Six Numbers That Run a Therapy Practice — and What Each One Decides

Ask a practice owner how the practice is doing and you will usually get one of two answers: a feeling, or a spreadsheet with forty tabs. Both fail the same way — neither tells you what to do next. The feeling has no denominator, and the forty-tab dashboard buries the handful of numbers that drive real decisions under a pile of numbers that exist because the software could export them. A working dashboard for a therapy practice is smaller and stricter than either: six metrics, each with a formula you can compute from records you already have, each attached to exactly one decision, and each carrying a warning label about the volumes at which it stops being trustworthy. This article defines all six that way — formula, decision, warning label — so the dashboard you build this month is one you will still be reading next year.

The dashboards that fail are rarely missing data. They fail because no metric on them has an owner, a formula anyone could restate, or a decision anyone has agreed it drives. “Productivity” appears without anyone specifying the numerator. “Revenue” trends up and to the right while the practice quietly writes off a tenth of what it earns. And when a number does dip, nobody can say whether the dip is a problem or just arithmetic — because at the volumes a small practice runs, a single family’s vacation can move an attendance percentage by more than the change anyone is worried about.

The fix is not more data; it is admission criteria. A metric earns a slot on the dashboard only if it passes three tests. It has a precise formula — numerator and denominator you could compute by hand from the schedule, the remittances, or the referral log. It drives a named decision — when this number moves past a line you drew in advance, a specific action follows, taken by a specific person. And it has a stated failure mode — you know the volume below which it turns to noise and the ways it can look healthy while the practice is not. Six metrics pass those tests for most outpatient therapy practices. Here they are, one table, and then each in enough depth to actually use.

Read the table as a set of contracts, not a scorecard. The definition column is exact on purpose — most metric arguments inside a practice turn out to be two people using the same word for different fractions. The decision column is the admission ticket: if you would not take that action when the number moves, do not track the number. And the last column is the one dashboards always omit — the specific way each metric misleads when the denominator is small, which for a solo or two-clinician practice is most weeks.

Six is not a magic number, but the shortness is load-bearing. Each additional metric dilutes the attention the others get, and every number on this list already implicates the ones next to it: utilization falls when arrival rate falls, collections per visit falls when first-pass rate falls, and referral lag grows when utilization is genuinely maxed. The six work as a system — which is also why they divide cleanly into three families with three different review clocks.

Arrival rate and utilization are the pair that describe whether the hours you planned to sell became sessions. They fail differently, and conflating them is the most common dashboard mistake in therapy practices. A practice can run high utilization — every bookable hour has a name in it — and still bleed revenue to a poor arrival rate, because booked and attended are different events. The reverse also happens: families who show up reliably, spread across a schedule so fragmented that a third of bookable time sits in unusable twenty-minute gaps.

Two definitional rules keep the pair honest. First, arrival rate’s denominator includes no-shows and late cancellations — a cancellation with enough notice to refill the slot is a scheduling event, not an attendance event, and mixing the two hides the pattern you can act on. If your arrival rate is the number that is slipping, the fix lives in reminder cadence, policy, and slot design, which is its own topic — see the no-show playbook for the intervention side. Second, utilization’s denominator is hours you genuinely intended to book. If Friday afternoon admin time counts as bookable, utilization reads low and the dashboard quietly argues for cutting a clinician’s documentation time — the wrong conclusion drawn from a dishonest denominator.

The three money numbers describe one journey — a delivered visit becoming spendable dollars — measured at three points: how much arrives, how fast it arrives, and how much friction it meets on the way. They come from the same records: your remittances and your billing software’s claim history. Professional associations converge on essentially these measures for gauging financial health; APTA’s practice-finance guidance, for example, centers on collections performance measured against contractual allowed amounts and on how long receivables age before they are paid.

Collections per visit is deliberately not the contracted rate. It is banked dollars divided by delivered visits, per payer, so it already contains your denials, underpayments, and uncollected patient balances — which is exactly why it, and not the fee schedule, is the number that should drive payer decisions. The per-payer split is non-negotiable: a blended number averages your best contract with your worst and recommends keeping both. If the per-payer arithmetic is the decision you are facing, the private-pay-versus-insurance analysis works that decision end to end.

