The compounding practice: why several small improvements add up faster than you expect

By Paul Gruensfelder · Healthcare Growth Strategies

Many practices try to grow by finding one big lever. A new ad channel. A rebrand. A different scheduling vendor. The instinct is understandable, because a single dramatic fix is easier to picture than a handful of quiet ones. But the arithmetic of how a patient actually reaches your chair rewards improvement at several steps at once, because the steps multiply — and that compounding is easy to underestimate.

I want to work through that math here, using public benchmarks drawn from the common touchpoints of retail healthcare — how patients find, choose, and show up for care. It’s built as a cross-specialty planning example, not a universal benchmark — substitute your own channel, specialty, and location data where you have it. The endpoint is a clean, operational one: a kept first visit, a new patient actually in the chair. By visitors I mean unique prospective new-patient website sessions — not existing-patient traffic, duplicate visits, or spam. The claim I’m testing is simple: a practice that gets meaningfully stronger at several non-clinical steps at once can turn the same traffic into more new patients, and the multiplication is where the surprise lives.

A word on what this is, because it’s easy to oversell. It’s an illustrative sensitivity model, not a measured result. Each rate is conditional on the one before it, and the scenarios assume a practice could reach all of them at once without changing its traffic mix or running out of capacity — which real practices only ever approximate, and the strong and stretch scenarios in particular assume improvements that may not all be attainable together. Read the output as the shape of the opportunity, not a forecast for a specific office.

There’s a reason this is easy to miss, and it isn’t a lack of discipline. It’s how budgets are built. Marketing spend is legible: it gets a line item and someone asking what it returned. The handoffs after the click — the phone, the schedule, the follow-up — live in operations, where nobody reports a conversion rate to anyone. So a practice optimizes the one number it can see and quietly tolerates the ones it can’t.

The funnel multiplies, it doesn’t add

Picture the path many new patients still take: they find you, they land on your site and reach out, they get through to a human on the phone, they book, and they show up. Each step passes along only a fraction of the people who entered it.

Because each step keeps a fraction, the end result is the steps multiplied together, not added. Improve one step by 10% and you get about 10% more at the end. Improve four steps by 10% each and you don’t get 40% more — you get about 46% more, because the gains stack. Push each step further and the arithmetic gets steep quickly.

Here’s the same idea with real benchmark numbers. Start every practice with 1,000 qualified visitors a month and walk them down the phone funnel. The left column is a baseline practice, using published benchmarks where they exist and conservative planning values where they don’t; it’s an assembled illustrative scenario, not an observed average. The right column is the same practice performing meaningfully stronger at every step — the strong scenario.

Comparison of two phone funnels, each from 1,000 monthly visitors. Baseline practice: 1,000 visitors, 51 inquiries (5.1% site conversion), 36 reach a human (70% answered), 12.6 booked (35% of answered), 9.6 kept first visits (77% show rate). Meaningfully stronger everywhere: 1,000 visitors, 66 inquiries (6.6%), 56 reach a human (85%), 22.4 booked (40%), 19.1 kept first visits (85% show) — a total of about 19 versus 10, roughly twice as many.

Same 1,000 visitors. Four conversion steps, each meaningfully stronger, and the model yields about twice the kept first visits. Figures from the HGS funnel model; sources cited below.

Same demand at the top. Roughly twice the new patients at the bottom in the strong scenario — no one was added to the funnel; more of them simply carried through each handoff.

It helps to put a rough dollar sign on that, using your own numbers. Say a new patient is worth about $1,000 in first-year net production — use your own figure. In the baseline scenario, that month’s new-patient cohort represents roughly $9,600 in modeled first-year net production; in the strong scenario, about $19,000. That gap, in the model, isn’t from more advertising — it’s the same visitors carried further down the funnel. The figure is illustrative: it’s modeled net production, not collections, lifetime value, or margin, and it assumes value per patient holds and the practice has room to see them.

Walking the funnel, step by step

Getting found. Before any conversion step, someone has to see you, which increasingly means a map listing and a review profile. In BrightLocal’s 2025 consumer survey, 83% of U.S. adults said they use Google to find and read local business reviews, and only 4% said they never read reviews¹. I treat traffic as an input you set, because visibility gains vary so much by market — but the same reviews that help you get found feed the flywheel later.

