Real World Data Vendors: Types, Costs and How to Choose One

What real world data vendors sell, in one paragraph

Real world data vendors license patient-level health data collected outside clinical trials: insurance claims, electronic health records (EHR), lab results, oncology and genomic records, and patient-reported data. Pharma teams use them for launch planning, patient finding, market sizing, health economics and regulatory evidence. The right vendor depends on the question you need to answer, not on the size of the database, so the first decision is the data type, then the vendor, then the contract.

That order matters because the market has no single winner. CB Insights interviewed pharma buyers in 2023 and heard the same complaint repeatedly: there is no one-stop shop, pricing has no standard model, and the data rarely joins cleanly to what the company already holds. IntuitionLabs, in its RWD platform comparison updated on 13 November 2026, recommends that larger companies run two to four platforms in parallel for exactly that reason. A smaller brand team cannot afford that, so it has to choose well the first time.

This guide covers the main types of RWD vendor, what each is good and bad at, how deals are priced, the contract terms that decide whether the data is usable, and a step-by-step way to choose. It is written for brand, commercial operations, market access and HEOR leads at mid-sized pharma and biotech companies who are buying RWD for the first time or replacing a licence that is not delivering.

The RWD market in four numbers

The main types of real world data vendors

Most RWD vendors fall into five groups. The patient counts each one advertises are not comparable with each other, because a claims vendor counts anyone with a claim and an EHR vendor counts anyone with a record in a participating health system. Compare them on the data type you need, not on the headline number.

Two points trip up first-time buyers. First, open and closed claims answer different questions, which we explain in our Komodo Health vs IQVIA comparison. Second, the big syndicated vendors sell both prescription audits and patient-level data, and the two are licensed separately. Our guides to Symphony Health vs IQVIA and Veeva Compass vs IQVIA break those products down.

Which type of RWD vendor fits your question?

Commercial and launch

If the question is how many patients are diagnosed, who treats them, what they are on today and how fast they switch, claims data answers it at national scale. Most brand and commercial operations teams need claims first and clinical data later, if at all.

HEOR and regulatory

Outcomes research and regulatory submissions need clinical variables such as lab values, disease severity and reasons for stopping, which claims do not carry. FDA's July 2024 guidance asks sponsors to show that study variables are captured and validated in the source, so data provenance matters as much as size.

Oncology

Line of therapy, biomarker status and progression are rarely coded in claims. Oncology-specific vendors abstract them from charts and pathology reports, which is why buyers in the CB Insights interviews named Flatiron, Tempus and Syapse when asked about cancer data.

Trial feasibility

To test whether enough eligible patients exist at real sites, you need counts against inclusion and exclusion criteria, by site. Federated EHR networks were built for this and often price it separately from full data access.

How much does real world data cost?

Real world data cost is rarely published. IntuitionLabs notes that none of the major vendors publishes a comparable public rate card, and every deal is quoted against a scope. The only broad benchmark we have read is the CB Insights buyer research from July 2023, which put annual spend between $75,000 and $5 million, with buyers saying the vendors "don't really have a standard [pricing] model" and that pricing is negotiated case by case.

What moves the price is predictable even when the number is not:

• Scope of patients. One indication costs less than a whole therapy area or the full database.
• Lookback and refresh. More years of history and more frequent updates cost more. A monthly refresh is a different product from a one-off extract.
• Delivery model. Data delivered to your own cloud warehouse, access inside the vendor's analytics platform, or only aggregated outputs. Raw delivery to you is usually the most expensive and the most useful.
• Rights of use. Internal analytics only, publication, regulatory submission, or sharing with agencies and consultants. Each extra right is priced.
• Linkage. Joining the vendor's data to your hub, specialty pharmacy or CRM data through tokens is usually a separate fee and sometimes a separate vendor.
• Term and exclusivity. Multi-year deals buy a lower annual price; exclusivity on a dataset, where offered, costs a lot more.