Days from visit to cash is the same journey on a clock. Its dashboard job is sizing your operating reserve and naming your slowest payers — a payer that pays slowly is charging you the float on your own payroll. First-pass claim rate is the friction gauge: the share of claims that get paid as submitted, with no human touching them twice. Every failed first pass has a reason code attached, and at small volumes the reason codes are the metric — five rejections with the same eligibility error is a broken intake step, not a five-point rate change. Payers explain those adjustments on the remittance advice using standardized adjustment reason codes, so the diagnostic trail is already in your hands; the denial-management workflow covers what to do once a pattern shows up.

The sixth number is the one most practices never compute, because it lives in nobody’s software by default: the days between a referral arriving and the evaluation actually happening. It is the earliest honest capacity signal you have. Utilization tells you the schedule is full today; referral lag tells you demand is outrunning capacity before families start giving up — and families do give up quietly, by booking elsewhere, long before anyone calls to complain. It is also a referrer-relations number: the pediatrician who sends three families and watches all three wait weeks has learned something about your practice that no marketing visit will unteach.

Track it per referral in a simple log — date received, source, date of evaluation — and read the distribution, not the average. The blended mean hides the payer whose authorization step adds days, and the referrer whose patients always land in your scarcest slots. This is the number that should trigger hiring conversations and hours changes ahead of the pain, and it pairs directly with how you run intake and the waitlist; the waitlist workflow is the operational other half of this metric.

Every metric above is a fraction, and fractions get jumpy when the denominator is small. A solo practice delivering twenty-five visits a week moves its arrival rate four points with one absence. A ten-claim week moves first-pass rate ten points per claim. None of this means small practices should not measure — it means the reading discipline differs from what a hospital revenue-cycle team does with thousands of claims. Three habits keep small-denominator numbers honest. Widen the window before you react: compute weekly, but act on four-to-twelve-week trends. Read the names before the number: at small volume, every metric movement is a short list of specific patients, claims, or referrals, and the list is more informative than the percentage. And distinguish a level from a change: a number that has always sat at one level is a fact about your model; a number that moved is a question about what happened — only the second one is an alarm.

A dashboard is a meeting schedule wearing a spreadsheet costume. The six numbers stay useful only if each family of metrics has a standing time on the calendar, a fixed short agenda, and the thresholds written down from the previous quarter. The rhythm below fits a solo or small-group practice; the point is not the specific minutes but that each number is read on the clock that matches how fast it can genuinely change.

1. 01

   Weekly, fifteen minutes: the schedule numbers

   Arrival rate and utilization for the week just ended, next to their four-week trends. Read the names behind any movement. The only decisions on this agenda are scheduling decisions: a family conversation, a slot change, a reminder-sequence tweak. Money questions are out of scope on purpose — they cannot be answered weekly and trying poisons the habit.

2. 02

   Monthly, thirty minutes: the money numbers

   Collections per visit, days from visit to cash, and first-pass rate, each per payer. Pull the reason codes on everything that failed first pass and group them — the groups are the to-do list. Reconcile what the aging report says against what the dashboard says; when they disagree, the aging report is telling the truth.

3. 03

   Quarterly, one hour: thresholds and contracts

   Recompute every threshold against the trailing quarter, per payer. This is where collections-per-visit trends become contract questions, where referral-lag distributions become hours-and-hiring questions, and where you retire any metric that produced zero decisions in three months — the admission test runs in both directions.

4. 04

   Annually: the model questions

   Once a year, the six numbers feed the big decisions — payer mix, headcount, hours, and pricing — with four quarters of per-payer history behind them. An annual decision made on annual data is judgment; the same decision made on last month’s blended numbers is a coin flip with a spreadsheet attached.

“

“A metric earns its place by having a formula, an owner, and a decision. Everything else on the dashboard is decoration — and decoration that pages you on a Friday night.”