ROI Calculator
The cost of answering late.
Every enquiry goes to whoever replies first. Put your numbers in and see the revenue and hours a workforce that answers straight away gives back.
Three numbers to have ready
- Enquiries you take a month
- Average value of a won client
- How fast you reply today
Two minutes, no email needed. The result stays on screen.
The ROI calculator shows what unanswered enquiries cost you over a year, and what an Autonomous Digital Branch recovers. Enter your enquiry volume, their value, and your current response time; it estimates the leads lost to slow replies and after-hours gaps that an AI workforce answering at any hour would capture and book instead.
Every figure here is an illustrative estimate, not a guarantee. We use conservative caps on purpose, so the number is one you could defend to a finance director. Your real result depends on your market, your offer, and how well the front of your business already runs.
The number
What a fast first reply is worth.
Most enquiries go to whoever answers first. Put four honest numbers in and we will show the revenue and the hours you leave on the table by replying slowly.
Start with your volume.
Round figures are fine. We only need the shape of your funnel.
Calls, forms, and messages combined.
A single job, or a year of a retainer.
Worked example
The model, worked through once.
The same maths the calculator runs, on its starting values, line by line. This is an illustrative example of the model, not a quote or a promise; put your own numbers in above and the figures change with them.
Would-be wins lost, by reply speed
- Within a few minutes
- 10%
- Within an hour
- 30%
- Same day
- 55%
- Next day or later
- 80%
Every other figure this model uses, the caps included, is set out with its source in the assumptions ledger below.
| What goes in | |
|---|---|
| Enquiries a month | 60 |
| Average value of a won client | £2,000 |
| Reply speed today | Within an hour |
| Enquiries currently won | 20% |
| What the model does | |
| Share of would-be wins lost at that speed | 30% |
| Share of that loss recovered, at most | 45% |
| Conversion after answering instantly | 31%, never allowed past 95% |
| What comes out | |
| Extra clients a year | 78 |
| Extra revenue a year | £155,520 |
| Hours handed back a year | 144 |
| Setup, plus a year of the monthly fee | from £11,400 |
| Time to cover that cost, at UK pricing | 1 month |
Hours handed back counts 12 minutes per enquiry to answer and qualify by hand, across 720 enquiries a year. Time to cover divides the one-off setup by the monthly recovered revenue less the monthly fee, rounded up to a whole month. It is worked here at UK pricing; the calculator above runs it against the price for your own region. The client value and the revenue recovered are your money, so the currency switcher in the calculator above changes their symbol and leaves the amount alone. Setup and the year of fees are our prices, so they move with the price set for your region, and each is a from figure, illustrative until the work is scoped.
Payback counts recovered revenue before what it costs you to deliver the work you win, so a thinner margin stretches it. It runs from the month the workforce goes live, which leaves the build outside the count. Every figure here is illustrative and your own result will differ.
Assumptions
Every number behind the figure.
A figure you cannot check is a figure you should not trust. Everything the calculator runs on is listed here, read from the same constants the maths uses, so this ledger and the model cannot come apart. Where a number is our judgement rather than a measurement, it says so in those words.
The loss factors, and the evidence beside them
These four are house estimates, and this paragraph used to defend them differently: it measured each one against a decay curve of conversion by response interval, published on our briefing on the five minute cliff. In August 2026 we went back and read the sources behind that curve, and it did not survive: the conversion column traced to a single vendor's unauditable marketing page, so we removed it from the briefing and from this ledger. What the vetted research does license is the direction. The odds of qualifying a lead fall steeply with delay, twenty-one-fold between minute five and minute thirty (Oldroyd, MIT and InsideSales, 2007), and sixty-fold at a day (Harvard Business Review, 2011). The factors below are our judgement of magnitude, held by the build to the shape that evidence supports: each band treats more business as lost than the one before it, and none reaches certainty.
| Reply speed today | We treat as lost | The vetted finding it leans on |
|---|---|---|
| Within a few minutes | 10% | the odds of qualifying a lead have already fallen four-fold between minute five and minute ten, per Oldroyd's MIT and InsideSales lead-response study |
| Within an hour | 30% | a firm that makes contact within the hour is nearly seven times as likely to qualify the lead as one an hour slower, per the Harvard Business Review audit |
| Same day | 55% | by thirty minutes the qualification odds have fallen twenty-one-fold from their five-minute peak, per Oldroyd's MIT and InsideSales lead-response study |
| Next day or later | 80% | a firm replying within the hour is more than sixty times as likely to qualify the lead as one that waits a day, per the Harvard Business Review audit |
The two findings above are the ones that survived reading at source: Oldroyd, MIT and InsideSales (2007) and the Harvard Business Review audit (2011). They measure qualification odds rather than deal conversion, which is why this ledger no longer converts them into implied loss percentages: that arithmetic would be ours, and a derived figure presented as published evidence is the exact habit this page exists to avoid. The magnitudes are ours and say so; the direction is theirs.
