Why audits expose a weak vendor risk platform
Most banks choose third-party risk management software on the strength of the workflow demo: questionnaires go out, risk scores roll up, dashboards turn green. An audit or supervisory review checks something else. Can you produce, on request, a complete and internally consistent inventory of who supports each critical function? Can you show how and why you tiered those providers, what you've done to monitor them since onboarding, and how you'd exit if one failed?
Supervisors in many jurisdictions ask versions of those same questions. In the EU it's DORA and the EBA outsourcing guidelines. Singapore has the MAS outsourcing requirements, India has the RBI's IT outsourcing directions, the UK has PRA SS2/21, and US banks work under the 2023 interagency guidance on third-party relationships. The wording differs; the evidence expected overlaps heavily. The eight mistakes below are the ones that most often turn into audit findings, and each comes with a fix you can test during selection or apply to the tool you already run. This is an operating guide, not legal advice. Your compliance and legal advisers should confirm which obligations apply to each entity.
Looking to streamline vendor RFP selection?
Access pre-negotiated tier discounts through our independent institutional network.
Treating the DORA register of information as a report you'll export later
The DORA register of information isn't a narrative document. The implementing technical standards set fixed templates that link contractual arrangements, the ICT services delivered under them, the functions those services support, and the providers involved, including their LEIs and their place in the subcontracting chain. National authorities collect the register in a structured format and pass it to the European Supervisory Authorities. The ESAs' 2024 dry run reportedly exposed data-quality problems such as missing identifiers and broken links between templates, which is exactly what loose vendor profiles produce.
Trouble starts when the platform holds vendor records as free-form profiles and the register gets assembled in spreadsheets each reporting cycle. Every manual join is a chance to introduce an inconsistency that a validation rule will catch.
What to do instead: in the demo, ask the vendor to generate the register live from its own data model, not from a prepared file. Run it through the current validation rules. Then ask who pays for and delivers taxonomy updates when the ESAs revise the templates. If the answer is "professional services," price that in. The same test applies to outsourcing register software for non-EU entities: the register should be a view of the data, not a separate project.
Inventorying contracts instead of the functions they support
Every regime here asks the same core question: which arrangements support functions whose disruption would hurt the firm, its customers or the market? DORA and the EBA guidelines say "critical or important functions." The PRA and RBI use "material outsourcing." MAS applies its own materiality test, and the US guidance speaks of "critical activities." A contract list can't answer any of them.
One cloud master agreement can support payments, core banking and fraud analytics. One business function can depend on six contracts. If your platform stores contracts but not the many-to-many links between contracts, services and functions, you can't show an examiner your concentration exposure or what happens when a single provider fails.
What to do instead: make the function taxonomy the anchor of the data model. Before go-live, get the business-continuity and operational-resilience teams to agree on one list of important business services or critical functions, and require every arrangement to map to it. During evaluation, ask vendors to show one provider's footprint across all functions on a single screen.
Tiering vendors by spend or a single questionnaire score
Spend is easy to pull from procurement, which is why it ends up driving tiering. It's also the wrong measure. Supervisors expect criticality to reflect how much disruption a failure would cause, how sensitive the data is, how hard the provider would be to replace, and whether the service touches a regulated activity.
Hypothetical example: say a regional bank pays USD 30,000 a year for a sanctions-screening data feed and USD 2 million for facilities management. Tiering by spend puts facilities at the top and the screening feed near the bottom. But if the feed fails, the bank may be unable to onboard customers or release payments by the next morning. If the facilities provider fails, it's an inconvenience.
What to do instead: build inherent-risk tiering from a short set of weighted criteria: function criticality, data classification, substitutability, regulatory sensitivity and concentration. Keep it separate from the residual score that due diligence produces. The software should record who approved each tier and when, so you can show the rationale at audit, not just the result.
Looking to streamline vendor RFP selection?
Access pre-negotiated tier discounts through our independent institutional network.
Stopping at the contract counterparty and ignoring fourth-party risk
Your core banking vendor probably runs on a hyperscaler, uses an offshore support center and embeds a third-party database. DORA expects firms to capture subcontractors that effectively underpin ICT services supporting critical or important functions. PRA SS2/21 and the RBI directions both expect firms to understand material sub-outsourcing. The US guidance expects banks to consider a third party's own reliance on subcontractors. This is where groups often find, too late, that five "diversified" providers all depend on the same cloud region.
What to do instead: require subcontractor disclosure at onboarding and at each review for critical tiers, and store fourth parties as linked entities with their own identifiers, not as free text. Then ask the vendor to show a concentration view that rolls up shared fourth parties across your portfolio. A platform with only a "subcontractors" comment field fails this test.
