Use the live example, understand the maths, copy the prompt, and adapt the design for your own product. Everything on this page is free to take. The calculator below runs entirely in your browser, nothing you type is sent anywhere, and the formulas are written out in full further down so you can check the numbers rather than trust them.
Try it first
Year one result
Return on investment
88%
Stress-test the hours-saved assumption
Only the hours saved per person per week changes here. Every other input you entered is held constant.
50% of entered hours
0.5 hours per week
ROI
-6%
Payback
15.8 months
75% of entered hours
0.75 hours per week
ROI
41%
Payback
4.7 months
Entered assumption
1 hour per week
ROI
88%
Payback
2.7 months
Total annual benefit
£36,400
Year one investment
£19,400
Net year one benefit
£17,000
Payback
2.7 months
Where these numbers came from
Annual time saving value: 20 people x 1 hours x 52 weeks x £35 = £36,400
Total annual benefit: £36,400 plus other annual benefit of £0 = £36,400
Year one investment: £5,000 implementation plus twelve months at £1,200 = £19,400
Monthly net benefit: £3,033 gross per month minus £1,200 recurring = £1,833
Payback: £5,000 implementation divided by £1,833 monthly net benefit = 2.7 months.
One year horizon. This is a model of expected value, not a record of savings achieved.
This model assumes the benefit starts immediately. If implementation delays benefit realisation, reduce the first-year benefit or model the ramp separately.
ROI counts all twelve months of recurring cost inside year one investment. Payback deducts the recurring cost from the monthly benefit as it occurs, so only the upfront implementation cost sits in the numerator. Counting the recurring cost there as well would double-count it.
What this calculator actually tells you
This is a year one investment case. It converts two things you can quantify, time saved and other annual benefit, into money, then sets that against what the first year costs you. That is all it does, and knowing the boundary is what makes it useful.
It is a model, not evidence of realised savings. Nothing here proves that the hours will actually be released, that released hours turn into value, or that the recurring cost stays flat. If you present the output to somebody who controls a budget, present the assumptions in the same breath, because the assumptions are the argument. The number is just arithmetic on top of them.
It is also deliberately a single year. Most software business cases look better over three years, and a three year view is a legitimate thing to model, but it is a different model with discounting and renewal assumptions in it. Do not quietly stretch a one year calculator over three years to make the result look stronger.
The maths, without the black box
Annual time saving value is people multiplied by hours saved per person per week, multiplied by 52 weeks, multiplied by the fully loaded hourly cost.
Total annual benefit is the annual time saving value plus any other annual benefit you entered.
Year one investment is the one off implementation cost plus twelve months of the recurring monthly cost.
Net year one benefit is total annual benefit minus year one investment.
ROI percent is net year one benefit divided by year one investment, multiplied by 100. When year one investment is zero there is nothing to divide by, so the calculator says the figure is not defined rather than inventing one.
Payback uses a cash flow definition rather than the shortcut you often see. Monthly gross benefit is total annual benefit divided by twelve. Monthly net benefit is that figure minus the recurring monthly cost, because the recurring cost keeps consuming the benefit every month. Payback months is the implementation cost divided by monthly net benefit. If monthly net benefit is zero or negative the implementation cost is never recovered, and the calculator says payback is not reached instead of printing a large but meaningless number.
The two figures appear to treat the recurring cost differently, and they should. ROI includes all twelve months of recurring cost inside year one investment. Payback deducts the recurring cost from the monthly benefit as it occurs, so the numerator holds only the upfront implementation cost. Putting the recurring cost in that numerator as well would count it twice.
The model also assumes the entered benefit starts immediately. If implementation delays benefit realisation, reduce the first-year benefit or model the ramp separately rather than reading the headline figure as though the saving began on day one.
