Loaded cost means salary plus employment costs and tooling — usually 1.2–1.4× the headline salary, not the salary itself.
These four are where every ROI model is won or lost. The defaults are what we would use for a first pass on a typical back-office process; they are not promises, and the honest range is wide.
The straight-through rate. 70% is what we actually see on most projects — often higher, sometimes lower, and it is usually the client who decides where the line between automatic and reviewed should sit. Above 85% is rare in the first year and should be treated with suspicion in any proposal, including ours.
Exceptions do not become free. Around 50% is what we see in practice: a case that took 9 minutes takes about 4½ when it lands on a person with the work already done.
Model and infrastructure usage, licences the flow needs, and the maintenance it takes to keep working when a supplier changes a format. Around 400 a month is what our own projects actually cost to run, but it moves with volume — the same flow at ten times the cases costs more.
Move this until it matches a real quote. Ours average around 10,000, pulled down by how many small builds we do, and the full span runs from about 2,000 for a single narrow automation to six figures for a platform touching many systems. If payback only works at the bottom of the range, the project is marginal — that is useful to know now.
Hours spread thinly across many people. Twenty minutes saved for forty people is a nicer day, not a line in the P&L. Count it only if a role is not backfilled or a hire is not made.
Cycle time, error rates and rework. Often worth more than the labour saving and frequently easier to prove — but they belong in their own units, not converted into money with an invented multiplier.
Your internal project time. Realistically a few hours a week from whoever owns the process, and more during the pilot. It is not on our invoice and it is still a cost.
Volume growth. A process whose volume is rising is worth more than this shows, because you are also buying the headcount you will not add.
On these assumptions the build pays for itself well inside a year. Worth checking the straight-through assumption hardest — it is the one carrying the result.
Your numbers travel with the request. The mini-audit either confirms them or tells you where the model is wrong.
Why the baseline has to be measured before the build, and which hours actually count.
If the payback here is marginal, the process may be the wrong one.
What goes on the cost side: free mini-audit, fixed audit, fixed build.
{{ rateNote }}
{{ s2Sub }}
{{ autoNote }}
{{ excNote }}
{{ runNote }}
{{ buildNote }}
{{ hn }}
{{ verdictBody }}
{{ ctaNote }}
We use cookies for analytics — to see which pages bring enquiries. Nothing else, and nothing before you agree. Cookie Policy