The Voice Agent ROI Calculator: Now Per Industry, and Honest About Its Own Limits

Every business we talk to has the same blind spot: they know roughly how many calls they get, and they have no idea how many they lose.

Not "lost" as in the caller left a voicemail. Lost as in the line was busy, or it was 19:30, or it was a Saturday, and the person calling simply dialled the next result on the page. That call never appears in any report. It is invisible, and it is expensive.

We built a calculator to make it visible. This is the second version: Voice Agent ROI Calculator.

What changed, and why

The first version had six sliders and one industry-agnostic assumption set. It answered exactly one question — what does it cost you in staff time to handle the calls you already answer — and it answered it the same way for a driving school and a car dealership.

That was the problem. A dealership and a restaurant have almost nothing in common on the phone. A restaurant call is the booking attempt, so it converts at better than half. A dealership call is the top of a funnel that ends weeks later on a forecourt, and converts at around two percent. Feeding both through one model produces a number that is wrong for both.

So the calculator now asks which business you are first, and changes its inputs, its defaults, and its wording to match:

Car dealership — deal values come from an editable four-row price-band grid, and the weighted average falls out of it
Restaurant — transaction value is spend per guest × party size
Medical clinic, driving school, supermarket, mobile repair, and a general fallback

It also models two value streams instead of one:

Labour saved — the share of answered call minutes the agent resolves end to end, priced at your real fully loaded staff cost per minute, and capped at what your actual headcount could physically handle
Revenue recovered — the calls that currently ring out, answered around the clock, filtered down to the ones with revenue behind them, and converted at your rate

Against those two streams it puts our actual price — a monthly retainer plus a per-minute rate on the agent's own call minutes — and shows the net.

The reality check is the point

Here is the thing about ROI calculators: the failure mode is not being slightly wrong. It is producing a number the prospect knows is false, at which point everything else on the page loses credibility too.

So the calculator now computes the number you would use to catch us out, and shows it to you unprompted:

On these numbers your business converts 44.4 % of relevant calls today. You have set the calculator to 20 %.

That first figure is derived from your own inputs — your current transactions divided by the relevant calls you already answer. If the rate you have dialled in is more than double it, or if the calculator is claiming an uplift of more than 40 % on your current volume, the panel turns amber and offers you a button: use my actual rate.

A calculator that talks you down to a defensible number closes more deals than one that inflates. We would rather lose the impressive headline and keep the meeting.

Two related things it now says plainly:

When your implied conversion comes out above 100 %, that is not a bug — it means you close more business than the phone alone can explain, because walk-ins, online booking and regulars are carrying part of it.
For a supermarket, the relevant-call share sits at 25 % on purpose. Most of those calls are "are you open?" and "do you carry X?", with no revenue behind them at all. The case there is staff time, and the results panel says so instead of inventing a revenue number.

Four charts, all showing their working

The value bridge is a waterfall: labour saved, plus revenue recovered, minus our monthly price, equals net monthly gain. One picture, whole model.

The call funnel exists purely so you can audit the headline. Total calls → missed today → relevant and recovered → new transactions, with the percentage that produced each stage. When you want to know where a number came from, it is right there.

The break-even curve shows the cumulative net position across twelve months, red below the line and green above, with the break-even month marked. The one-off setup cost lands in month one, where it belongs — we are not amortising it quietly into the background.

The sensitivity chart is the one for sceptics. It sweeps conversion rate from zero to twice your setting and plots annual net benefit, with markers at your value and at the rate your business actually achieves. It shows exactly how load-bearing that single assumption is. Conceding the uncertainty is more persuasive than hiding it.

What we still do not assume

Some vendors in this space model savings by quietly reducing headcount — a "workforce optimization factor" that means fewer people on payroll next year.

We do not model it that way, because it is not what happens with our clients. The agent absorbs the repetitive, high-volume, after-hours calls. The people you already employ stop being interrupted every four minutes and start doing the work that needs a human. Your payroll line does not move in our calculator. If it moves in yours, that is your decision, not our sales pitch.

We also cap the labour saving at your real capacity. If you tell the calculator you take 5,000 calls a month with one person answering, it will not credit you with saving minutes that person could never have worked in the first place.

Containment stays at 60 % — the share of call minutes the agent resolves with no handover. The remaining 40 % still costs you people time in the model. Anyone quoting 90 %+ containment is describing a demo, not a Tuesday morning.

Sharing a configured link

The industry selection writes itself into the URL, so /voice-agent-roi?industry=restaurant opens on restaurant defaults. Useful when you forward it to a colleague, and useful for us when we prepare for a call.

For clinics specifically

Call content in a medical practice is health data. Processing it needs a data processing agreement under Art. 28 GDPR, and our agent neither records nor repeats diagnoses — it books appointments and routes anything clinical to your team. That is stated in the assumptions panel rather than left for you to ask about. German practices always ask, and pre-empting the question is the whole point.

Try it

Spend a minute with it: Voice Agent ROI Calculator.

If the number looks too good, tell us — we would genuinely rather correct an assumption than win a project on a bad estimate. And if it looks worth pursuing, book a free discovery call and we will pressure-test it against your actual call logs.