Scheduling templates and real capacity

Paper capacity and real capacity

Start with the arithmetic that nobody does. Take one clinician. Count the sessions their template says they hold in a year. Then count the sessions that actually happened: subtract vacation, meetings, administrative time, the academic half-day, the sessions cancelled because they were covering elsewhere, the ones released too late to fill. Then, of the slots in the sessions that did happen, count the ones that were bookable at the time somebody was calling — not the ones marked held for a purpose that never materialized. The number you end with is real capacity, and in most practices it is a good deal lower than the number in the template.

This matters because every conversation about access is conducted in terms of the paper number. Leadership sees a template with a certain number of slots per week, multiplies, and concludes the practice has enough capacity for its panel. The schedulers know otherwise and cannot prove it. The clinicians know otherwise and are asked to explain why their sessions look underbooked while the wait is five weeks. Everybody is arguing from a different number, and none of the numbers is real capacity.

Producing that one number, per clinician, per quarter, is the highest-value analysis in this domain. It is not hard. It is mostly counting, and it changes the conversation immediately.

Measure access the way patients experience it

The standard operational measure for access is the wait until the third next available appointment of a given type with a given provider, sampled on a regular cadence. It is deliberately the third rather than the first, because the first opening is frequently an artifact — a cancellation from ten minutes ago that nobody could have planned around. The third is a much better estimate of what a person calling today will actually be offered.

Measure it per provider, per visit type, and per location, sampled the same way on the same day of the week, and keep the series. A single number for the practice hides everything interesting. The pattern you are looking for is not the average; it is the spread between providers in the same specialty, and the trend within a provider over months.

Alongside it, two supply measures that most systems can produce.

Fill rate is the fraction of bookable slots that had an appointment on them. Low fill with a long wait is the diagnostic signature of a template problem: the openings exist but not in a form the schedulers can use, usually because they are the wrong visit type or held for a category that has no demand.

Utilization in minutes is arrived minutes divided by available minutes. This one catches what slot counting misses. A session running twelve fifteen-minute slots and a session running six thirty-minute slots look different in slot counts and identical in minutes, and the second may be entirely appropriate. Counting slots produces the argument; counting minutes produces the discussion.

Holds are the first thing to look at, every time

Nearly every template contains slots reserved for a purpose: same-day acute, post-operative follow-up, a referral partner, a procedure, a specific clinic day. Holds are a legitimate and useful instrument. They also decay, silently and continuously, and the decay is the single most common source of phantom capacity we encounter.

What happens is ordinary. A hold is created for a real reason. The reason ends, or the volume shifts, and nobody removes the hold because removing it requires knowing it is stale, and nothing reports staleness. So the slot sits unbookable through the entire period a scheduler could have used it, and releases automatically at some short horizon, or never.

Three measurements will tell you the whole story. What fraction of held slots were eventually used for their held purpose. What fraction released and then went unfilled because they released too late. And how old each hold rule is, and who created it.

The fix is a policy rather than a system: every hold has an owner, a stated purpose, an automatic release horizon, and a review date. Holds that are used for their purpose less than some agreed fraction of the time get shortened or removed at review. This is a quarterly meeting with a report, not a software project, and it routinely recovers more bookable capacity than anything else on the list.

Release timing deserves its own thought. A hold that releases the day before is nearly worthless, because the patients who needed it were told five weeks and made other arrangements. A hold that releases seven to ten days out is usually recoverable. Look at your own booking-to-appointment interval distribution and set the horizon from it rather than from habit.

Slot length: settle it with the timestamps

The recurring argument in every practice is whether a visit type needs twenty minutes or fifteen. It is usually conducted as a contest of impressions between operations and clinicians, and it is entirely settleable with data the system already collects.

Most scheduling and rooming workflows record when a patient checked in, when they were roomed, when the clinician entered, and when the visit ended. Those timestamps are imperfect — people forget to click — but in aggregate they are more than good enough. Compute the actual duration distribution per visit type per provider, and look at the median and the upper tail rather than the mean.

