Vendor Risk Data Quality: Five Lessons for Defensible Records

A vendor record looks backwards. It holds what the vendor said, or what a scanner saw, on the day someone last looked. Five lessons on dating, sourcing, scoping and testing vendor risk data so it holds up under DORA and NIS2 scrutiny today.

Vendor Risk Data Quality

A vendor record looks backwards. It holds what the vendor said, or what a scanner saw, on the day someone last looked.

When your third-party risk management programme treats the record as a description of the present, it is overstating what it can know.

The vendor most likely to expose that gap was low risk when it was onboarded. A supplier that began by hosting a marketing site and now sits inside the payment flow still carries the questionnaire from the year it joined, with a rating that was accurate at the time. Most teams ask whether a vendor has been assessed. The more useful question is how old the answer is, where it came from and what question it was answering.

EU cyber and operational resilience regulation is moving towards the same question. DORA Article 28(3) requires financial entities to maintain and update a register of information for contractual arrangements on ICT services, and EBA Q&A 2025_7388 confirms that the obligation is not limited to critical or important ICT services. The ESAs describe the register as a tool for monitoring ICT third-party risk, and a live Register of Information can meet that standard where a register filed once cannot. In the ESAs’ voluntary, best-effort 2024 dry run, 6.5% of the 947 registers analysed passed all 116 data quality checks, and missing mandatory information accounted for 86% of the errors.

For in-scope essential and important entities, NIS2 requires Member States to ensure that management bodies approve the cybersecurity risk-management measures, oversee their implementation and can be held liable for infringements (Article 20(1)). Those measures must include supply chain security, covering security-related aspects of relationships with direct suppliers and service providers (Article 21(2)(d)), and entities must take into account the vulnerabilities specific to each direct supplier and the overall quality of its cybersecurity practices (Article 21(3)). Taken together, these provisions put the same question in front of the management body, supplier by supplier.

Five habits separate a record that is documented from one that can be defended.

At a glance:
  1. Date every answer, because vendors keep changing after they respond
  2. Record where each answer came from and weigh it accordingly
  3. Check the answer against today’s question, not the one you asked at onboarding
  4. Test the data against a conversation before you rely on it
  5. Write down what you do not know, and treat the list as part of the record

1. Date every answer, because vendors keep changing after they respond

Leave the date off an answer and it reads as current. Verizon’s 2026 Data Breach Investigations Report covers incidents from 1 November 2024 to 31 October 2025 and states that breaches involving a third party reached 48% of the total, up 60% from the prior year’s dataset. Drawing on exposure data collected from inside third-party cloud environments, it also found that only 23% of third-party organisations fully remediated missing or improperly secured multi-factor authentication on cloud accounts, and that weak passwords and permission misconfigurations took almost eight months for half of findings to be resolved.

Vendors change at very different speeds. A control confirmed in March may have been fixed, broken or abandoned by September. Without a date, the record cannot say which. Skip a periodic review for several years and the record still looks current while the evidence behind it ages.

In an understaffed team the review calendar decides how old the oldest answer becomes, and Mitratech’s 2025 TPRM Study found nearly 70% of teams in that position.

2. Record where each answer came from and weigh it accordingly

A vendor describing its own controls, a third party measuring its exposure from outside and your own team observing how the vendor behaves under a real request produce three different kinds of evidence. Store them in one field and every answer inherits the weight of the most flattering one. Self-attestation captures what the vendor intends when it signs, and outside-in data, the kind gathered through third-party risk intelligence, captures what the internet can see when the scan runs.

Only your own observation, such as how quickly an incident notice arrived and how complete it was, tests the relationship you actually have. Each answers a different question, and the record should say which one it is answering.

3. Check the answer against today’s question, not the one you asked at onboarding

Scope follows risk tier and use, and both change. If a vendor has begun using AI in the service it delivers to you, a questionnaire written before that point contains no questions about it. If a law has come into force since the last assessment, the answer predates the obligation you now rely on it to evidence.

