MedTech case study: qualified radiologists in 6 weeks

In six weeks, a French MedTech company specializing in medical imaging went from a mass of raw contacts to a database of qualified radiologists, each one carrying a buying signal.

The result did not come from a purchased list, but from cross-referencing more than ten public healthcare sources (RPPS, ARS, BOAMP, HAS, PubMed), qualified by AI agents specialized in the medical field. Every radiologist on the final list carried a verified buying signal, not just a name and an email address.

This case study covers the starting problem, the qualification method, the CRM cleanup that followed, the results in numbers, and what carries over to another MedTech company.

Why did 15 reps across 8 countries have no usable prospect database?

Fifteen sales reps spread across eight European countries each worked from their own Excel lists, built up over the years, with no enrichment and no buying signals. Before every outreach attempt, a rep spent 30 to 45 minutes manually researching a radiologist’s context, usually for a cold contact. A physician with no identified imaging need had no commercial value, but nothing made it possible to tell the one who was buying from the one who would not budge.

The problem wasn’t a lack of names; it was a lack of context. A list of radiologists with no view of their imaging setup, no equipment history, and no renewal signal is a cold list disguised as a database.

Yet the pool of radiologists in a given country is finite and identifiable: France has about 8,900 (DREES, 2020). An exhaustive database per country is therefore within reach, as long as you know how to read the right sources instead of buying back a list everyone already has.

How do you qualify thousands of radiologists on a buying signal?

Qualification cross-references more than ten public sources for each physician: the RPPS (France’s national registry of healthcare professionals), LinkedIn, Hospimedia, PubMed, the ARS (regional health agencies), the BOAMP (France’s official public procurement bulletin), the FINESS code (France’s national registry of healthcare facilities), the HAS (France’s National Authority for Health), the BODACC (France’s official bulletin of company announcements), and the trade press. AI agents specialized in the medical field detect real buying signals there (an MRI nearing the end of its term, an open tender, a job change, a recent publication), verify contact details, and reconstruct the history. Only a minority of the scraped contacts come out qualified. This is the work of Radar, the AgifyMD module that detects buying signals.

Every record delivered includes a profile summary (specialty, institution, seniority) and a research section (recent publications, areas of expertise). The FINESS code serves as the pivot key: it links a physician to their facility and to every registry that describes it, from the PMSI (France’s hospital activity database) to the BOAMP. That is what turns an isolated name into a lead with real context.

This event-based qualification logic is the same as prospecting with buying signals in the hospital market: you keep a radiologist because a public fact identifies their facility as a potential buyer, not because they appear on a list. The reps could then work precise micro-segments: lists of five to ten physicians, 3T MRI as the criterion, a targeted campaign instead of a mass send. When this qualification feeds a direct outreach channel, it produces engagement rates well above the market, as shown by this LinkedIn outreach campaign to physicians.

If you want to know which buying signals actually exist for your product catalog, that is exactly what we scope in 30 minutes during a prospecting scoping call.

How do you clean up a CRM without redoing it by hand?

A B2B contact database decays by about 22.5% a year (HubSpot, citing MarketingSherpa): outdated addresses, duplicates, empty fields. A second AI agent took over the CRM records. It cross-referenced the RPPS, Hospimedia, and facility data to reconstruct each center’s imaging equipment, merged duplicates, and updated outdated records.

For each center, the agent reconstructed the available MRI scanners and their field strength (1.5T, 3T), the installed brands (GE, Siemens, Philips), the incumbent competitors, and renewal signals. It structured fields reps could use directly: specialty, seniority, region, decision-maker status, involvement in research, preferred language, with a confidence score for each record.

The challenge is not a one-off cleanup; it is keeping it clean over time. A CRM cleaned once becomes dirty again in less than two years, mechanically. An agent that continuously re-enriches records from public sources keeps the completion rate up instead of letting it slide.

What measurable results did the qualified database deliver?

The engagement delivered a database of qualified prospects in six weeks, followed by an expanded database segmented by region. Prep time per meeting dropped from 30 to 45 minutes to just a few minutes. A third agent, dedicated to inbound leads, now qualifies new prospects as they come in.

The 15 reps no longer spend their time on research, but on prospects who all carry a verified buying signal. The ratio flipped: less manual qualification, more selling.

“We can target exactly the people we want to reach: lists of five to ten physicians, 3T MRI as the criterion, a small, well-targeted campaign with a brochure and a message.” a rep on the team, feedback session (translated from French)

What can another MedTech company reuse?

Three mechanisms carry over. First, multi-source medical cross-referencing: more than ten public registries cross-referenced beat any purchased list, which is already out of date the day it is created. Second, the CRM agent that maintains quality over time, rather than a one-off cleanup that quickly decays again. Third, the buying-signal principle: qualify a physician on a verified event rather than on their title alone. This logic has a name, reverse prospecting, and it applies to any MedTech company with a long sales cycle and a high deal value.

The buying signal can take different forms depending on the product: a tender, an authorization renewal, the appointment of a decision-maker, or what the HAS quality overview reveals about a facility. For sales that go through public procurement, detecting a BOAMP tender before the specifications are written is the direct extension of this logic. In the private clinic market, the same method produced a map of 4,342 HAS-listed facilities in nine days. At the European level, this structured-database logic made it possible to build more than 10,000 medical leads across 18 countries in 8 days, segmented into four distinct groups by buying role.

One honest limitation: the investigation cost per physician rises with the precision of the qualification. The return shows up across the database and the campaign, not lead by lead. That is why the initial scoping phase, which defines the relevant signals, weighs so heavily in the final result.

Frequently asked questions

What share of a scraped database comes out qualified?

In this case, only a minority of the scraped contacts came out qualified. The rate depends on the specialty, the geographic scope, and how strict the disqualification criteria are. The metric to track is not the raw volume scraped, but the share that carries a verified buying signal and usable contact details.

How long does it take to qualify a database of physicians?

Six weeks for a database of qualified radiologists across eight countries, in this case. The timeline depends mainly on the initial scoping phase, which defines the relevant signals, more than on the processing itself. Once the sources and criteria are set, enrichment runs continuously and the database keeps growing.

Is using the RPPS and the BOAMP to target physicians GDPR-compliant?

Yes. The RPPS, the BOAMP, the HAS, and the ARS are public sources. Using them for B2B prospecting falls under legitimate interest (GDPR, Article 6(1)(f)), provided you respect the right to object of the people concerned. The data can legally be used to qualify a physician.

Can an AI agent keep a CRM clean over time?

Yes, as long as it runs continuously and not just once. A database decays by about 22.5% a year (HubSpot). A one-off cleanup is therefore lost again in less than two years. The agent described here continuously re-enriches records from public sources, which keeps the completion rate up instead of letting it slide.

Conclusion

Qualifying thousands of radiologists in six weeks does not come down to scraping volume, but to signal quality. Three takeaways: cross-referencing more than ten public healthcare sources beats a purchased list that is out of date the moment you buy it; a CRM agent running continuously maintains a database that time degrades by 22.5% a year; and qualifying on a verified buying signal, not a job title, turns a raw database into prospects you can contact at the right moment.

The open question is which signals matter for your catalog. Let’s look at your case in 30 minutes: together, we will identify the public sources relevant to your products and how to turn them into a qualified pipeline.

---

Agify deploys autonomous AI outbound agents within B2B sales teams, with a specialization in the French MedTech market and its public healthcare data sources.