What you will find is consistent across settings. Providers legitimately differ, sometimes by a factor approaching two, for reasons including panel complexity, teaching, language and their own style. The distribution is right-skewed, so a template built on the median will run late every day and a template built on the ninetieth percentile will waste capacity. And some visit types are systematically mis-slotted in one direction — usually a type that was created years ago and whose content has changed since.

The productive move is not a uniform slot length. It is a per-provider template that fits that provider's measured distribution, with an agreed rule about what happens in the tail. A clinician shown their own duration data will usually engage with it seriously, because it is their data and it is not an accusation. A clinician told that the standard is now fifteen minutes will not.

One caution that is worth stating plainly: shortening slot length increases throughput on paper and increases the chance of running late, and running late has a compounding effect through a session that ends with everybody waiting. If the practice already finishes an hour behind most days, the honest finding is that the template is already too tight, and the answer is a longer slot or fewer of them, not a shorter one.

Symptom	Usual cause	What to measure	Effort to fix
Long wait, low fill rate	Holds, or wrong visit types in the openings	Held-slot usage; fill by visit type	Low — policy and template edits
Long wait, high fill rate	Genuine demand above supply	Demand arithmetic; panel size	High — staffing or scope
Sessions run late every day	Slot length below measured duration	Duration distribution per provider	Moderate — and clinicians will engage
Idle afternoons with a long wait	Late releases; unbookable slot types	Release horizon vs booking interval	Low
Added a provider, no improvement	Ramp, template not built, panel not assigned	Real sessions delivered vs planned	Moderate
Two providers, wildly different templates	Historical accretion, some of it justified	Duration and case mix per provider	Moderate — converge, do not standardize
New patients cannot get in at all	New-patient slots held and released late	New-patient third next available specifically	Low

The demand side, in one calculation

Supply analysis without a demand estimate produces confident recommendations that do not survive contact. The estimate is not complicated.

Take the panel of patients attached to a provider. Compute visits per patient per year from your own history, ideally split by a rough complexity or age band, because the difference between bands is large. Multiply. That is annual demand from the existing panel. Add the new-patient demand you want to serve. Compare against real capacity as computed at the top of this article — sessions actually delivered, times bookable slots, times the arrival rate.

Two things usually fall out. The gap is bigger than expected, because real capacity is lower than paper capacity and visits per patient per year is higher than people guess. And the gap is not uniform across the week — demand concentrates on Mondays and around holidays, and template supply is flat.

That second finding is often more actionable than the first. Matching the shape of supply to the shape of demand costs nothing but template editing, and it is invisible until somebody plots both curves on the same axis.

Working the backlog down is a project, not a policy

The approach that shortens waits durably is to work the existing backlog down and then keep supply and demand matched day to day, so that most of what arrives today can be handled soon. It works, and it is misunderstood in one specific way that causes most of the failures.

The misunderstanding is treating it as a template change. It is not. Changing the template to make room for near-term demand while a backlog exists produces the worst of both: the backlogged patients still wait, the new openings are consumed by the backlog anyway, and everyone concludes the approach does not work. The backlog has to be worked down first, deliberately, with temporary extra capacity, a review of who on the waiting list still needs to be seen, and a decision about which visits can be handled another way. That is a project with an end date, and it is the part organizations skip.

The second thing that undermines it is carve-outs. Every additional protected category fragments the pool of slots, and a fragmented pool cannot absorb variation. Practices that hold slots for six different purposes have six small queues instead of one large one, and small queues have much worse waiting behaviour than large ones for exactly the same total capacity. Fewer categories is nearly always better, and it is the recommendation clinicians resist most.

Third, some visits do not need a slot. A result review, a medication adjustment, a stable follow-up in some specialties can be handled by message, by a nurse visit, or on a longer interval. Every visit moved off the template is capacity created at zero cost, and the decision is clinical, which means it has to be led by clinicians rather than presented to them.

