The training ROI calculator that counts forgetting.
Most training ROI calculators multiply optimism by headcount. This one starts from the number everyone omits — how much of what you paid to teach is gone within a quarter — and shows what a retention engine recovers. Every assumption is printed and adjustable.
Why forgetting belongs in the ROI math
Training budgets are approved as if delivery were the product. But the delivered hour is only the purchase; the asset is what remains in heads when the work needs it. A century of memory research puts unreinforced loss at 50–80% within weeks — which means most training ROI models begin by ignoring the largest line item on the cost side.
The model here is deliberately simple: what you spend, what decay takes, what scheduled retrieval — the 30–40% effect, taken at its midpoint — buys back. Simple is the point: every term is inspectable, and the conclusion survives even hostile assumptions.
The calculator
Four sliders, three numbers. Defaults describe a 500-person program; drag to match yours. The model is printed below the results — change any assumption you disagree with and the argument still holds at half strength.
Model: hourly cost = salary ÷ 2,000 working hours; investment = people × hours × hourly cost; loss = investment × forgotten share; recoverable = 35% of the loss — the midpoint of the 30–40% long-term retention improvement for spaced retrieval over massed practice. Figures update as you drag; the static numbers shown are the defaults.
Where the recovered value comes from
Scheduled retrieval interrupts decay per concept, per person. The recovered slice isn’t new training — it’s the training you already paid for, kept alive instead of re-purchased annually.
What the calculator can’t know
Your incident costs, your rework rates, your audit exposure — the losses forgetting causes downstream. Add them mentally; the calculator’s number is the floor, not the ceiling.
From model to measurement
The pilot replaces these assumptions with your data: measured retention on a real cohort, against the projection without practice. The calculator makes the case; the pilot proves it.
The line item, argued in numbers
Hours trained, hours forgotten, value recovered — the model behind the calculator, live on one screen.
Interface shown as an illustration with representative numbers, not a screenshot — the layout is the product’s.
Swap assumptions for measurements.
A pilot on one cohort turns every slider on this page into a measured value from your own workforce.
Questions buyers ask
Where does the 50–80% forgetting figure come from?
From the forgetting-curve literature, replicated from Ebbinghaus onward: without retrieval, most newly learned material fades within days to weeks. The research page on the forgetting curve walks the studies and their conditions.
Isn’t 35% recovery conservative?
Deliberately. The spacing literature reports 30–40% improvements in long-term retention for scheduled retrieval over massed practice; we take the midpoint and apply it only to the value already being lost. Real programs also compound gains from targeting and skipped redundant training, which this model ignores.
Why salary ÷ 2,000 for hourly cost?
A standard full-time-year approximation. If your loaded costs run higher — they usually do — the loss figure only grows.
Can I get this as a spreadsheet for our CFO?
The model is four multiplications — it rebuilds in a spreadsheet in a minute, which is exactly what we’d encourage: the argument should survive your own modelling, not depend on ours.
What would the pilot measure, concretely?
Measured retention per concept on a live cohort over a quarter, against the no-reinforcement projection — the two-line chart that replaces this page’s assumptions with your evidence.
Run the numbers on a real cohort.
Book a demo and we’ll set up the pilot that turns this calculator’s model into your measured curve.
- 30 minutes, on your calendar — pick a slot here
- Run on your own content wherever possible, not a canned deck
- You see the dashboards, the learner surface and the evidence exports
- No commitment — and pilot data stays yours either way