Worked example using the sample values, which are editable examples rather than a Sentway claim. Twenty people saving one hour a week at 35 pounds an hour is 20 times 1 times 52 times 35, which is 36,400 pounds of annual time saving value. With no other benefit, total annual benefit is 36,400. Year one investment is 5,000 plus twelve months at 1,200, which is 19,400. Net year one benefit is 17,000 and ROI is roughly 88 percent. Monthly gross benefit is about 3,033 pounds, monthly net benefit after the 1,200 recurring cost is about 1,833, so the 5,000 implementation cost pays back in roughly 2.7 months.
Read that example again and notice how much of it rests on one input. Halving the assumption to half an hour a week takes the annual benefit to 18,200, which is below the 19,400 year one investment, so year one ROI turns negative and payback stretches past the twelve month horizon. Holding the full hour leaves ROI at roughly 88 percent and payback under three months. That is the distance between a yes and a no, and it is one estimate. The stress-test strip under the headline result runs that comparison against whatever you have typed in. Do not include benefits you could not defend in a meeting.
Design anatomy, and why the interface is shaped this way
Inputs and results stay on screen together. A calculator that scrolls the answer out of view when you change an input breaks the only interaction that matters, which is adjusting a number and immediately seeing what it does.
The estimated values are plain numeric inputs, not sliders. Sliders look interactive and read well in a screenshot, but they make it hard to type an exact figure, they are awkward with a keyboard, and they imply a range that you have quietly invented. A slider is the right control when the range is genuinely meaningful and bounded. Cost per hour is not that.
The primary result is large, but the assumptions are not hidden behind it. A big number with no visible workings is the design pattern of a page that wants a signature rather than a decision. Showing the breakdown costs a little visual drama and buys the thing that actually converts, which is that a sceptical reader can check you.
Labels use the language of the business, not of the model. People affected, hours saved per person per week, fully loaded hourly cost. Nobody has to guess what a variable means, and the phrase fully loaded is doing real work because salary alone understates the cost of an hour.
The breakdown appears before any lead capture. If you gate the methodology behind an email form, the reader has to trust a number they cannot inspect in order to give you their address, which is exactly backwards. Ask for the email after you have been useful, if at all.
The layout is responsive because calculators travel. They get pasted into messages and opened on a phone by somebody in a meeting. A layout that overflows sideways on a narrow screen fails at the moment it is being used most.
How AI created it, step by step
The quality of an AI built calculator is decided almost entirely before any styling happens. This sequence is what separates a page that computes correctly from one that looks right and is subtly wrong.
One, define the business variables. Name every input, state its unit, and say who supplies it. Vague inputs produce vague models. Writing pounds per hour, fully loaded rather than cost removes an entire class of misunderstanding.
Two, define the formulas explicitly. Give the AI the arithmetic rather than asking it to work out how ROI should be calculated. If you leave the definition open you will get a plausible one, and plausible is not the same as the one your finance team recognises. Payback in particular has several defensible definitions and you must choose.
Three, define the error and edge cases before the interface exists. Zero investment, empty fields, text pasted into a number field, negative entries, very large numbers. State what should be shown in each case. This is where most generated calculators fall over, because the happy path is the only path anybody specified.
Four, define the interface and the hierarchy. Say which result is primary, what appears in the breakdown, and that inputs and results must be visible together. Specify semantic labels and keyboard usability as requirements, not as polish.
Five, test with known values. Give the AI a fixture: a set of inputs and the answers you have already calculated by hand, and ask it to confirm the page reproduces them. Then give it a deliberately broken set, such as a zero investment, and check the message rather than the number.
Six, only then style and publish. Visual work done before the model is correct usually has to be redone, and a good looking wrong calculator is more dangerous than an ugly right one.
The master prompt
Paste this into whichever AI you build with. It does not mention Sentway, so it is useful whether or not you ever publish through us.
Master prompt: build the calculator
Build a production quality, responsive ROI calculator as a single self contained web page. Use semantic HTML with a real <label> for every input, no external dependencies, and no frameworks unless you need one.
Inputs, all numeric, all in GBP where money applies: one-off implementation cost, monthly software or service cost, number of people affected, hours saved per person per week, fully loaded hourly cost, and an optional other annual benefit. Use editable example values rather than claims, for instance twenty people saving one hour a week at 35 pounds an hour.
