Build a portfolio of sites the way a VC builds one of companies.

Every site I have ever qualified, I have qualified believing it would enroll. I have done the diligence. I have walked the patient path. I have talked to the PI personally. I have pulled the claims data, audited the patient flow, sat with the coordinator, asked about the parking lot. The sites I have brought onto studies have been the ones that passed every check we know how to run, and I have taken responsibility for that pass at every contract signing.

Approximately a quarter of those sites, in any given study, end up underperforming what the qualification predicted. Not because the qualification was wrong. Because the site is a small business — usually a single-physician or small-group practice — and small businesses are vulnerable to things that have nothing to do with the study. The PI's spouse gets sick. The lead coordinator quits without notice. The practice's lease comes up for renewal and the partners are arguing about whether to move. The EHR vendor pushes an update that breaks the patient-flow report. Any one of these things can take a site offline for a quarter, and none of them are visible during qualification.

This is the part of site selection that the conventional model — qualify rigorously, contract with the qualified, expect them to enroll — does not handle well. The model treats each site as an independent unit whose performance can be predicted from its attributes. The model is wrong on its assumption. Sites are small businesses. Small businesses have a baseline rate of operational disruption that no amount of qualification can predict away. The right response is not to qualify harder. It is to build the site list as a portfolio.

The analogy is to venture capital, and it is more than an analogy. A VC building a portfolio knows that any individual investment can fail for reasons unrelated to the diligence. The founder gets sick. The market shifts. A competitor lands a deal. The diligence cannot eliminate this risk; it can only inform the bet at any individual company. The portfolio absorbs the risk by being a portfolio — by sizing positions against an expected dispersion of outcomes, by building enough investments that the failures are absorbed by the successes, and by reserving capacity for follow-on bets when an early investment proves out.

Site selection on an IVD study has the same structure. Even with the best qualification, a quarter of sites will underperform. The right portfolio strategy is to size the active site list above the apparent need, on the explicit assumption that some fraction of activated sites will lose months for reasons we cannot see in advance. Activate eight sites for a study that needs six. Reserve activation budget for two more sites you have qualified but not yet brought live. When a site goes offline because the lead coordinator quit, you replace them from the qualified-but-not-active bench, not from a fresh qualification cycle that takes another six weeks you do not have.

This is how we now build site lists, and it is the most uncomfortable conversation we have with sponsors. Sponsors push back on activating "more sites than the study needs." From a finance perspective, the pushback is rational: each activation has a cost, and activating extras looks like over-investment. From a portfolio perspective, the pushback is wrong. The cost of a site that goes offline mid-study is materially higher than the cost of activating one more site at startup — because the offline site costs not just its own activation but also the timeline slip and the rescue work to bring its enrollment over to a replacement.

The portfolio approach also changes how we evaluate sites that look small. A solo practice with a great PI and a tight patient-flow fit, in conventional qualification, is a high-quality site that the CRO might activate. The same site, viewed through a portfolio lens, is a high-variance position — high expected value, high concentration of single-point-of-failure risk. The portfolio response is not to skip the site; it is to balance it with a partner site, ideally a multi-physician practice with comparable patient flow, so that if the solo PI has a bad month, the partner site can absorb the gap.

What this means, practically, for sponsor evaluations of CRO bids: a site list that names ten sites for a study that needs six is not a markup. It is a portfolio. The sponsor whose CRO is over-allocating sites is the sponsor whose study is more likely to hit its enrollment timeline. The CRO that promises to enroll the study on six perfectly qualified sites is the CRO whose study is more likely to slip when one of those six has an unforeseeable bad quarter.

None of this is an argument against rigorous qualification. The qualification work is what makes the portfolio possible — without it, you are not building a portfolio, you are guessing. The qualification produces the universe of investable sites. The portfolio decides which of them to back, in what proportions, with what reserves. Both disciplines are required. CROs that confuse the two — that treat heavy qualification as a substitute for portfolio construction — are running the rigorous half of the discipline and missing the half that absorbs the risk no qualification can predict.