The ISPOR Value and Outcomes Spotlight article on sourcing RWD (Foster and Karandikar, 2023) is blunt about the trade-off: vendors move fast and pool many sources, but "the costs tend to be very high" and the buyer has very little control over what is collected or how it is pooled. Budget for the people and systems needed to use the data, not only the licence. A dataset nobody can query is the most expensive line in the budget.

Worked example: reading a feasibility count before you sign

Every vendor will run a feasibility count before a deal. Here is how to read one, using illustrative numbers rather than any vendor's real figures.

Say you are launching a therapy for a rare autoimmune condition and a claims vendor returns 40,000 patients with at least one diagnosis code in the last three years. That is the headline. Now apply the filters your brand team actually needs:

1. Two or more diagnosis claims, at least 30 days apart, to remove rule-out codes. Suppose that leaves 22,000.
2. Continuous enrolment visible for 12 months, so you can see their full treatment history. In open claims this filter is softer because enrolment is not observed; ask how the vendor handles it. Suppose 15,000 remain.
3. On a relevant prior therapy in the last 12 months, the patients your label targets. Suppose 6,000.
4. Treated by a prescriber you can identify with a valid NPI. Suppose 5,400.

The useful number for targeting is 5,400, not 40,000. Ask every vendor on your shortlist to run the same written definition, with the same codes and windows, and compare the counts at each step. If one vendor loses far more patients at the enrolment step than another, you have learned something about its payer coverage that no sales deck will tell you. Then ask how many of the 5,400 can be linked by token to your own hub or specialty pharmacy data, because that is the number that determines whether you can measure your own patients' journeys.

How to choose a real world data vendor, step by step

Contract terms to negotiate before you sign

• Refresh cadence and lag

  Write the delivery schedule and the maximum lag into the contract, with what happens if a delivery is late.

  • Monthly, quarterly or one-off
  • Lag from service date to delivery
  • Notice of restatements

• Rights of use

  Confirm in writing who can use the data and for what: internal analytics, publications, regulatory submissions and agency or consultant access.

  • Named third parties allowed
  • Publication and submission rights
  • Derived data ownership

• Delivery and format

  Agree where the data lands and in what model, so your team or partner can load it without rework each month.

  • Delivery to your cloud warehouse
  • Data dictionary and change log
  • Common data model such as OMOP, if offered

• Linkage and tokens

  If you will join the data to your own patient services data, agree the tokenisation vendor, who pays for it and the expert determination needed under HIPAA.

  • Token provider named
  • Re-identification risk review
  • Match-rate reporting

• Coverage changes

  Sources leave networks. Ask to be told when a payer, pharmacy or health system drops out and how it affects trend lines.

  • Notice of source loss
  • Coverage report each refresh
  • Price review if coverage falls

• Exit and retention

  Know what you keep when the licence ends: the raw data, derived tables, or only published outputs.

  • Destruction or retention terms
  • Derived analytics you keep
  • Renewal price caps

Getting RWD to work with the data you already own

Most of the value in real world data comes from joining it to your own sources: CRM call activity, hub and specialty pharmacy status feeds, copay and free goods data. Done well, you can see the full journey of your own patients against the wider market. Done badly, you hold a second, disconnected version of the truth that nobody reconciles.

The join happens through privacy-preserving tokens. Each data source runs the same tokenisation software over patient identifiers before the data leaves it, so records for the same person can be linked without names or dates of birth being shared. HIPAA allows this route to de-identified data through an expert determination, under 45 CFR 164.514(b), that the linked dataset carries a very small risk of re-identification. Our guide to patient tokenization in specialty pharma explains how tokens work and where match rates fall down.

After linkage, the data needs a home. We recommend landing RWD in the same warehouse as your other commercial data, with prescriber and account identifiers mastered so that a physician in the claims feed is the same record as in CRM. That is the design we describe in building a pharma commercial data warehouse, and the platform question is covered in Snowflake vs Databricks for pharma. If your RWD arrives through a specialty pharmacy aggregator, see our specialty pharmacy data aggregator guide for how those feeds differ. This linking and reporting layer is the core of our pharma commercial analytics consulting: we work with the licences you already hold rather than selling data.

Common mistakes, and when not to buy RWD yet

Questions about real world data vendors