Use these formulas exactly:
- annual time saving value = people * hours saved per week * 52 * hourly cost
- total annual benefit = annual time saving value + other annual benefit
- year one investment = implementation cost + (monthly cost * 12)
- net year one benefit = total annual benefit - year one investment
- ROI percent = (net year one benefit / year one investment) * 100, only when year one investment is greater than zero, otherwise display that ROI is not defined
- monthly gross benefit = total annual benefit / 12
- monthly net benefit = monthly gross benefit - monthly cost
- payback months = implementation cost / monthly net benefit, only when monthly net benefit is greater than zero, otherwise display that payback is not reached under these assumptions
Input handling: negative numeric inputs are invalid. Prevent them where practical, for example with a min of zero on every field, and otherwise normalise them to zero. Never use a negative input in any calculation. Treat blank or unparseable entries as zero. Never display NaN, Infinity or a negative payback.
Negative ROI is a valid result and must be displayed as a negative percentage. Do not clamp it to zero and do not hide it.
Formatting: money as GBP with no decimals, ROI as a whole number percentage, payback to one decimal place.
Assumptions to state on the page: this model assumes the entered benefit starts immediately, so if implementation delays benefit realisation the reader should reduce the first-year benefit or model the ramp separately. Also explain that ROI includes all twelve months of recurring cost inside year one investment, while payback deducts the recurring cost from the monthly benefit as it occurs, so only the upfront implementation cost sits in the payback numerator; including the recurring cost there as well would double-count it.
Layout: keep inputs and results visible together on desktop and stack cleanly on mobile with no horizontal scrolling. Make ROI the primary result, then total annual benefit, net year one benefit, year one investment and payback. Below the results show a plain English breakdown of how each figure was derived, and keep the assumptions visible on the page rather than hidden behind a control.
Accessibility: every input keyboard reachable and labelled, visible focus states, and colour contrast that passes WCAG AA. Do not put aria-live on the whole recalculating results container, because it announces every keystroke. Instead add one small dedicated aria-live="polite" summary region that holds a short sentence with ROI, payback and net year one benefit, and update it only after input changes settle, around 300 to 500 milliseconds. The visible results update immediately.
Do not add testimonials, logos, ratings, invented statistics or a lead capture form. Use plainly editable CSS with the colours and fonts declared as variables at the top so I can restyle it without touching the markup.
Produces: A single page ROI calculator that computes correctly, degrades gracefully on bad input, and can be restyled without rewriting the logic.
Publish it with Sentway
Once the page is right, this second prompt puts it at a public URL through the Sentway MCP connection.
Free plan
Publish the ROI calculator you just built as a Sentway site and give me the public URL. Keep the page exactly as it is. Do not put an access gate in front of it unless I ask for one.
Produces: A live public URL serving the calculator page, ready to share.
Add a gate later only if you actually want to know who is using it. A calculator that is open is shared more.
Compares what a team spends today against what it would spend on your product, rather than valuing time. Replace people, hours and hourly cost with current annual spend and proposed annual spend. Annual benefit becomes current spend minus proposed spend, and the rest of the model is unchanged. This version is easier to defend because both inputs are invoice figures rather than estimates.
Change the calculator so the benefit comes from cost replacement rather than time. Inputs: current annual spend on existing tools, proposed annual spend, one-off migration cost. Annual benefit = current annual spend - proposed annual spend. Year one investment = migration cost + proposed annual spend. Keep every other formula, edge case and layout rule the same, and label clearly that this compares spend, not time.
Sales business case calculator
Built for a buyer to take into an internal approval meeting rather than for a browser. Adds a current cost versus proposed cost pair, a confidence factor applied to the benefit, and a printable summary. The confidence factor matters: it lets a cautious buyer discount your figures themselves instead of dismissing the whole page.
