Your company’s payroll runs on a hosted platform. Customer documents sit with a cloud storage provider. A small agency has admin access to your website. A software library you never audited sits inside your product. Each of these is a third party that can hurt you, and several large incidents in recent years began at a supplier, not at the victim. Third-party risk management is the discipline of deciding how much trust each of them deserves. There are two broad styles, and I will argue for one.

Style one: the questionnaire

The familiar approach. You send a spreadsheet of 150 questions: do you encrypt data at rest, do you have an incident response plan, do you run annual penetration tests, do you have ISO 27001. The vendor answers, someone reviews, and the vendor is marked low, medium or high risk. Standard sets exist, such as the SIG questionnaire from Shared Assessments and the CAIQ from the Cloud Security Alliance, which reduce the effort of writing your own.

Strengths are real. It scales across hundreds of suppliers. It forces a baseline conversation. It creates a record, which matters for auditors and for contract clauses. It is also cheap for you.

Weaknesses are equally real. Answers are self-reported and often written by a sales engineer under a deadline, rounded towards “yes”. “Do you encrypt data?” gets a yes whether the encryption covers one database or all of them. Questionnaires measure what a vendor says it has, at a single moment, and age quickly. And 150 questions treat a missing policy and a missing control as equal weight, which hides what matters.

Style two: evidence

Here you ask for proof, and you look at the thing itself. Independent assurance reports: an ISO 27001 certificate together with its scope statement and statement of applicability, or a SOC 2 Type II report from an auditor, which describes controls and tests them over a period. You read the scope, since a certificate covering one office does not cover the product. You read the exceptions the auditor listed. You review the summary of a recent penetration test and whether the findings were fixed. You examine the architecture for how your data is separated from other customers’, who at the vendor can access it, where it is stored, and who the vendor’s own subprocessors are.

You can add outside-in signals: scanning the vendor’s public footprint for expired certificates, exposed services or leaked credentials, or using rating services. These are rough, can be wrong, and do not see inside the company, but they are independent, repeatable, and can be run continuously.

The cost is effort. Nobody has the time to read an audit report for every stationery supplier. That is where the real skill comes in.

My side

Opinion: use the questionnaire as a screening tool, not as the decision. Spend your real effort on evidence, but only for the small number of vendors where a failure would hurt, and judge that by what they touch, not by how big they are.

A workable tiering looks like this, with illustrative boundaries:

TierWhat the vendor touchesDepth of review
CriticalPersonal or financial data at scale, production access, or a service you cannot run withoutAudit reports, architecture review, contract security schedule, annual reassessment, exit plan
ImportantLimited sensitive data, or a significant but replaceable serviceShort questionnaire plus one independent attestation
LowNo data, no network accessBasic checks and standard contract terms

The contract does half the work

Assessment finds out what the vendor does today. The contract makes them keep doing it. For critical suppliers, look for clauses on: breach notification within a defined short window, so you can meet your own deadlines under CERT-In directions and the DPDP Act; the right to audit or receive audit reports; restrictions on subcontractors and data location; data return and deletion at exit, with certification of deletion; minimum security controls named explicitly; liability and insurance sized to the realistic loss. Under the DPDP Act a company deciding why and how personal data is processed remains accountable even when a processor does the work, so a vendor’s failure is not a defence you can hand to the regulator. Regulated sectors have further rules, such as the RBI’s expectations for outsourcing by banks and NBFCs, so check the sector regulator’s current requirements.

Access you grant is risk you own

A neglected angle is what you give the vendor in practice. Many supplier-originated breaches ride on connections you provided: a VPN account, an admin login on your CMS, an API key with wide permissions. Treat those like any privileged access. Give vendors named accounts with MFA, limit them to what the job needs, expire access when the project ends, and log what they do. That reduces your exposure whether or not their paperwork is perfect.

Fourth parties and software dependencies

Your vendor has vendors. A SaaS tool may be hosted on a major cloud and rely on an email provider and a support platform, any of which can fail or leak. You cannot assess everyone, but ask critical suppliers for their subprocessor list and how they manage them. For software components, a software bill of materials and dependency scanning serve a similar purpose inside your own engineering.

The questionnaire is a fine first filter. It becomes dangerous the moment a green rating on a spreadsheet is mistaken for assurance. Ask for proof where the stakes are high, accept a shorter check where they are not, and write the rest into the contract.

Figures and names in the examples are illustrative.

Leave a Reply

Trending