The data will fight you a little

Five defects show up in almost every scheduling dataset. Find them in week one.

Status codes are used differently by site. Cancelled by patient, cancelled by clinic, rescheduled, bumped, left without being seen. Two clinics reporting different numbers is frequently a coding difference rather than a performance difference. Write one definition and get agreement before computing anything.

Sessions cancelled in bulk look like patient behaviour. When a provider is out and thirty appointments move, that is a supply event, not a demand event, and it has to be separated or every metric is contaminated.

Slot definitions change without notice. Templates are edited continuously by people who are not thinking about your report. Any analysis has to tolerate that the grid was different last month, which means capturing template state over time rather than reading only the current one.

Providers and resources are scheduled differently. Rooms, equipment and staff may be scheduled as separate resources, and a clinician's apparent capacity can be constrained by a room that is booked elsewhere. If your analysis only looks at provider schedules, you will recommend capacity that physically does not exist.

Double-booking is normal in some settings. In several specialties an overlapping booking is a deliberate and appropriate pattern rather than an error. Do not report it as a defect without asking; you will lose the room.

Run the review with the clinician, not about them

The template belongs to the person working it. Every durable template change we have seen came from a conversation with a clinician holding their own data, and every failed one came from a standard applied uniformly from above.

A quarterly template review that works has a short, fixed shape. Their measured visit duration distribution by type. Their fill rate and their third next available. Their holds, with how often each was used for its purpose. The gap between sessions planned and sessions delivered. And one or two specific proposals, framed as questions.

What that does is move the conversation from impression to evidence without turning it into a scorecard. It also surfaces the legitimate reasons a template looks the way it does, which are frequently invisible from the data alone: a teaching obligation, a complex subpopulation, a room shared with another service. Those reasons are real, and a system that overrides them produces a template that gets worked around within a month.

Where you do not need us

Most of what is in this article is a spreadsheet and a week of somebody's time. If you have one site and twenty providers, an operations analyst with access to the scheduling data can produce the real-capacity number, the hold usage report and the duration distributions, and those three artifacts will drive a year of improvement without buying anything.

Software earns its place at a different scale and for a different job: many sites, continuous monitoring rather than periodic study, or a template state that changes fast enough that a quarterly manual pull is always stale. And optimization software applied to a template full of stale holds will optimize around the holds, which is worse than useless because it confers the appearance of rigor on the wrong baseline. Clean the template first. Then decide whether you need anything.

What we see go wrong

• Planning against paper capacity instead of sessions actually delivered
• Holds with no owner, no release horizon and no review date
• A uniform slot length imposed across providers with different measured durations
• Reporting the mean visit duration when the distribution is right-skewed
• Changing the template while a backlog still exists, and concluding it did not work
• Adding carve-outs, fragmenting the slot pool into small badly behaved queues
• Ignoring room and equipment constraints and recommending capacity that cannot exist
• Optimizing over a template nobody cleaned first

What good looks like

• Real capacity computed per provider per quarter, from sessions actually delivered
• Third next available sampled on a fixed cadence, by provider, type and location
• Fill rate and minute-based utilization reported together
• Every hold with an owner, a purpose, a release horizon and a review date
• Held-slot usage measured and reviewed quarterly
• Visit duration distributions per provider per type, medians and tails
• Demand estimated from panel size and visits per patient per year
• Supply shape plotted against demand shape across the week
• Template state captured over time, not only its current version
• A quarterly review held with each clinician, using their own data

Bottom line

The gap between the template and reality is made of ordinary things: holds that outlived their reason, sessions that were never going to happen, and slot lengths that no longer match the visits being delivered. All three are measurable with data the practice already has, and measuring them changes the conversation from an argument about impressions into a set of specific, small decisions that clinicians will actually agree to. Do that first. Work the backlog down as a project with an end date. Then, if you still have a problem, it is a genuine supply problem and the answer is staffing, scope or a different care model — and you will have the numbers to make that case.

Frequently asked questions