Everyone knows recruitment runs late. Far fewer teams can say which stage consumed the time, and that gap is why the same delay repeats on the next study.
Around 80 percent of trials fail to meet their initial enrollment target and timeline, with published estimates putting the resulting lost revenue as high as eight million dollars per day for a high-value product.
That figure appears in virtually every recruitment deck in this industry, and it no longer changes anyone's behaviour. The reason is worth saying plainly. The number describes an outcome. It says nothing about a cause, and therefore nothing about an intervention.
What would actually be useful is knowing, for a study running 90 days late, which part of the recruitment funnel produced the 90 days.
Most studies measure enrollment as a single line: participants randomised against a planned projection. When the actual falls below the plan, the study is behind. Every study dashboard in the industry shows this.
What the curve cannot show is where the divergence started. A study can fall behind because candidate volume never materialised, because interested candidates were never contacted, because contacted candidates could not get an appointment, because consent conversion was poor, or because screen failure ran well above assumption.
Those have entirely different remedies. Adding media spend to a study whose real constraint is a four-day lag between referral and first contact is money applied to the wrong stage. But because the measurement stops at the aggregate, the diagnosis usually does too, and the default response becomes adding sites, which raises start-up cost and oversight burden while often making per-site performance look worse.
Here is the structural problem. In most studies, the recruitment funnel is instrumented at both ends and almost nowhere in the middle.
Campaign platforms report impressions, clicks and form completions. The CTMS reports screened, consented and randomised. Between those two datasets sits the entire candidate journey: pre-screening completion, eligibility determination, referral to a specific site, first contact attempt, successful contact, appointment offered, appointment booked, appointment attended.
That middle section is where most candidates are lost, and in a typical study it is tracked, if at all, across a media agency's dashboard, a call centre's log, a referral spreadsheet and the site's own records. No common identifier joins them, so nobody can say what proportion of qualified candidates ever reached a site, or how long they waited.
A sponsor in this position cannot distinguish a demand problem from a conversion problem. Those require opposite responses, and one of them is considerably cheaper.
The useful model measures each transition as both a conversion rate and an elapsed time, so that volume loss and speed loss are separately visible.
Reach to pre-screen started. How many people who saw the study began a pre-screening questionnaire, by channel and by campaign.
Pre-screen to qualified. What proportion met the protocol's pre-screening logic, and which single criterion disqualified most of them. This last detail is disproportionately valuable and almost never captured.
Qualified to referred and contacted. How long between qualification and first contact attempt, how many attempts were needed, and how many candidates were never reached at all.
Contacted to scheduled. The conversion that quietly destroys recruitment performance, because a motivated candidate who cannot get an appointment for three weeks is frequently gone.
Scheduled to attended to consented. Attendance rate, then consent conversion, which isolates whether the consent process or the protocol burden is causing loss at the final step.
Consented to randomised. Screen failure rate against assumption. A high rate here almost always traces to eligibility criteria, which means it was created months before recruitment began.
Two findings recur when teams instrument the funnel for the first time.
The first is that site selection is doing more damage than recruitment tactics. More than a third of sites selected for clinical trials under-enroll, and around 11 percent fail to enroll a single subject. A study that activated forty sites and got meaningful enrollment from twenty-four had a selection problem, and the recruitment spend was applied downstream of the actual failure.
The second is that the plan was frequently unachievable. More than half of studies, 53 percent, take longer than planned to give activated sites enough time to reach 85 percent of target enrollment. Execution was being measured against a target that competent delivery could not have hit.
Both locate the cause upstream of the teams usually held accountable for the result.
This is a data-joining problem before it is an analytics problem, and it is the specific thing Cloudbyz Patient Recruitment is built to solve.
Candidates enter through omnichannel outreach across voice, SMS, WhatsApp and IVR, which matters because response rates differ sharply by population and channel, and because a channel that generates volume but not qualified candidates should be identifiable within days rather than at study close.
Pre-screening logic applies the protocol's eligibility criteria before anyone consumes a site visit slot. This converts a portion of what would have been screen failures into pre-screen exclusions, which cost almost nothing, and it surfaces which criterion is disqualifying most candidates while the protocol team can still act on it.
Referral routing sends qualified candidates to sites by capacity and geography rather than by whoever responds first, and scheduling closes the gap between interest and appointment that loses so many motivated candidates. eConsent then measures the final conversion directly, with comprehension checks, rather than leaving it as an unexplained drop.
Because all of this sits on the same platform as CTMS, EDC and RTSM, the candidate record connects to the enrolled subject record. Campaign attribution runs through to randomisation rather than stopping at form submission, and a screening bottleneck appears alongside a drug supply constraint on the same curve, which matters because the two are routinely mistaken for each other.
When the next enrollment review reports the study is behind plan, the productive follow-up is not how many more sites can be added.
It is: of the candidates who qualified last month, how many reached a site, how long did they wait, and where did the rest go? If nobody can answer, the intervention that follows will be a guess, and it will usually be more sites or more media, because those are the interventions available when the diagnosis is missing.
The 80 percent figure will not move until the industry stops treating enrollment as one number.