DORA applies the register obligation across ICT contractual arrangements and leaves the depth of wider third-party risk management to proportionality. Tier, which reflects criticality, service type and dependency, therefore sets how much evidence an answer needs. A vendor whose use has changed keeps its onboarding tier until somebody moves it, so the tier needs a review date of its own.

Reginald Youngblood, recently retired Associate Vice President of Corporate Compliance at AT&T, described in our recent webinar how he applies the same logic to fourth parties: he asks which ones “would put you on the front page the fastest” and starts the review there. NIS2 Article 21(2)(d) names direct suppliers and service providers, so extending the review to fourth parties is a judgement the programme makes for itself.

4. Test the data against a conversation before you rely on it

Continuous third-party risk monitoring narrows the gap between what a vendor does and what you learn, and for higher-risk vendors it belongs in the programme. When a feed signals a change, you still do not know what it means for the service you receive, and the person best placed to say is the relationship manager who talks to the vendor every week.

Youngblood shares the limitation: “The data could say one thing, but reality could be something different.” A short conversation, held when a signal arrives or a renewal approaches, tests the record against the relationship. A feed cannot do that, because it has no view of what was agreed at the last service review. His other rule belongs in the record itself: “Document that you did reach out to them.” When a vendor stays silent, the logged attempt shows you did your part. Then, let your upper management team know the vendor has been non-responsive. It could cause a change in vendors.

5. Write down what you do not know, and treat the list as part of the record

The regulations discussed here do not prescribe a formal completeness statement for each vendor record, so no rule makes you write one, and the other four habits hold up better when you do. Each critical vendor record should say which questions were not asked, which answers are out of date and when the gaps will be closed.

The ESAs’ dry run suggests where the gaps cluster: the template for the assessment of the ICT services, including provider substitutability, accounted for 20% of all missing values. An organisation that knows its record is thin on substitutability and has a dated plan to fix it stands in a stronger position than one that filed a complete-looking register nobody has tested. Boards expect the same candour. A board can act on a gap that is stated and dated, and that detail gives the evidence behind DORA and NIS2 vendor controls somewhere to start.

Open the record of your highest-risk vendor

Find the date of its most recent answer, who gave it and which question it was answering. If any of the three takes longer than a minute to find, your record is still describing the past. Reginald Youngblood and I take these questions further in the on-demand webinar, Assessed and Filed: Why Vendor Risk Data Isn’t the Same as Vendor Risk Visibility.

Preguntas frecuentes

Why do vendor risk assessments go stale?

An assessment records what a vendor said or what a scan showed on one day, and the vendor keeps changing after that. The review calendar then decides how old the oldest answer becomes. Date each answer and you can see which records no longer describe the vendor.

How often should a vendor risk assessment be refreshed?

No single interval applies to every vendor. The depth of third-party risk management is proportionate, so refresh timing should follow the vendor’s risk tier, the service provided, and any change in use, scope or dependency. Under DORA, the register of information must be maintained and updated for contractual arrangements on ICT services.

What should a vendor risk record contain besides questionnaire answers?

Each answer should carry its date, its source (the vendor’s own statement, outside-in data or your own observation) and the question it was answering. The record should also state which questions were not asked and when the gaps will be closed.

Does DORA require the register of information to be kept up to date?

Yes. DORA Article 28(3) requires financial entities to maintain and update the register of information for contractual arrangements on ICT services, and EBA Q&A 2025_7388 confirms that the obligation is not limited to critical or important ICT services. In the 2024 dry run, 6.5% of the 947 registers analysed passed all 116 data quality checks, in an exercise that was voluntary and best effort.

How does vendor risk data differ from vendor risk visibility?

Vendor risk data fills the record. Visibility lets you say how old each answer is, where it came from and whether it still answers today’s question about that vendor.

Is your vendor data current?

Ver el seminario web