When metrics measure more than they improve

Clinical trial sites are measured constantly: activation timelines, enrolment, screen failure rates, data quality, deviations, safety reporting, patient retention, and more. These metrics matter: they reveal operational risk, quality concerns, and areas needing intervention.

Yet, despite all this measurement, site performance often doesn't improve. Why? Because measuring performance isn't the same as improving it. The question we should be asking is not, "Are we measuring site performance?" but, "Are the metrics we collect providing actionable insights early enough to improve performance?". Too often, the answer is no.

The data visibility gap

Sponsors and CROs use centralised analytics, risk-based quality management (RBQM), benchmarking, and cross-site comparisons to identify trends and outliers. They can often see whether one site is performing differently, whether a screen failure pattern is isolated or study-wide, or whether query ageing is becoming a risk.

Sites, however, often cannot see that same picture. A site may know its enrolment number, screen failure rate, or query backlog, but not whether those figures align with comparable sites on the same protocol or reflect a broader study issue.

That is the real performance gap: sites are expected to improve against metrics they often cannot fully see, compare, or contextualise.

From retrospective metrics to early warning signals

Site metrics are frequently treated as retrospective evidence, rather than operational signals. A missed enrolment target, accumulated deviations, or unresolved queries may all be important indicators. But by the time they appear in formal reporting, the opportunity for early correction has narrowed. Sites and CROs need more leading indicators of operational risk, such as prescreening-to-screening conversion trends, screen failure reasons, time since last patient enrolled, query ageing, repeated query themes, and delayed first data entry.

These signals help sites adjust staffing, re-train coordinators, revisit recruitment sources, clarify protocol requirements, or prepare for patient conversations earlier. Sometimes they point to a protocol issue that may justify clarification or an amendment. If deviations spike, is the EDC unclear? If under-enrolment persists, are sites interpreting the inclusion/exclusion criteria differently? Is the sponsor seeing similar themes across sites and geographies, or is the concern specific to one site or monitor?

CRO operations leader Nicole Stansbury, senior vice president of global clinical operations at Premier Research, captured the opportunity this way: “RBQM has helped the industry identify the right risks earlier. The next evolution is making sure those signals reach the sites in a form they can use, while there is still time to change the outcome.”

Enrolment alone is a blunt instrument

Enrolment remains one of the most visible and heavily weighted site metrics. It is easy to count, compare, and tie to timelines. But enrolment alone is a poor proxy for site performance.

A site that enrols quickly may later struggle with retention, visit completion, data quality, or protocol adherence. A slower enrolling site may be prescreening carefully, serving a complex patient population, or producing cleaner data with stronger retention.

Enrolment can also be misleading when it ignores site activation timing and actual recruitment opportunity. A rescue site may start up quickly, but see enrolment close soon after opening, leaving it with poor-looking numbers despite limited opportunity. Enrolment per month, enrolment relative to activation date, or enrolment against site-specific opportunity may provide a fairer picture.

Screen failure rates make this even clearer. A high screen failure rate may suggest poor prescreening. It may also reveal protocol misfit, difficult eligibility criteria, recruitment messaging that attracts the wrong candidates, or unrealistic feasibility assumptions. If the same pattern appears across many sites, pushing each site harder will not solve the problem.

Context is the start of root cause

The same principle applies to protocol deviations, query resolution, and activation timelines. Deviations may reflect site error, but also protocol complexity, ambiguous instructions, technology failures, participant burden, supply delays, or insufficient training. A delayed activation milestone may reflect site readiness, but also contracting delays, budget negotiation, or regulatory approval timelines.

Without context, the metric identifies where to look, but not what to do. Site executive Nancy Cleverley, senior vice president of growth & expansion at AMR Clinical, summarised the issue well: “A raw number can tell you where to look, but it rarely tells you why something happened. If there is no way for the site to add context, the performance story is incomplete before the conversation even begins.”

This is not a call to excuse poor performance. Sites are accountable for what they control. Sponsors and CROs need reliable partners who can enrol appropriately, protect participants, follow the protocol, enter clean data, resolve issues, and sustain quality. But accountability works best when expectations are visible, data are timely, and performance can be interpreted accurately. If metrics influence escalation, oversight, trust, or future site selection, sites should have enough access to those metrics to understand and respond.

Site performance is a business question

A more complete view of site performance should include the business realities of running research. Traditional metrics focus on sponsor-facing execution, but they do not fully capture portfolio mix, sponsor mix, start-up burden, payment milestones, investigator workload, or disconnected technology burden.

Study start-up is a common pressure point. Many sites aim to open Phase 3 studies within 90 or 120 days, yet, documentation provided before the site initiation visit (SIV) is sometimes incomplete or in draft form. Without complete information, sites may delay pre-award work to avoid unreimbursed rework across budgets, coverage analysis, regulatory review, IRB submission, and contracts. A study with incomplete documents may fall behind a start-up-ready study due to reasonable business prioritisation.

These factors affect performance directly. If payment timing is unclear, sites struggle to forecast cash flow. If technology, protocol, or vendor requirements are unclear early, sites may later look inefficient for reasons built into study execution.

A shared operating model

A better model would treat site performance metrics as shared operational infrastructure.

That means providing anonymised benchmarking where possible, so sites know whether they are aligned with study norms or trending toward risk. It requires comparative context across enrolment, screen failure, data quality, protocol deviations, and compliance, plus leading indicators defined before the trial begins.

Sites need timely visibility into signals they can act on, such as query trends, screen failure reasons, inventory risks, reconsent triggers, protocol deviation patterns, and enrolment projections.

An improved model would let sites annotate metrics, validate data, and contribute to root cause assessment. It would also recognise that business and operational burden are not separate from site performance; they are part of what makes performance possible.

Sites do not need every sponsor or CRO report, nor identifiable peer-site data. They need timely, comparative, contextual information that helps them understand whether a metric reflects their own performance, a study-wide issue, or an operational dependency outside their control.

Measurement should improve performance

Sponsors, CROs, and sites ultimately want the same outcome: trials that meet enrolment targets, protect participants, maintain compliance, produce usable data, and conclude on time. The industry has invested heavily in tools that can detect risk. The next step is making those insights more usable.

The industry does not need additional ways to inform sites how they performed after the fact. It needs better ways to help sites perform while there is still time to change the outcome.

About the author

Christine Senn, PhD, is senior vice president of site-sponsor innovation at Advarra, an industry leader in regulatory reviews and clinical research technology. Senn began her clinical trial career at the University of Vermont College of Medicine while pursuing her Doctorate in Psychology. She then spent 16 years in various leadership positions, including chief operations officer at the site network IACT Health, which later became Centricity Research. Senn was also the 2023 Chair of the Association of Clinical Research Professionals (ACRP), is an ACRP Fellow (FACRP), and is double certified as a Certified Clinical Research Coordinator (CCRC) and Certified Principal Investigator (CPI).