Extend the calculator into a business case tool. Add a current annual cost input, a proposed annual cost input, and a confidence percentage between 10 and 100 that multiplies the total annual benefit before any other figure is derived. Show the unadjusted and adjusted benefit side by side so the discount is visible. Add a print friendly summary view that lists every input, every formula and the final figures on one page.
Service or project ROI calculator
For agencies and consultancies where the benefit is revenue uplift rather than saved hours, and the costs are a project fee plus an optional retainer. Swap the time inputs for baseline monthly revenue and an expected uplift percentage, and keep the recurring cost so the payback definition still holds.
Rework the calculator for a service engagement. Inputs: project fee, optional monthly retainer, current monthly revenue, expected uplift percentage, and expected months until the uplift is fully realised. Monthly gross benefit = current monthly revenue * uplift percentage / 100, applied only after the ramp months have passed. Year one investment = project fee + (retainer * 12). Keep the same payback definition and the same edge case handling, and show the ramp explicitly in the breakdown.
Before you put this in front of customers
Use assumptions you can defend out loud. If you cannot say where the hours saved figure came from, do not ship it as a default.
State the time horizon in the interface. This model is one year. Say so, rather than letting a reader assume three.
Explain what fully loaded hourly cost means, and say whether your default includes employer costs and overhead.
Decide honestly whether the other annual benefit field is credible for your case. If it becomes a place to park a number that makes the total work, remove the field.
Test zero in every field, a very large number in every field, and text pasted into a number field. Nothing should ever render NaN or Infinity.
Open it on a phone at a narrow width and check nothing scrolls sideways.
Do not hide the methodology behind an email form. Gate a downloadable summary if you must, never the arithmetic.
Check the claims around the calculator match what the calculator computes. A page promising proven savings above a page computing a model is the part that damages trust.
When an ROI calculator is the wrong format
When the benefits are mainly qualitative. Better decisions, less frustration and lower risk are real reasons to buy things and they do not survive being forced into pounds. A calculator that monetises them produces a number that any careful reader will discount to zero, and the rest of your case goes with it.
When the assumptions are too speculative. If three of the six inputs are guesses, the output is a guess with extra steps and a decimal point. Precision that the underlying data cannot support reads as manufactured, and it is worse than saying you do not know yet.
When the buyer's situation varies so much that any default is misleading. If your honest answer to every input is it depends, a calculator will make you look either naive or evasive.
In those cases a short case study with real figures from one named situation, a straight comparison of options, or a scenario walkthrough with a low, likely and high case will usually be more credible than a calculator. All three admit uncertainty instead of hiding it.
Where Sentway fits, and when you do not need it
Your AI built the page. That part happens in whatever tool you already talk to, and nothing on this page depends on us.
Sentway is useful for what comes after: publishing that page at its own URL without setting up hosting, putting an access rule in front of it, capturing what visitors do and what they submit, and selling access if the calculator is the product rather than the advert. Because the connection runs over MCP, your AI can read the responses back in the same conversation where it built the page.
You do not need us when the calculator simply belongs on a website you already run. In that case take the master prompt above, drop the output into your existing site, and you are done. That is a genuinely good outcome and we would rather say so than pretend otherwise.
Questions people ask
Is this ROI calculator free to use?
Yes. It runs in your browser, it does not send what you type anywhere, and there is no sign up to see the results or the formulas.
How is payback period calculated here?
Implementation cost divided by monthly net benefit, where monthly net benefit is the annual benefit divided by twelve minus the recurring monthly cost. If monthly net benefit is zero or negative, payback is never reached and the calculator says so rather than showing a number.
Can I use this ROI formula in my own calculator?
Yes. The formulas are written out in plain English above and repeated inside the master prompt so you can hand them to an AI directly.
Why does it show ROI as not defined sometimes?
ROI divides by year one investment. When implementation cost and monthly cost are both zero there is nothing to divide by, so the honest answer is that the figure does not exist rather than a very large number.
Do I need Sentway to build an ROI calculator?
No. The master prompt produces a self contained page that works anywhere. Sentway is for publishing it as a URL, gating it, capturing responses or selling access.