Everybody optimizes the first 20% of patient recruitment. Trials die in the other 80%.

Nobody ends up in clinical research by accident. You don’t give a decade to a molecule because the hours are good. Every person I meet in this industry got here because at some point, something broke their heart and they decided to fix it.

Then the work happens to them. Enrollment slips two quarters. The protocol gets amended twice. The financing conversation swallows the operational one. By year six the patient has quietly turned into a row in a report. Not because anyone stopped caring. Because there’s so much else to carry.

I can’t forget the patient. I was one.

I have Stargardt’s disease. It takes your central vision first, which means the thing I’m looking at is always the thing I can’t see. Years ago a late-phase study that could have treated it shut down. Good science. Not enough people. One of its sites sat a half mile from my retina specialist’s office, and I drove past it while it was actively recruiting. I found out it had failed from a newsletter.

So I built a recruitment company. And I built it wrong.

The confession

I came out of e-commerce. Two Blind Brothers, the apparel company my brother and I built, taught us how to move people online better than almost anyone. So I looked at clinical trial recruitment and saw a familiar shape: better targeting, better creative, fill the top of the funnel. A conversion problem. My team could win conversion problems in our sleep.

But in e-commerce, the relationship ends at checkout, the final 20%. The customer has what they came for and you’re done.

In patient recruitment, checkout is the first 20%.

Here’s the part that made the mistake expensive. From day one, Clinical Enrollment was paid on success: qualified referrals and randomizations, not impressions, not clicks, not leads. If patients didn’t enroll, we didn’t get paid. And we still built the company around the same top-of-funnel metrics every other vendor sells.

The ads worked. The screeners filled. The referrals flowed. And the enrollment curve barely moved, because a qualified lead is not an enrolled patient, and everything that turns one into the other happens after the click. Other vendors invoice for the top of the funnel. We had bet our revenue on the bottom of it and were optimizing the top. Every patient who vanished between the screener and the site was money we had already spent and would never see again.

If you didn’t click through that: a thousand people raise their hands, two honest screens keep ninety-two, and nobody in this industry will tell you how many of their ninety-two ever stand in a clinic. So here is ours, counted both ways: one in five of the referrals we deliver to sites signs a consent form, and of the patients who physically make it to an on-site visit, 89% sign. The distance between those two numbers is patients who never reach the door, and closing that distance is the advocate team’s entire job. The complete program, stage by stage, is at the end of this article.

Nobody’s board approved a program to buy two million video views. They approved a program to enroll seventy patients in a CNS study.

What does staffing for the ICF look like? Around 85% of our people work after the patient raises their hand. The phone team running eighteen qualification questions. The patient concierge group I call our Mother Goose team, whose entire job is making a scared person feel seen. The account managers keeping sponsors current on every patient. The site relations people who live inside one study and know every coordinator’s vacation schedule.

The first call still goes out fast, because interest is perishable. An AI caller reaches a candidate in an average of 87 seconds from form submission and hands off live to a human the moment they qualify or ask for one. Speed is a tool. It gets you the conversation.

It cannot have the conversation. That takes a person.

I know how this sounds. A company paid on enrollment, employing the warm voice that talks a scared patient through her placebo question. You should ask what that incentive does to the conversation. We asked first. It’s why our people run on careful language: a patient prequalifies for a next step, and nobody on our phones ever implies enrollment or benefit, because the site’s own screening decides who’s eligible. We don’t. The concierge’s job is that the patient understands and feels safe, and if the honest answer to her question ends with her walking away, that’s the job working. A patient talked into a trial is a screen fail waiting to happen, and we don’t get paid for those.

Which brings me back to Maggie.

“Why are we behind?”

Every clinical operations lead gets asked some version of the same question, usually by someone senior enough that the answer matters. Why are we behind?

Answering it usually means pulling an IRT report, emailing eleven sites, waiting, and chasing a recruitment vendor who responds with a slide deck a week later. By the time an answer exists, it describes a situation that’s already changed.

We got so tired of watching sponsors live that way that we built them their own product. We call it The CE Script. A script is what clinicians call a prescription, and the joke is only half a joke: it’s not always flattering to read, and reading it once a day is the treatment.

