper coverage from proprietary and multi-source data, surfacing up to three times more companies than other sources; this discovery test is exactly where you would see a difference like that in your own market. Details on what the platform covers are on our data coverage page.

Step 4: Check accuracy and where the data comes from
Accuracy is not a percentage a vendor can hand you. It is the result of how the data was collected, combined and checked. Ask three questions about each field that matters to you.
Where did it come from? Official registries, filings, company websites, trade records and patent offices are different kinds of source, with different reliability. A vendor that can show the source of a record lets you settle a dispute in a minute. A vendor that can only show a score leaves you to trust it. This is why we built access to primary source data into the product, including through a browser extension; see how direct access to primary source data works.
How were conflicting sources reconciled? A company has one legal name in a registry, another on its website and a third in a customer database. Multi-source data is only better than single-source data if the matching is sound. The principles are in our practical guide to multi-source data integration.
Does it hold up on a spot check? Take 20 of the companies you know best and verify every important field by hand: owner, headcount range, revenue band, key executives. Count the errors. Twenty companies is small, but one vendor with three errors and another with eleven tells you what you need to know.
For contact data, accuracy means the person still holds the role and the details work. Check a sample of 25 contacts per vendor against current professional profiles, and check whether email addresses are verified rather than guessed from a pattern. The same section of our guide on sales intelligence data covers the types and sources behind contact records.
Step 5: Check compliance before you sign
Contact data is personal data. If you will hold or use details of people in the EU or UK, the GDPR applies to the way you use them, whatever your own location. One concrete example: when personal data was not collected directly from the person, Article 14 of the GDPR requires the controller to give that person specified information, such as who the controller is and how the data is used.
So your evaluation should ask, in writing:
- Which regulations does the vendor’s collection and processing follow, and for which regions?
- How are deletion and opt-out requests handled?
- What does the contract say about your obligations versus the vendor’s?
- Where is data stored and processed?
Bring your legal or privacy team in at this point, since they will sign off on the contract and own the answers. Milda positions its platform as GDPR-compliant and source-verified, so ask for the documentation in the same request you send to every other vendor, and score all vendors on the same sheet.
Step 6: Judge freshness and signals
Company data decays. People change jobs, companies are acquired, ownership shifts and financials are filed late. Two things matter: how often records are refreshed, and whether the vendor surfaces change as an event you can act on.
Event-driven growth signals are the practical test. If you care about expansion, check how quickly a new facility, funding round, product launch or leadership change appears in the record for a company you know it happened to. Pick three recent events in your own market and see which vendors had them, and when.

It also helps to ask how signals are built. A signal derived from a primary source (a filing, a patent publication, a trade record) is a stronger basis for outreach or investment screening than one inferred from a news snippet.
Step 7: Test integration and workflow fit
A dataset you cannot reach in your normal workflow does not get used. Ask how each vendor delivers data:
- Web app: for analysts, strategy teams and deal teams doing interactive research.
- Browser extension: for people who research while browsing company websites and profiles and want data in context.
- API: for teams who need enrichment inside a CRM, a data warehouse or an internal tool.
Milda delivers its company intelligence through all three: search, browser extension and API. In the evaluation, ask a real user on your team to complete a real task end to end, for example building a list of 50 target companies with ownership and contacts and exporting it. Time it. A tool that takes forty minutes to produce what another produces in ten will lose to the second in practice, however rich the data.
For API access, have an engineer check documentation quality, rate limits, and how matching handles ambiguous company names. Our Data Platform page describes how the data is delivered.
Step 8: Count how many vendors you really need
Most teams end up with several vendors because each is good at one thing: one for contacts, one for financials, one for ownership, one for patents. Each extra vendor adds a contract, a security review, a different company identifier and a reconciliation job. That hidden work is rarely in the scorecard, so add it: for each vendor, estimate the hours per month the team spends moving or matching data between systems.
That is the logic behind scoring “breadth in one place” as its own criterion. A single platform that covers firmographics, ownership and shareholders, financials, risk, trade data, patents, publications, growth signals and competitor mapping removes whole categories of matching work. For a fuller view of how this plays out in practice, a resource on AI-assisted link prospecting workflows shows the same principle in another field: the fewer disconnected inputs a workflow depends on, the less of it breaks.
Step 9: Compare price on the right basis
Price comparisons fail when the units differ. One vendor charges per seat, another per credit, a third per API call, and a fourth by record exported. Convert each to the cost of your actual workload: the number of users, the number of companies researched or enriched per month, and the number of contacts you need. Then compare the total for twelve months, including onboarding, integrations and any minimum term.
Milda’s plans adapt to team size and data usage level, and the offers are Browser Analyst, Strategy Terminal and Data Solutions. No prices are published, so the pricing page is where you start a conversation for a quote that matches your workload.
Putting the scores together
After testing, multiply each 1 to 5 score by its weight and add up. Then do three sanity checks before you decide:
- Does the ranking match the team’s instinct? If not, find out why. Either the weights were wrong or the instinct is about something you did not measure.
- Is the top vendor weak on any must-have? A vendor can win on total score and still fail on compliance or a critical field. Treat must-haves as pass or fail, not as points.
- Would the users choose it? Let the people who will use the data run the final session. Adoption decides whether any of this was worth it.
Keep the test set and the scores. In twelve months, rerun a smaller version at renewal. It takes a day, and it shows whether the data held up.
What Milda offers for this evaluation
If your shortlist is about company data for sourcing, market mapping, due diligence or business development, Milda is built for that evaluation. The platform covers 420 million companies and 50 million verified contacts, with firmographics, ownership, shareholders, financials, risk assessments, trade data, patents, publications, growth signals and competitor mapping in one place, available through search, a browser extension and an API.
Bring your test set and your scorecard. Review the product lines on our pricing page to see which of Browser Analyst, Strategy Terminal and Data Solutions fits your team, or look at how we support market research and sales and GTM research.

Our view: evaluate the data, not the demo
Most vendor evaluations we see are decided too early, on the strength of a database size or a polished interface. Neither tells you whether your companies are in there.
We built Milda around a different idea. Coverage is only useful if it is deep, with ownership, financials, risk, patents and signals attached to the same company, and if you can trace each fact to its source. That is why we combine proprietary and multi-source data, and why we claim up to three times more companies surfaced than other sources.
Our advice to any team choosing a provider is simple: bring your own list, weight your criteria before you see a demo, and judge by what the data does in your workflow. We are comfortable being evaluated that way, because it is the way the platform is designed to be used.
Vilius Stanciauskas, Founder and CEO
FAQ
What are the most important criteria in a data vendor evaluation?
Coverage of your own target companies and the accuracy of each field matter most, so give them the largest weights. Next come compliance, freshness, integration and price. Treat legal requirements as pass or fail rather than as points.
How many vendors should I compare?
Three to five is enough for a serious test. Fewer leaves you without a benchmark, and more makes it hard to run the same sample through every vendor. Run the same test set and the same scorecard on all of them.
How big should my test sample be?
Use 100 to 200 companies for coverage and match rate, and spot-check 20 companies and 25 contacts by hand for accuracy. Include private, small, non-US and recently changed companies, because those are where vendors differ most.
How do I compare vendors that price differently?
Convert every quote to the cost of your real workload over twelve months: users, companies researched or enriched per month, and contacts needed. Include onboarding, integration work and any minimum term so that seat, credit and API pricing line up.
Who should be involved in the evaluation?
Include the people who will use the data daily, an engineer if you need API access, and your legal or privacy team for contact data. The users’ verdict at the end often settles close calls.