We all do oversight. At least, that's what we say. We run QC checks on vendor deliverables. We review data management plans. We double-check critical variables. We raise queries and chase them to closure. We sit in governance meetings. We tick the box. But here's the uncomfortable question — is any of that actually oversight?

In most cases, no. It's QC. It's being part of the process. And that is a very different thing.

We've confused "doing something about it" with "overseeing it"

Oversight has a simple, sharp definition: ensure that the vendor's process and quality of work meet the established requirements — and correct the process or performance, not the data or documents.

That last phrase is where most of us lose it.

When you QC data and raise queries, you're fixing the data. When you review a plan and comment until you're satisfied, you're shaping the plan. When you double-check the vendor's work and find the same issues quarter after quarter, you're not overseeing — you're running your own parallel process on top of theirs.

The data got cleaner. The plan got better. You added value. But the process that produced it didn't change one bit. And next trial, you'll be doing it all again.

ICH E6 (R3) now makes this distinction explicit: active, continuous, risk-proportionate oversight of the process — not compensating for gaps in it. But even without the regulatory push, this is just a more effective way to work. Doing "more oversight" the old way doesn't fix anything. It just burns more sponsor hours.

The way forward — our two cents

The question that changes everything is simple:

Ask it every time you touch a vendor deliverable. It's uncomfortable at first, because the honest answer is often "I'm part of it." That's fine. The discipline isn't to stop doing QC — QC has its place. The discipline is to know when you're doing it, and to capture the oversight step separately.

When you find an issue, resist the reflex to fix it yourself. Go back to the vendor and ask them to fix the process that produced it. Then check, at the next deliverable, whether the fix stuck. That tiny extra step is oversight. Without it, you're just forever catching the same fish.

And when you don't find issues on the next review — that's the win, not an anti-climax. It means the process is working. Log it as oversight, not as wasted time.

Taking action — as simple as it is difficult

None of this is rocket science. It is, however, a change of habit — and habits are harder to change than processes. Here's where to start.

As a sponsor or CRO

• Stop duplicating the work you pay for.

  If you outsource it, oversee it. Don't redo it. Every hour your team spends re-running the vendor's QC is an hour not spent on the process that produced it. And the vendor's process doesn't improve while you compensate for it.

• Structure oversight in three tiers.

  Company (vendor selection and qualification), portfolio (cross-trial trends and expectations), trial (daily work). Organise each activity at the right level. The higher up you push something, the less operational effort it costs later — and the greater the overall effect.

• Link oversight to risk.

  Use your trial risk picture to decide where oversight focus belongs — and, just as important, where it doesn't. Strong performance for a year? Reduce. Slipping? Increase. Don't oversee everything equally; you'll oversee nothing well.

• Document the non-findings, too.

  Oversight that isn't documented didn't happen — at least from a regulatory perspective. A one-line log entry saying "reviewed, no action needed" takes thirty seconds. Over a year, those thirty-second entries are worth more than any thick retrospective audit binder.

As an individual

• Ask the question. Every time.

  Am I part of this process, or am I overseeing it? Make it a habit, before every review and every meeting. The honest answer will often be "part of it" — that's fine, as long as you know.

• Separate the roles.

  When you have to do both — and you often will — do them consciously. One line in the oversight log, even if the underlying QC work took half an hour. If you don't separate them deliberately, the oversight component disappears into the operational noise.

• Change the question you ask your vendor.

  Stop asking "can you fix this query?". Start asking "what would it take to stop this happening at all?" It changes the conversation — and slowly, it changes the process.

• Reward process, not rescues.

  A quiet quarter — one where you found less, not more — is a sign that oversight is working. Celebrate it internally. The instinct to celebrate "good catches" is how we end up rewarding parallel QC instead of real oversight.

The views expressed in this article are strictly the opinions of the author. It must not be seen as a reference with regards to the authorities' interpretation of regulatory guidelines or laws. No liability or responsibility of any kind (to the extent permitted by law) including responsibility for negligence, is accepted.