Open it and every referral we’ve ever delivered has a live status, from second contact attempted to randomized. Days since referral. Days since the site last acted. Business days waiting. And when a site has gone quiet, one click drafts the follow-up email, with our team copied to clear whatever’s blocking. No IRT pull. No vendor chase. Nine seconds.

A site’s job is to run the trial. Ours is to make that easy.

Sites aren’t the problem in patient recruitment. Sites are drowning. A coordinator running four studies doesn’t need another portal, another login, or another vendor’s referral dropped into a shared inbox with a first name and a phone number attached.

You saw TOGO above. What it hands a coordinator is a decision, not homework: the patient’s own words, every screening answer, when she can actually take a call, and her records already pulled, in an average of about four days, with a face sheet on top. Logging what happened is one tap, which sets the follow-ups and reminders automatically, because we built it from site feedback instead of a product roadmap. No chasing. No guessing.

Our site relations team lives inside one study at a time. They know how many patients each site still needs, which of our referrals were good, and when the coordinator takes her kids to the beach in July. That last one sounds trivial. It’s the difference between a referral that gets called and a referral that ages out.

[SITE QUOTE 1. Ideal shape: a coordinator on the difference between our referrals and everyone else’s. Records-in-hand and “I don’t have to guess” is the strongest version.]

Study Coordinator, [site type]

Every screen fail is the protocol talking

When a patient doesn’t qualify, that’s not a dead record. That’s data somebody should be acting on.

If washout is knocking out a third of otherwise perfect candidates in the Southeast, the sponsor should know this month, not at the next steering committee. If one site fails phone screens at twice the rate of the others, that’s a training issue or a question being asked in a way that scares people, and either way it’s fixable. If an exclusion criterion is quietly eliminating a population the protocol never meant to exclude, that’s an amendment conversation worth having while there’s still time to have it.

So we report every failure reason, at every site, every time. The sponsors who use it well make the next thousand patients cheaper than the last thousand.

Everyone’s optimizing the ends. The middle is where it’s hard.

Look at how this category is built. One camp optimizes site databases and feasibility. The other pushes as much raw volume into the top of the funnel as the media budget allows. Both are legitimate businesses. Both live at the ends of the process, where the metrics are fast and the work is automatable.

The middle gets skipped because the middle is unglamorous, headcount-heavy, and impossible to automate your way out of. The hard part of recruitment was never attention. It’s conviction. A qualified person genuinely deciding to walk through a site’s door. A site trusting your referrals enough to call them first. A sponsor trusting your data enough to stop double-checking it.

Those three sentences are the same sentence. It’s all trust, and trust is built by hand.

And if you’re reading this mid-trial and behind, with sites already fatigued by somebody else’s referrals: that’s exactly when the middle matters most. We don’t start those engagements from scratch. We keep a first-party community of 8,000 to 10,000 patients with confirmed diagnoses across our therapeutic areas who’ve already told us they want to hear about research. We’re not finding them. We’re activating people who are already waiting.

There are bigger numbers I could put in a band up there. Trials we support recruit 4.4x faster. On-site screen fails drop 18%. Our average sponsor saves $19.2 million against the cost of the delay they were headed toward. And you should be exactly as skeptical of those averages as I’d be in your chair, because averages are where vendors hide.

So instead of another average, here’s one program, opened. A Phase 2 CNS study we just finished, from our end-of-trial report, counted top to bottom.

Read the bottom of that table again. One in five of our referrals signed a consent form; the industry’s referral-to-ICF norm is under 5%. At fourteen of the trial’s sites, our referrals made up more than half of everything they randomized. That’s what the unscalable middle produces when you refuse to skip it: not better traffic, better arithmetic, at the stages your board actually asked about.

[SITE QUOTE 2. Ideal shape: the operational proof. Show rate, screen-fail rate, or responsiveness. “These patients actually show up” is the money version.]

Study Coordinator, [site type]

Six weeks after a woman iced her feet at midnight and almost scrolled past an ad, she sat in a clinic waiting room with her records already on the coordinator’s desk and a text from Leslie on her phone. She wasn’t a lead. She was never a lead. She was Maggie, and everyone who touched her file knew it.

[PATIENT QUOTE 2. Ideal shape: closing beat. Someone who enrolled, in plain language, on what it meant to be found. Short. This lands right before the CTA.]

Enrolled participant, [indication] study

None of those numbers came from having better software than everybody else. They came from refusing to skip the middle.