The website. Once someone lands, the question is whether they take a booking-intent action. Unbounce’s 2024 Conversion Benchmark Report, built from tens of millions of conversions, puts the median healthcare-industry landing page at about a 5.1% conversion rate, and the medical-treatment subcategory at 5.3%². That’s a landing-page conversion benchmark, not a whole-site rate, and I use it as the anchor for a visitor taking a booking-intent action — placing a phone call or completing an online booking. The next two steps follow the calls; online bookings run their own path, covered below. (Web-form leads would need their own contact path; to keep the model clean I limit it to calls and online bookings.)

The phone. Two steps live here — whether the call is answered, and whether an answered call becomes a booked visit. The best evidence is from call-tracking vendors. Patient Prism, which reports analyzing more than 2.3 million patient calls across 127 practices³, states a typical industry missed-call rate above 20% against a sub-10% target. For booking, Patient10x reports that answered calls convert to a scheduled appointment at roughly 25–40%, depending on specialty and urgency⁴. These are vendor figures, so I treat the two rates as planning values: 70% answered, and 35% booked — a planning value within that cited range. The point isn’t the exact number: a call that rings out is a patient who did everything right and still didn’t reach you, after you paid to make the phone ring.

The show. A booked visit isn’t a kept one. A systematic review of 105 studies across specialties and settings found an average no-show rate of about 23%⁵ — the 77% show rate in the baseline scenario (it spans many specialties and countries, so read it as a general anchor, not a US-specific figure). That review is from 2018, so it’s fair to ask whether it still holds; MGMA’s 2025 polling found a majority of practices reporting no-shows steady or improving year over year⁶ — though a one-year directional poll can’t establish whether the 2018 review’s 23% average is still representative. Confirmations, reminders, and a simple deposit policy move it.

Those are the four conversion steps behind the funnels above. Two rest on published benchmarks (the website and the show), and two are planning values informed by vendor call data (the phone and the booking). Each is sourced below.

Online booking is a second path

Everything above followed the phone. But a growing share of patients would rather not call at all, and when a practice lets them book directly online, those patients skip the two steps where the phone funnel leaks most — the unanswered call and the call that never converts. That’s why online booking isn’t just a convenience; in the model, routing an action straight to a completed booking skips the phone stages, so it converts at a higher modeled rate — a property of the routing math, not a measured efficiency claim.

Adoption is still modest. In an MGMA poll, only 11% of practice leaders said a majority of their patients self-schedule, and a majority reported a quarter or fewer doing so⁷. So I model online booking as an optional second path that takes a share of booking-intent actions — a fifth, an illustrative default, not an adoption benchmark — directly into a booked visit, and lets the phone funnel handle the rest. Mechanically: at fixed traffic and a fixed conversion rate, routing a fifth of those actions straight to an online booking raises modeled kept visits from about 9.6 to 15.6, roughly 60%. That’s a routing sensitivity, not an estimate of the causal lift from installing online booking: it holds because those routed actions bypass the answer and booking stages and are credited the online show rate, and it assumes they’d otherwise have completed the phone funnel at the baseline rates. It says nothing about adoption, whether online creates additional booking-intent, which patients choose each path, or implementation quality.

Two honest caveats keep that number from being a free lunch. First, the model treats those online bookings as if they would otherwise have behaved like average phone inquiries; if online instead captures net-new demand the gain is larger, and if it mostly shifts calls you’d have booked anyway it’s smaller. The model has no separate parameter for that split, so read the number as a routing sensitivity. Second, online-booked patients don’t always show at the same rate, and the evidence cuts both ways. A US primary-care study found online self-scheduled no-shows lower than staff-scheduled, about 4.5% versus 7.4%⁸, and a private practice that let patients self-book and cancel saw the same direction; but a triage-gated hospital in a related two-center study saw online no-shows higher.⁸ Online show rates can differ by patient mix, lead time, and how booking is implemented, so the calculator leaves that rate adjustable: self-book with easy cancellation and reminders tends to help; friction and long lead times hurt.

