Patient Revenue Engine ROI: a simple model with a worked example

The model in one line

Here's the whole model:

Monthly ROI = (recovered revenue × contribution margin − monthly fee) ÷ monthly fee

Each piece is simple, but each one has a common mistake. Getting them right is the difference between an ROI figure you can rely on and one that only works in a sales deck.

Piece 1: Recovered revenue

Recovered revenue is the value of visits that would have been lost without the tool: refilled cancellations, answered missed calls that booked, returning overdue patients, rebooked no-shows and walkouts.

Before you buy, this number is an estimate based on your own leakage. After you go live, it should come from an itemized ledger: each visit, its date, the action that recovered it and its dollar value. Don't use a vendor's industry average in your ROI model. Use your practice's numbers, and once you have them, use the ledger.

Piece 2: Contribution margin

A recovered visit isn't pure profit, but it's close to it. Your rent, staff salaries and provider time are already paid whether the slot is filled or not. Filling the slot adds only variable costs: supplies, card processing fees, perhaps some billing costs.

That's why the right margin to use is contribution margin, the share of a visit's revenue left after those variable costs, not your overall practice profit margin. Your accountant can help you estimate it. For this example, we'll assume variable costs of 15%, so a contribution margin of 85%. That's an illustrative assumption; yours may differ.

Piece 3: The full cost

A common mistake is counting the fee alone while ignoring add-ons. If pricing grows with message volume, the cost rises in exactly the months the tool is doing the most work, which makes ROI look better in the sales estimate than on the invoice. Ask for a quote at your expected volume, and at double it.

Include everything you'll pay: the monthly fee, any setup or onboarding charges spread over the year, and any per-message or per-user fees at your expected volume. Also consider staff time. If the tool adds work, count it. If it removes work, like the front desk no longer calling through a recall list, that's a benefit you can note, even if you leave it out of the math to stay conservative.

A worked example

Take an illustrative single-location practice with an average visit value of $170, paying a flat $1,149 a month with no extra fees. In a typical month, the recovery ledger shows 30 recovered visits.

1. Recovered revenue: 30 × $170 = $5,100.
2. Contribution at 85%: $5,100 × 0.85 = $4,335.
3. Net gain: $4,335 − $1,149 = $3,186 a month.
4. Monthly ROI: $3,186 ÷ $1,149 = about 277%.
5. Annual view: $13,788 in fees against about $52,000 in contribution, or roughly $38,200 net.

Break-even and sensitivity

Break-even tells you how little has to go right. Each recovered visit in this example contributes $170 × 0.85 = $144.50. Divide the fee by that: $1,149 ÷ $144.50 ≈ 8 visits a month to cover the cost. At a lower fee of $699, break-even drops to about 5 visits a month.

Then test a range instead of a single guess:

• Low case, 12 visits: $2,040 recovered, $1,734 contribution, about $585 net. Still positive.
• Base case, 30 visits: $5,100 recovered, $4,335 contribution, about $3,186 net.
• High case, 50 visits: $8,500 recovered, $7,225 contribution, about $6,076 net.

How a guarantee changes the math

A results guarantee puts a floor under the low case. Suppose a vendor guarantees recovered revenue of at least two times its fee every month, or the next month is free. At $1,149, the floor is $2,298 in recovered revenue. At 85% contribution, that's about $1,953, or roughly $804 net even in the guaranteed-minimum month.

And if a month falls below the floor, as in the low case above, the following month's fee goes to zero, which lifts the average ROI across the two months. In other words, the guarantee turns the downside from a loss into a smaller gain.

Run this model with your own visit value, margin and leakage estimate. If it only works in the high case, wait. If it works in the low case, the decision is mostly about timing.