n the early 1990s, sponsors stopped buying clinical research one study at a time. Faced with growing pipelines, fragmented suppliers, and rising trial complexity, they reorganized how they engaged contract research organizations (CRO). They moved from transactional, study-by-study contracting to preferred-provider relationships, strategic partnerships, and later to functional service provider (FSP) models that allocated entire categories of work to a small number of trusted partners. The shift was uneven, but the direction was clear: Sponsors wanted fewer relationships, deeper integration, and a way to buy reliability and predictability rather than capacity.

A similar shift is now underway at the site layer, and it has gone largely underdiscussed. Sponsors are moving from site-by-site selection to network-level allocation. The pattern is familiar to anyone who watched the CRO market 30 years ago: a small but growing number of credible delivery organizations, professionalized operations, standardized quality systems, and an emerging willingness on the sponsor side to commit volume in exchange for predictability. Multisite Clinical Research Corporations, or MCRCs, are to the site layer what strategic partner CROs became to sponsors in the 2000s. They are organizations large enough, consistent enough, and operationally mature enough to be allocated work rather than merely selected for it.

This shift raises an objection I have heard in almost every sponsor conversation. Concentrating sites with one provider, the argument goes, puts too many eggs in one basket. This Commentary argues that this objection has it backwards: Allocating volume to a single accountable network does not concentrate risk but in fact reduces it, and it does so in ways that align with where regulators are already pushing the industry.

The Real Risk: Variability, Not Concentration

The concentration-risk argument assumes that every site is equally reliable and that the only variable worth managing is how many baskets you spread your eggs across. Anyone who has run trials at scale knows this is not how sites perform. In my CRO years, we had a saying: Some sites will carry a study, and some will barely move it; the frustrating part was never knowing in advance which was which. Independent research points the same way: roughly half of all sites in a typical global phase 3 trial either fail to enroll a single patient or fall short of their target, a pattern that, according to Tufts Center for the Study of Drug Development research, has held for more than two decades. This uncertainty, not the number of sites, is what undermines trials. It shows up as slow starters, nonenrolling sites, and a long tail of underperformers that quietly drag a study past its timelines.

Variability in site performance is the risk that sponsors should manage. Network-level allocation addresses it directly. An MCRC does not promise that every site performs identically but provides a structure that narrows the spread between the best and worst sites. It does this by applying the same standard operating procedures, the same training, the same technology, and the same governance across all of them. The value of a network is not that any single site is guaranteed to excel. This approach reduces the uncertainty introduced by multiple sources of variability.

The Evidence: What Benchmarking Data Shows

This is more than just my experience talking. The Association for Multisite Research Corporations (AMRC) has benchmarked MCRC performance against other site types in partnership with one of the world’s largest pharmaceutical companies. The analysis covered 40 studies across 100 MCRC sites and 736 non-MCRC sites in the US, all completed since 2020. Some of the findings, on quality specifically, have already been discussed in Global Forum. I want to draw out what they mean for how sponsors buy delivery.

Start with speed, because it is the most visible. MCRC sites moved from final protocol to site ready in an average of 199 days, against 264 days for non-MCRC sites. From site ready to first subject in, MCRC sites took 57 days against 87. That is a meaningful head start on any timeline.

But speed is not the finding that matters most for this argument: Predictability is.

The site-ready-to-first-subject-in cycle was not only faster at MCRC sites, it was more predictable, with a smaller spread around the average. Across the other cycle-time measures, where average performance was broadly similar between site types, MCRC sites were still consistently the more predictable of the two. This is the variability thesis showing up in the data. A sponsor allocating volume to an MCRC is buying a narrower range of outcomes, which is exactly what makes a program plannable and more predictable.

Predictability did not come at the expense of quality. Important deviations as a share of total deviations were lower at MCRC sites, at 13.3% against 14.7%. Query volumes and resolution times were comparable between the two groups, and data-entry timeliness was similar, with MCRC sites again showing less variation. MCRC sites were as productive as their counterparts on enrollment and, critically, had a lower proportion of nonenrolling sites. The long tail of dead-weight sites, which quietly wrecks a recruitment plan, was shorter.

From Part-time to a Professional Undertaking

There is a structural reason MCRC performance is less variable. For most of its history, the site industry grew up as an amateur undertaking. Trials were run by part-time physician-investigators, fitting research around a clinical practice and supported by staff who often moved within a year or two. The result was the well-documented “one-and-done” problem, where a large share of investigators ran a single trial and never returned. Every one-and-done PI is a site that starts from zero, with no accumulated operational learning. Variability is the natural consequence of an industry built this way.

MCRCs change the underlying economics. Research is the core business, not a sideline, so the people running trials stay. AMRC’s benchmarking bears this out starkly. Principal investigator turnover at MCRC sites ran far below that of the wider industry, at 7.8% versus 54.2%, and coordinator turnover was also markedly lower, at 15.4% versus 33%. Stable career research teams are the mechanism that produces low variance. They carry operational memory from one study to the next, and that accumulated competence is precisely what a sponsor is buying when it allocates to a network.

The Diagnosis is Settled

Put the pieces together and the concentration-risk objection collapses. Allocating volume to a single accountable network does not pile up risk; it compresses the variability that causes trials to miss their timelines, and it does so without trading away quality. The data supports the diagnosis. The harder question is what sponsors and MCRCs should each do about it.

Part two of this commentary will set out the practical model: why this shift aligns with where regulators are taking risk-based quality management, what the emerging enrollment data—including a diversified, blinded portfolio of real-world case studies spanning six therapeutic areas—tells us about the power of scale, and the specific commitments that both sides must make for network-level allocation to work. The lesson from the CRO market is that these partnerships succeed only when the metrics, governance, and accountability are built in from day one.