Clinical Trial Site Software: Buy the Management System, and Build Only the Operational Layer Above It

When is off the shelf genuinely the right call here?

Clinical Conductor and RealTime CTMS are the two clinical trial management systems most independent research sites end up on, and for a large share of sites they are the correct and final answer. They hold the clinical trial agreement, the budget, the per visit payment terms and the invoicing, and they hold them properly. Rebuilding any of that is money spent to arrive back where you started.

Buy, and stop there, if this describes your operation:

• One site running under about eight concurrent protocols.
• Two or three coordinators who can each name every active subject without looking.
• A deviation log dominated by clinical judgement calls rather than missed visit windows.
• One therapeutic area, or one or two sponsors whose own tooling already covers most of your day.
• Paper source that your monitors have never raised a finding against.

At that shape, the arithmetic a build automates is arithmetic a person can still do reliably. Eight protocols is a whiteboard someone reads across in thirty seconds. The subscription costs less than the discovery phase of a custom project, and the difference buys a second coordinator.

There is a second thing you should not build at any size, and it is worth saying before anything else. Do not build electronic data capture. Medidata Rave, Veeva CDMS, Oracle Clinical One and Medrio belong to the sponsor and are mandated study by study. Your system sits upstream of them by design, and any proposal that includes replacing them is a proposal to enter a fight nobody asked you to have.

When does a custom build actually pay off?

The threshold is not revenue, it is cognitive load. Five signals tend to arrive together, and two or more of them makes a build case:

• More than a third of your deviations are visit window, missing source or missing signature, which means the problem is operational rather than clinical.
• Coordinators lose six to nine hours a week to transcription, chasing signatures and reconciling paper against the sponsor system.
• Monitoring preparation takes two days of binder pulling before every visit, four times a year, across every active protocol.
• You lost a study at feasibility because you could not evidence your enrolment history with anything better than an anecdote.
• Your best coordinator maintains a personal spreadsheet the whole site depends on.

That last one is the tell that ends the argument. The spreadsheet is your requirements document, and the fact that somebody had to write it means the packaged tool already failed at the thing you needed most. Open it and you will usually find two things: a versioned schedule of assessments, and a days to window close calculation across every protocol at once. Those are precisely what a build formalises, and precisely what the packaged tools do not attempt.

Check your own deviation log first. If fewer than a third of your deviations are window or source related, fix process before writing a cheque.

How do they compare on the things that matter in this industry?

Schedule versioning. A packaged system carries the visit schedule that existed at study startup. When an amendment moves the Week 12 window from plus or minus seven days to plus or minus three and adds an unscheduled draw, most sites re-enter it by hand or rely on the lead coordinator remembering. Ask a vendor how a subject who re-consented on a specific date gets forward windows recomputed while subjects still on the prior version do not. A configuration screen cannot express that. Dated protocol versions can.

Cross protocol ordering. The question a coordinator has at 7:40am is not what is happening today, it is what closes soonest across everything. Packaged tools sort by date and by study because they were built one study at a time. Sorting by days to window close across all protocols at once is the single screen that changes a working day.

Source verification burden. Generic electronic source products exist, and sites abandon them around protocol four because every worksheet is rebuilt as a form from scratch. Ask what happens when you hand over a sponsor's Word source worksheet. If the answer is a two week build per protocol, your operation will drift back to paper by the third study.

Capacity in hours rather than names. Staff assignment as a name field cannot tell you that a Day 1 oncology dosing visit is five and a half hours with a timed series while a device follow up is forty minutes. No packaged tool models procedure time, because it varies per protocol and per site.

Data portability under audit. Ask how the complete record leaves whatever you run, including the audit trail with reason for change capture rather than only current values. A regulated record you cannot export intact is a finding waiting for a date.

What does total cost of ownership look like at your scale?

On the build side, from Digital Heroes delivery experience, three shapes recur. A window engine alone, meaning protocol and schedule of assessments with versioning, a subject roster with consent version pointers and the cross protocol coordinator dashboard, runs $35,000 to $65,000 in seven to ten weeks. A focused first release adding electronic source for two or three protocols with typed fields, ranges, an append only audit trail and compliant electronic signature on the principal investigator review step runs $60,000 to $130,000 over twelve to sixteen weeks. A full platform adding the enrolment funnel and forecasting, coordinator capacity, agreement linked per visit invoicing, health record cohort integration and the retention engine runs $150,000 to $400,000 phased over six to twelve months.

