Last updated: August 2026 · Reviewed by Aaraddhya Bhatalkar, Co-CEO, New Resilience.

The short version: Low census usually traces to one leaking stage in the admissions funnel rather than a shortage of leads. The common instinct is to buy more leads at the top, which pours volume into a mid-funnel leak and wastes money. The better move is to measure conversion between each stage, find the single worst-leaking one, and fix it before adding volume. Doing that well depends on seeing each inquiry's true source and the stage where it actually stalled, which is the attribution problem most programs get wrong.

Treat admissions as a funnel, not a number

Whether you run residential, PHP, or IOP for mental health or substance use disorder (SUD), your admissions process is a series of stages, and someone falls out at each one:

• Inquiry received

• First contact made, ideally within minutes

• Benefits verified and financial picture clear

• Clinical fit confirmed

• Admit scheduled

• Admitted

Census is the output of that whole chain. When it drops, the useful question is not how many inquiries you got. It is which stage lost more people than usual.

Low census is almost always one leaking stage

When beds are empty, the reflex is to assume you need more leads, so programs turn up ad spend or push business development for more referrals. Sometimes that is right. More often the top of the funnel is fine and the problem is that a specific stage in the middle is leaking, so every extra lead you buy leaks out at the same place. You spend more and census barely moves.

This is why finding the bottleneck comes before buying volume. Fixing a stage that is losing half the people who reach it is usually cheaper and faster than doubling the number of inquiries, and it makes every future lead worth more.

How to find the bottleneck

Two measurements find it:

• Stage-to-stage conversion. For each step, what percentage of the people who reached it moved to the next one. The stage with the steepest drop, measured against what that step should convert at, is your constraint.

• Time in stage. How long inquiries sit at each step before moving. A stage where deals pile up and age is often the real bottleneck, because in admissions, time kills. A family waiting on a callback or a verification does not wait long before choosing somewhere else.

Look at both. A stage can convert acceptably on paper while quietly taking two days, and those two days are where you lose the people who were ready.

The usual suspects, and how each one is fixed

• Slow first contact. If inquiries sit before anyone responds, this is almost always your biggest leak and the highest-return fix. Speed to first contact matters more than almost anything else in the funnel.

• Slow or unclear verification. If people clear first contact but fall out before admit, verification of benefits is a likely culprit. A same-day, clearly explained verification keeps ready families moving.

• No follow-up on the undecided. Many families do not say yes or no on the first call. If nobody owns the second and third touch, these quietly disappear, and this is the leak programs most often cannot see.

• Scheduled but never admitted. If people are scheduled and then no-show, the gap between scheduling and arrival is your bottleneck. Confirmations, logistics help, and staying in contact through that window recover a surprising number of them.

The attribution problem: where did it come from, truly?

To fix the right stage you have to see the funnel clearly, and the hardest part of seeing it clearly is knowing where each inquiry actually came from. Attribution in this field is genuinely hard, and getting it wrong sends you to fix the wrong thing.

The reasons it is hard are worth naming. Families self-report the source inaccurately, saying “Google” when they saw an ad or “my therapist” when the therapist named you because a family member found you through ChatGPT. First-touch and last-touch both mislead, because a single admit often has several touches: an AI search, a directory listing, a therapist's recommendation, and a friend in recovery, all pointing at the same program. Pick any one as the source and you will misjudge the others.

The honest goal is to be directionally right, not perfect. A few habits get you there:

• Capture a source on every inquiry at first contact, even if it is imperfect. An imperfect source recorded consistently beats a perfect one recorded sometimes.

• Ask the open question at the right moment. “How did you first hear about us?” asked naturally during the first conversation beats a guess from a dropdown, and it catches the AI and word-of-mouth paths that tracking misses.

• Keep the referral source in the record. Relationship-driven admits are invisible unless you write down who sent them, and that is the data you need to protect your best referral relationships.

• Let more than one source count. The admit that was AI-surfaced and therapist-recommended is both. Forcing a single label throws away the truth.

• Read at the source level over time, not deal by deal. Any single inquiry's attribution is noisy. A source's pattern across a quarter is stable enough to act on.

Attribution and the bottleneck connect directly. Once source is captured, you can read conversion by source and by stage at the same time, and you will usually find that different sources stall at different stages. Online and AI-driven inquiries often convert well but only if first contact is fast. Referral-driven inquiries may need less speed and more relationship follow-through. Knowing which source stalls where tells you exactly which fix to make. We wrote more about crediting channels honestly in our piece on allocating a marketing budget.

Why this beats buying more leads

A program that fixes a stage losing half its people effectively doubles the yield of every lead it already pays for, at no extra acquisition cost. A program that buys more leads without fixing the leak pays full price for volume that drains out at the same spot. When census is soft, the first question should be where the funnel leaks, and only after that is tight should the question become whether you need more volume. It is also worth pairing this with a look at your payer mix by conversion, since some leaks are really a payer-economics problem in disguise.

Where a system helps

You cannot find a bottleneck you cannot see. The requirement is that every inquiry moves through defined stages in one place, with the source captured at first contact and the time in each stage recorded. A CRM built for behavioral health gives you that: a real pipeline with stages, time-in-stage visible, source attached to the inquiry, and conversion reported by stage and by source. With that in place, the leaking stage stops being a hunch and becomes something you can point to.

Where to start

Write down your admissions stages and, for last quarter, the conversion between each one. Find the steepest drop and the slowest stage. That is your bottleneck. Fix that one thing before you spend a dollar on more leads, and make sure a source is captured on every inquiry so you can tell which sources are stalling where.

If you want to see how New Resilience makes the admissions funnel, time in stage, and source-level conversion visible so you can find the bottleneck, book a 15-minute demo.