Calling an annual questionnaire 'ongoing monitoring'
A questionnaire sent once a year and filed is a point-in-time attestation. Supervisors want evidence that you track performance against SLAs, follow incidents and changes of control, and step up reviews when risk signals move. DORA's ICT risk framework, MAS's technology risk expectations and the US lifecycle model all treat monitoring as continuous and proportionate to tier.
Security ratings and news feeds help, but they don't replace monitoring. An outside-in score says nothing about whether the vendor met its recovery-time objective last quarter.
What to do instead: define monitoring by tier. For example: quarterly SLA and incident review for critical providers, annual reassessment for high-risk ones, and event-triggered review for the rest. Configure the platform to pull in SLA data, incident notifications and external signals, and to open review tasks automatically when thresholds are breached. Auditors will look for the trail from signal to action.
Leaving exit plans as a contract clause nobody has tested
Exit strategy is one of the most consistent supervisory themes. DORA expects documented exit strategies for ICT services supporting critical or important functions. The PRA expects stressed-exit planning for material outsourcing. The EBA, MAS and RBI all expect termination and transition arrangements, and the US guidance treats termination as its own lifecycle stage. Many banks can point to an exit clause. Few can show a plan that names an alternative provider, a transition timeline, data-return formats and an internal owner.
What to do instead: treat the exit plan as a structured record linked to each critical arrangement, with fields for the alternative provider, estimated transition time, data portability and the date it was last reviewed or tested. Run a tabletop exercise at least annually for your top few critical providers and store the output in the platform. If a vendor's tool can only attach a PDF, treat that as a red flag.
Configuring the platform for one regulator when the group answers to several
A group with entities in Dublin, London, Singapore, Mumbai and New York faces overlapping but different definitions, registers and notification rules. Australia's CPS 230, Gulf central bank outsourcing rules and South Africa's Prudential Authority add further variants. Build the tool around one regime and you'll rework it, or run parallel spreadsheets, for every other jurisdiction.
| Regime | Criticality concept | Register / inventory expectation | What auditors often probe |
|---|---|---|---|
| EU DORA and EBA outsourcing guidelines | Critical or important functions | Structured register of information submitted to authorities; outsourcing register for non-ICT | Data quality, subcontracting chain, exit strategies |
| UK PRA SS2/21 | Material outsourcing | Outsourcing register; notification of material arrangements | Stressed exit, concentration, intragroup arrangements |
| MAS (Singapore) | Material outsourcing arrangements | Register of outsourcing arrangements available to MAS | Due diligence depth, technology risk controls |
| RBI IT outsourcing directions (India) | Material outsourcing of IT services | Inventory of outsourced services and providers | Sub-outsourcing, data location, audit rights |
| US interagency guidance | Critical activities | Risk-based inventory, not a prescribed template | Lifecycle evidence, board oversight, subcontractors |
What to do instead: require jurisdiction tagging at the entity and arrangement level, with criticality fields that can hold several regulatory classifications at once. In the demo, ask the vendor to show one arrangement classified differently for two regulators without duplicating the record.
Budgeting for licenses and forgetting the data work
The subscription is usually the smallest number that surprises a bank after signing. The larger costs are data migration, function mapping, integration with procurement and contract systems, and the internal staff time to clean up years of inconsistent vendor records. Widely used platforms include ServiceNow, Archer, MetricStream, OneTrust, ProcessUnity and Aravo. They differ less in core features than in how much configuration they need and how they price modules.
| Cost driver | Typical 2026 range (USD) | Notes |
|---|---|---|
| Annual subscription, mid-size bank | 40,000 to 150,000 | Priced by vendor count, users or modules |
| Annual subscription, large or multinational group | 250,000 to 1,000,000+ | Multi-entity and regulatory register modules add cost |
| Implementation and configuration | 0.5x to 1.5x first-year subscription | Higher where function mapping is done for the first time |
| External risk intelligence feeds | 15,000 to 100,000 per year | Security ratings, financial health, adverse media |
| Regulatory taxonomy updates | Included to 30,000+ per year | Confirm contractually for DORA and similar registers |
These are typical 2026 market ranges. They vary by contract, region and negotiation, and local pricing often differs, so confirm with vendors before you budget.
What to do instead: before you sign, run a four-to-six-week data-readiness sprint on your 50 most critical arrangements. Map each to functions, identifiers, subcontractors and exit plans. If the sprint finds the data you need, choose the platform that generates your strictest regulator's register from it with the least manual work. If it doesn't, fix the data first. No tool will pass an audit built on records you can't defend.