Validation under 21 CFR Part 11 sits inside those bands rather than on top, and it adds 15 to 25 percent to the engineering line. Whether you land at 15 or 25 is decided in week two, because an append only audit trail designed into the data model is cheap and the same requirement retrofitted onto tables that have been overwriting rows is expensive. Health record integration is typically $20,000 to $45,000 of engineering, and engineering is not the constraint. Approval through a vendor partner programme can take months before code starts, while a nightly extract negotiated with your own health system information technology group often produces the same pre-screening list a quarter earlier.

After go live, budget 18 to 22 percent of build cost per year. Hosting for a three site network runs $400 to $1,200 a month. Each new protocol takes two to four weeks to onboard once the extraction pipeline exists, which at twenty concurrent protocols is a standing line. Change control is the recurring cost unregulated software does not carry: every material change needs impact assessment, testing evidence and a documentation update, so a change request that would be trivial elsewhere has a paperwork tail here.

On the buy side, the honest comparison is not the subscription, because you are keeping the management system either way. Price the gap instead. Take twelve months of deviations, count the share that are window, source or signature rather than clinical, and attach principal investigator time on each memo to file and coordinator time on each corrective action. Add monitoring preparation days. Then add the line everybody leaves out: replacing a certified coordinator lands around $40,000 to $60,000 once recruiting, ramp and the deviations during ramp are counted.

What does the hybrid look like, and when is it the honest answer?

For nearly every site that builds, this is the answer, and we give it whether or not it wins us work. Keep the management system. Build the operational layer above it.

In practice that is three pieces:

• The window engine, $35,000 to $65,000. Protocol versions with anchor events, day offsets, windows and required procedures. Subjects carry a consent version pointer, and the dashboard sorts by days to close across every protocol at once.
• Electronic source for your two or three worst protocols, inside the $60,000 to $130,000 release. Choose the ones that generate the most deviations rather than the ones that are easiest to model, because the extraction pipeline built for those makes every later protocol a two to four week job.
• Reconciliation against the sponsor system rather than push into it. Compare your source against what has been entered and surface the differences to a coordinator. That captures most of the operational benefit without a connector estate the next protocol will invalidate.

The split keeps your finance team out of the migration entirely, which matters more than it sounds. Contracts, budgets and payments stay where your controller already trusts them, the custom system becomes the daily driver, and nobody has to argue about which system is the record. It also keeps the number near $126,000 rather than near $400,000, because the expensive half of a full platform is the half you are deliberately not rebuilding.

Which should you choose, by operator size and stage?

• Single site, under eight concurrent protocols, two or three coordinators. Buy Clinical Conductor or RealTime CTMS and run disciplined paper source. Do not build.
• Single site, eight to fifteen protocols. Still buy, but start counting. Categorise every deviation for two quarters as window, source, signature or clinical. That count is the only evidence that makes the next decision cheap instead of theoretical.
• One site past fifteen protocols, or two sites. The decision point. Build the window engine alone at $35,000 to $65,000 and see whether the deviation category you were worried about actually falls before funding anything else.
• Three or more sites, or a network with mixed sponsors and therapeutic areas. Build the focused first release: window engine plus electronic source with audit trail and signatures for your two worst protocols, $60,000 to $130,000 over twelve to sixteen weeks.
• Network competing for sponsor selection on enrolment performance. Build the full platform, sequenced. Window engine and source first, then the pre-screening funnel with coded screen failure reasons, then capacity and invoicing. Twelve months of funnel history is what wins the next feasibility, and history is the one thing you cannot buy.

Two conditions apply to every build row. Design the append only audit trail in week two rather than month four, because that decision alone separates validation at 15 percent from validation at 25 percent. And do not migrate paper. Run new protocols on the new system and let existing binders age out with their studies, since digitising closed subject records creates a reconciliation question you will answer to a monitor for no benefit.

If you would rather someone argued with your brief than agreed with it, Digital Heroes starts every engagement with a signed specification covering the data model, permissions and acceptance criteria, which is what keeps a fixed price fixed. The document is yours whichever way you go.