Every missed call has a price, even when nobody writes it down. Working out your AI call answering ROI starts with one question: what is a missed call worth? A call that goes to voicemail after hours, or rings out during a busy lunch, may be a patient, a customer or a sale that goes to someone else. This article gives you a simple model to fill in with your own figures.
A simple way to estimate AI call answering ROI is to count the calls you miss, multiply by the share that would have become paying work and by the value of each job, add the staff time saved, and subtract the monthly cost of the service. Use your own numbers and treat the result as a range, not a promise.
The formula
The model has two parts: the value that comes back, and the cost that goes out.
Value = (missed calls × conversion rate × value per job) + (staff hours saved × cost per hour)
Return = value − monthly cost of the service
Each term means something specific:
- Missed calls. The calls you did not answer in a month, including calls that went to voicemail and were not returned in useful time.
- Conversion rate. The share of those callers who would have become a booking, a sale or a paid job if someone had answered.
- Value per job. The average amount a booking or sale is worth to the business, after the direct cost of delivering it.
- Staff hours saved. The time your team spends on calls the agent takes over, such as taking bookings, answering routine questions or chasing callbacks.
- Cost per hour. What an hour of that staff time costs, including overheads.
- Monthly cost. The regular cost of the service, plus its monthly share of any setup cost. The cost guide explains how to check a quote.
If you want to see the job the service actually does, the AI receptionist use case describes it in practice. The formula only works when you know what the job is worth.
Where to find each input
Most of the inputs already exist in your own systems. The table shows where to look and the mistake people most often make with each one.
| Input | Where to find it | Common mistake |
|---|---|---|
| Missed calls | Your phone system's call log, including calls that rang out or went to voicemail | Counting only the calls someone noticed as missed |
| Conversion rate | Your customer records: calls that became a booking or sale, divided by calls answered | Mixing in website enquiries as if they were calls |
| Value per job | Average invoice or contract value for the same job type, less direct costs | Using revenue when margin is what counts |
| Staff hours saved | Timesheets, or a week of timing how long routine calls take | Assuming the agent saves all the time on a call |
| Cost per hour | Payroll cost for the team, divided by paid hours, plus overheads | Leaving out overheads such as premises and phones |
| Monthly cost | The quote, with every line shown, turned into a monthly figure | Comparing a per-minute rate with a monthly licence without a total |
Pull three months of data for each input where you can. One quiet month can make the whole model look wrong, and one busy month can make it look too good.
A worked example (illustrative)
This example uses invented round numbers to show the arithmetic. It is not a result from any real business or vendor, and the money is in an unnamed currency.
Imagine a dental practice that misses 40 calls a month, mostly after hours and at lunchtime. Suppose one in four of those callers would have booked, so 10 bookings are lost each month. Each booking is worth 100 units to the practice. The team also spends about 10 hours a month on routine calls that an agent could take, and each hour costs 20 units.
- Value from missed calls: 40 × 0.25 × 100 = 1,000 units
- Value of staff time saved: 10 × 20 = 200 units
- Total value: 1,200 units
- Monthly cost of the service: 500 units
- Monthly return: 1,200 − 500 = 700 units
On these numbers, the value is more than twice the monthly cost. Change one input and the answer moves. If only one in ten missed callers would have booked, the value from missed calls falls to 400 units, the total to 600 units and the return to 100 units. That is why the sanity check below matters.
The table shows the same example under three assumptions, so you can see how much the answer depends on the conversion rate.
| Scenario | Conversion rate | Total value | Monthly return |
|---|---|---|---|
| Cautious | One in ten | 600 units | 100 units |
| Expected | One in four | 1,200 units | 700 units |
| Optimistic | Three in ten | 1,400 units | 900 units |
In this example, the conversion rate moves the answer far more than the staff time does, so that input deserves the most care.
The benefits that do not fit the formula
Some benefits are real but hard to put a number on. Leave them out of the figure and list them separately:
- Fewer after-hours enquiries going cold, because the first ring is answered.
- Better records, so the next conversation starts from the last one rather than a blank file.
- Fewer repeat calls on the same matter.
- Less interruption for staff, which helps them focus on the calls that need a person.
- Clearer information about what callers ask, which can shape your services.
How to present the number to your team
Lead with the range, not the single figure. Show the cautious, expected and optimistic cases side by side, and say which inputs are measured and which are estimated. Name the source for each input, so a finance reviewer can trace it back to a report.
Agree who updates the model, and how often. A monthly update is usually enough once a pilot has produced real figures. Keep the sanity-check list with the model, so the next person to read it asks the same questions you did.
A sanity check before you trust the number
A model is only as good as the questions you ask of it. Run these checks before the number goes into a board paper:
- Halve the conversion rate. Does the return still stay positive?
- Are the missed calls really lost to you, or are they picked up by someone else or called back later?
- Are the bookings completed and paid, not just made?
- Does the monthly cost include numbers, carrier charges, setup and any overage?
- Is the staff time genuinely saved, or will the team spend it checking the agent's work?
If the return stays positive after these checks, the case is sound. If it turns negative with one small change, test before you commit.
Turn the model into a test
The best check is a measured one. Record the missed calls and bookings for a month before the service starts, run it on one call type for a few weeks, and compare the two periods. The 30-day pilot guide sets out how to run that test with clear criteria, so the answer comes from your own calls rather than from the model alone.
How VoiFlow handles this
VoiFlow picks up on the first ring, 24/7, and books through connected calendars during the same call. Every call is logged, so the inputs to the model, such as missed calls, bookings and handoffs, come from the same record rather than from estimates. Plans are monthly licences with included minutes and messages, and the current options are on the pricing page.
Frequently asked questions
What if I do not know my conversion rate?
Estimate it from the last three months of calls you did answer, then check the estimate with your team. A rough figure checked against real bookings is better than no figure. Update it as the data improves.
Should voicemail count as a missed call?
Count a voicemail as missed if nobody returned it in useful time, for example before the caller had booked elsewhere. Be consistent, and say which rule you used when you share the number.
Is ROI the right measure for a small business?
It is a good start, because it forces you to write the inputs down. For a small business, track the things that matter day to day as well, such as whether the phone is answered at lunchtime or after closing, and keep them alongside the return.
How often should I update the model?
Once a month is usually enough, and always after a change in price, opening hours or the service itself. The first update after a pilot matters most, because it replaces estimates with your own call data.
Fill in the model with your own numbers, run the sanity checks and then test the answer. When the numbers are ready, contact our team to talk through your call volumes, or try a live agent to hear how a call is answered at the first ring.