The rest of the model
| Assumption | Figure | Where it comes from |
|---|---|---|
| Manual handling per enquiry | 12 minutes | A house estimate of the time one person spends reading an enquiry, answering it, qualifying it and logging it. It drives the hours figure only and never touches the revenue figure. |
| Share of the lost business we model as recoverable | 45% | A house cap. Nobody measured it and no client produced it. It is our judgement of how much of the revenue a slow reply loses a fast reply can win back, held low on purpose, and it is the single assumption the headline figure is most sensitive to. |
| Conversion ceiling | 95% | A backstop on the modelled win rate for any caller of this model. Inside this calculator it never binds, because the win rate needed to reach it sits above the highest the slider offers, so it is honest to call it a guard for future callers rather than one of the caps doing work today. |
| The horizon the payback figure covers | 12 months | The model runs a year, so a year is the longest payback it can state. Anything slower is reported as slower and no month is named. |
| Most enquiries a month the model accepts | 5,000 | A bound on the size of business this model can describe from a form. Above it the calculator stops estimating and asks for a conversation. |
| Most a won client can be worth | £1,000,000 | The same bound, applied to client value. It caps what goes in. What comes out is capped separately by the row below, because two sensible ceilings multiplied together still reach a figure we would decline to publish. |
| Largest yearly recovery this calculator will state | £25,000,000 | A house judgement about the size of business we serve. The model computes the same figure it always did; this decides whether we are willing to publish it. Past this point the model is describing a firm larger than the ones we build for, so the result stops naming a number and asks for a conversation instead. |
| Lowest win rate the slider offers | 1% | The model recovers a share of business you already prove you can win, so it needs a win rate above zero to have anything to work from. |
| Highest win rate the slider offers | 90% | A firm converting above this is not losing enquiries to a slow reply, so the model has nothing useful to tell it. |
| Setup, one off | from £3,000 | The published from figure for a managed AI employee, and the numerator of the payback sum. Illustrative until the role is scoped and agreed in writing, and never a quote. |
| Managed AI employee, a month | from £700 | The published from figure for the monthly fee, per employee, so a larger workforce moves it. It comes off the recovered revenue before the setup is divided by what is left. |
The conversion ceiling is the one cap doing no work on this page, and it is worth saying rather than letting it take credit. A win rate would have to pass 92.2% before the ceiling could bite, and the slider stops at 90%. Nothing on this site hands the model a figure that high, so it would guard a caller we have not written yet.
Why the number is rarely small
A slow reply hands a lead you paid for to a rival.
The lead you lose to a faster competitor cost the same to generate as the one you win. You paid for the advert, the referral, the reputation. The only difference was who answered first.
Each figure here names the study it comes from and the year it was published. They are published benchmarks, not results we are attributing to a single client.
55%
of firms reply in over five days, or never
In an audit of 433 B2B SaaS companies, 55% took more than five working days to reply to a web enquiry, or never replied at all (Drift Lead Response Report, 2017). By then the prospect has called someone else.
21x
more likely to qualify a lead in five minutes
Reply within five minutes and you are up to 21 times more likely to qualify the enquiry than a firm that waits thirty minutes (MIT and InsideSales lead-response study, 2007). After business hours, the gap is a chasm.
£48k
lost per accountant each year to admin
Full-time UK accountants lose nearly £48,000 each in billable hours every year to admin and outdated systems (Silverfin, All Accounted For, 2025). Most of it is invisible until you count it.
Reading your number
What your result is actually telling you.
Three figures come out the other side. Together, they are the gap between the business you run today and one that answers every enquiry, every hour, without another hire.
Extra revenue a year
The deals you would win if every enquiry were answered and qualified on the spot, instead of leaking to whoever replied first.
Hours handed back
The time your people currently spend answering, qualifying, and chasing by hand, freed for the work only a human should do.
The conversion you are leaving
The lift from your current win rate to what a branch that answers at any hour and never forgets a follow-up can hold.
Built to be defensible
We cap the recovery and the conversion deliberately. The number you see is one you can put in front of a finance director without flinching.
Questions
The calculator, answered.
It takes four numbers from you: monthly enquiry volume, average value of a won client, how fast you reply today, and your current win rate. It then models the share of would-be conversions lost to a slow first response, and recovers part of that loss, capped at 45 percent of it, so the result stays honest rather than flattering. Conversion is never allowed past 95 percent.
It divides the one-off setup by what the model recovers each month once the monthly fee is paid. Recovered revenue is spread evenly over 12 months, the monthly fee comes off that, and the setup divided by what remains is the payback, rounded up to a whole month. Setup runs from £3,000 and a managed AI employee from £700 a month, so the divisor is your monthly recovered revenue less £700. The count starts when the workforce goes live, so the build sits outside it, and it measures recovered revenue before what it costs you to deliver the work you win. Past 12 months the calculator says so plainly, because a year is the horizon this model covers.
It is an estimate, and we say so on the result. The figures show the shape of the opportunity, not a promise. We use conservative caps precisely so the number is one you can defend to a finance director. Your real outcome depends on your market, your offer, and how well the front of your business already runs.
Most enquiries go to whoever answers first. The MIT and InsideSales lead-response study (2007) found firms that reply within five minutes are up to 21 times more likely to qualify a lead than those that wait thirty minutes, and the odds fall sharply after the first few minutes. The figure is often credited to a 2011 Harvard Business Review article, which measured a different interval. An Autonomous Digital Branch answers and qualifies in moments, day or night, so fewer leads slip to a faster competitor.
No. The headline figure, the extra revenue and the hours saved, appears on screen as you work through the steps. You only share your details if you want us to send the breakdown and come back with a branch scoped to capture it.
Book a thirty-minute strategy call. We map your branch, the AI employees it needs, and the return it should make, then tell you exactly what we would build and what it costs. No pitch deck, no obligation.
Book a strategy call
Thirty minutes to turn your number into a plan.
Thirty minutes and you leave with the scope written down, the figure attached to it, and the date it could be live.
Book a thirty-minute call
Pick a time straight from the diary. If you would rather write first, email us and a person replies, usually the same working day.