The ROI Calculator that shows its working
Most ROI calculators multiply hours by a rate and print a number nobody believes. This one models four benefit streams separately. It subtracts the new work automation creates, waits for adoption before counting a cent, and reports payback on hard cash alone. The result is a figure that survives your finance director.
Four benefit streams, not one blended guess
They do not carry equal weight with a CFO, so the calculator never blends them together. Three are cash. One is not, and it says so.
Revenue captured
Demand you currently miss (abandoned calls, after-hours enquiries, bids never submitted), recovered and converted. Counted at gross margin, never at turnover.
Hard cashErrors & rework avoided
Fewer mishandled transactions, and the rework hours, credit notes, SLA penalties and write-offs that follow them.
Hard cashCash cost avoided
Only where money genuinely stops leaving: a role the business has committed not to backfill, or a per-transaction third-party fee the deployment removes.
Hard cashCapacity released
Hours handed back to your team. Shown in full, but kept out of the payback figure. It only becomes money if you redeploy those people or absorb growth without hiring.
Not cash yetFive steps, about three minutes
Everything recalculates as you type. Nothing is sent anywhere unless you ask for a copy.
Pick the workflow closest to yours
Seven sector presets: automotive service bookings, insurance claims, tender and bid response, property management, manufacturing quoting, contact centre and MSP support, or a neutral general option. Each loads cautious starting assumptions drawn from comparable South African deployments.
Replace the defaults with your own numbers
Monthly volume, handling time, fully-loaded staff cost, error rate, and, where the workflow earns revenue, your conversion rate, transaction value and gross margin. Every field tells you which system of yours the real figure lives in.
Choose how much you are willing to claim
Conservative claims 65% of the modelled benefit over a six-month adoption ramp. Expected claims 85% over four months. Optimistic claims all of it. The calculator opens on Conservative, and that is the basis we quote on.
Read the answer, and what drives it
Payback in months, net benefit, ROI and NPV at your own discount rate, a cumulative cash-flow curve with the break-even point marked, and a sensitivity ranking showing which single assumption moves the result most. Test that one first.
Open the working, or take it away
Every intermediate figure and every formula is on the page. If it is useful, have the full month-by-month breakdown emailed to you along with a list of exactly which reports we would need to replace the assumptions with measurements.
Three things it does that most ROI calculators will not
A calculator built to win an argument produces a number you cannot defend in a board pack. These are the parts that usually get quietly left out.
It subtracts the new work
Automation does not only remove effort, it creates some: exceptions that need a human, and QA sampling that did not exist before. Both are modelled and both come off the saving. Leaving them out is the single most common way a business case overstates itself.
It waits for adoption
Benefit accrues from go-live along an S-curve, not from the day you sign. A twelve-week deployment with a six-month ramp will often show a negative first year and a strong second. That is the truth, and far better said upfront than discovered in month nine.
It counts your side of the cost
Your exec sponsor's hours, your process owner's time, testing and training all go on the investment line, alongside model tokens, hosting and the managed operations fee. Your finance team would add them back anyway. We would rather be the ones who put them in.
Where the real numbers come from
The calculator's defaults are a starting point. Credibility comes from replacing them, in this order.
Your own systems: the gold standard
Volumes and cycle times from your DMS, CRM, ERP, claims platform or ticketing system. Loaded cost from payroll, as cost-to-company rather than basic salary. Margin and transaction value from the general ledger. Error and rework rates from QA logs, the credit-note register and the complaint log.
Gathered during: Operational Discovery, 1 to 2 weeks. The discovery questionnaire and the calculator's input form are deliberately the same document.
Measured on your data during the Impact Sprint
Discovery gives you volumes. It cannot give you the automation rate. That comes from a time-and-motion sample of 20 to 30 live transactions for a true handling time, and from replaying around 100 of your real historical records through a working prototype to observe the actual straight-through rate, accuracy and exception rate.
Gathered during: the 5-day Impact Sprint. This turns the three most-contested assumptions from guesses into observations, which is most of what the Sprint is for.
Our own production telemetry
We instrument every system we deploy to record containment rate, handling time, exception rate, human acceptance rate and cost per transaction. Those anonymised medians become the defaults you see here, and they get sharper with every deployment we run.
Gathered during: Managed Operations, the same dashboard that proves the business case afterwards against the baseline agreed at the start.
South African public benchmarks
Used only where nothing better exists: Stats SA for wage and labour data, the BankservAfrica Take-Home Pay Index for salary movement, SARB prime for the discount rate, and sector sources such as naamsa, SAIA, CIDB and the National Treasury eTenders portal for volumes.
Used for: the calculator's opening defaults only. Every one is meant to be overwritten by something from tier 1 or 2.
What this tool cannot do. It cannot tell you your real automation rate. That has to be measured on your records, and any calculator claiming otherwise is guessing. It produces an estimate for discussion, not a forecast, a quotation or financial advice. And it will happily show you a negative return: if the volume does not justify the deployment, we would rather you found that out here, in three minutes, than nine months into a project.