What modest gains actually add up to

Now put a range around the phone funnel, because a single number would be false precision. I ran it three ways — a moderate case where each step improves only slightly, a stronger case, and a stretch case where each step reaches higher illustrative input values. These are scenarios, not statistical bands, and only the moderate case is genuinely modest; the strong and stretch cases require substantial improvement across every step at once and shouldn’t be read as expected outcomes. The baseline scenario seats about 9.6 kept first visits per 1,000 visitors. Here are the four rates behind each scenario, and the totals they produce.

The four rates behind each scenario. Multiply them across 1,000 visitors to reproduce the totals below, or set your own in the calculator.

*Modeled net-production change at the assumed value per new patient, holding value and capacity constant — not a forecast or a probability-weighted expectation. Phone funnel only; traffic held flat; online booking modeled separately above.

Even the moderate scenario, where no single step does anything remarkable, lands near one and a half times the new patients of the baseline — the lower of the three modeled cases. The strong scenario, around 2×, assumes every step improves together, which is a lot to ask at once. The stretch scenario is valid arithmetic, but it requires higher values at every modeled stage at once and shouldn’t be read as a likely outcome. None of the three carries a probability, so none is the “expected” outcome.

I held traffic flat to isolate the funnel; improving visibility stacks on top. And I left three things out entirely — retention, referrals, and the review loop. They’re unmodeled mechanisms that could push the result up, leave it unchanged, or overlap with rates already in the funnel, so I didn’t try to price them. They’re the next section.

The funnel is really a flywheel

Everything so far treated the journey as a line that ends when someone shows up. It doesn’t. What happens after the visit quietly determines how expensive next month’s top of funnel will be.

Start with keeping the patients you already earned. The best-known work on retention, a long-standing cross-industry finding from Bain’s Fred Reichheld, holds that increasing customer retention by 5% raises profits by 25% to 95%⁹. It isn’t a healthcare-specific number, so treat it as a direction, not a dial — but the direction is sound: the patient who comes back on schedule is the cheapest growth there is.

Then the two engines that refill the top of the funnel. Referrals first: Nielsen’s Trust in Advertising work found that 88% of global respondents say they trust recommendations from people they know more than any other channel¹⁰. A referred patient arrives pre-trusted. Reviews second: they feed how you get found, and increasingly the AI-generated summaries that now sit between a patient and their search. Worth staying precise, because the ground is shifting: BrightLocal’s 2025 survey found 42% of respondents now trust reviews as much as personal recommendations, down from 79% in its 2020 survey¹¹. Reviews still matter for being found; they’ve weakened as a pure trust signal.

This is also where treatment acceptance lives — and why I left it out of the model. Whether a patient says yes to recommended care is shaped by the reputation and trust you build across the whole journey, so it tends to improve as a result of doing the rest well. But it isn’t a dial an operator sets a target on the way you can target a show rate or an answer rate, so folding it into the funnel as a controllable step would overstate what the model can honestly claim. Read it as an outcome of the flywheel, not a lever in the funnel.

Put those together and the line becomes a loop. A patient who beats their own expectations is more likely to come back, more likely to refer, and more likely to leave a review that helps the next stranger find you. None of that is in the 2×. It’s why a practice that’s genuinely good at the whole journey tends to pull away from its market over time.

What I’d take from this as an operator

Stop hunting for the one lever. The model’s whole lesson is that no single step, improved alone, does what the steps do together. A practice excellent at marketing and average at everything after the click leaves much of the multiplication unclaimed.

Measure each step, then prioritize. Because the steps multiply, a gap at any step caps everything downstream — so the best next move is the improvement with the best mix of attainable lift, cost, capacity, and downstream value, which often isn’t the step you’re already spending on. A practice pouring money into ads while many of its calls ring out is buying demand it then discards.

If you don’t offer easy online booking, the model treats it as an entire second path you’re not using — worth testing, though the model shows only the routing arithmetic, not the real-world lift.

Measure the steps you can’t currently see. Plenty of practices can quote their ad spend but can’t readily quote their answer rate, booking rate, or show rate. You can’t improve a step you don’t measure, and the unmeasured steps are exactly where a practice can quietly lose people.

Treat the after-visit as growth, not admin. Recare, referrals, and reviews can generate demand without incremental paid-media spend. Run them as an afterthought and you pay full price for demand you could partly generate for free.

None of this requires being the best in your market at anything. But if a real funnel behaves anything like this model, the compounding across several ordinary steps is where the leverage sits.