Ask most people at a CRO what the Trial Master File costs, and they'll point to the system: the licensing fee for whatever eTMF platform is in place. That's not where the money goes. The expensive part is the labor spent making the file match the work that already happened.
In one CRO organization I led biostatistics for, I watched this cost up close — not as an outside consultant reviewing a budget line, but as the person who owned it. TMF-related labor and internal audit time — people and internal audit time alone, not external inspection costs and not system costs, which were billed and borne separately — ran close to a tenth of the study budget. That figure is specific to that organization and that portfolio of studies. I wouldn't tell you it generalizes to every CRO or every study without more data, and we're actively gathering more of it. But I'd be surprised if it's an outlier.
Where the time actually goes
It doesn't go into creating documents. Monitors, coordinators, and study staff produce the underlying work — visit reports, training records, correspondence — as a normal part of doing their jobs. The cost is in the second pass: someone, later, reconstructing which document belongs to which task, confirming it was filed correctly, checking that access was restricted the way the protocol required, and closing the gap between what the file says happened and what actually happened.
That second pass is reconciliation. It's necessary because most systems treat documentation as a separate output from the work that produced it — something you upload afterward, tag afterward, and therefore have to check afterward.
The alternative isn't more QC. It's less reconciliation.
The instinct in most organizations, mine included at times, is to respond to reconciliation cost by adding more QC — another reviewer, another checklist, another pass before the file goes to the sponsor. That helps, but it doesn't address the root cause: the record and the work are still two different things that have to be matched up after the fact.
The alternative is to make the record a property of the work itself — generated at the moment a task is assigned and carried forward automatically as that task is completed, rather than reconstructed by someone checking a box later. That's the premise Opuspec is built on, and it's the reason I'm building it: not because I read about this problem, but because I owned the cost of it on a real budget, for years.
What we don't know yet
We haven't broken this figure down further — how much is coordinator and QC labor specifically versus broader audit preparation — and we're not going to claim more precision than we have. That work is in progress, alongside conversations with other CROs and sponsors to see how well this generalizes beyond one organization's experience.