How to plan a technology investment
A technology business case fails in one of two places: the hours were never saved, or they were saved and never became money. This is a sequence for getting both right, and for accepting a no.
Start at the constraint
Technology purchases usually begin with a demonstration. The demonstration is designed to show what a tool can do, not what your process needs. Before the first meeting, name the constraint: the one step in the process that limits everything downstream. It might be a manual inspection, a changeover, a queue before a pick, or a report that takes a person two days to assemble.
If the technology does not act on that step, its payback is somebody else's payback. This is the most common and least visible error in a business case: a correct calculation for the wrong bottleneck.
The exhibit at the top of this issue is a model, but its arrangement is the argument. There is a transfer cell, a roller line, a control cabinet and a staging area, and the time that escapes is not in the machine's cycle. It is in the wait before the pick and the queue after the gate.

Parts on paper. The worksheet in this issue is the same kind of object: a short chain of arithmetic written down where it can be checked.
Illustration for InSignal Works, 2026
Price the hours honestly
Count paid hours the process no longer needs — not the hours the machine runs. A machine that runs overnight while nobody was working overnight has saved nothing in labour, even if it produced a great deal.
Use a fully loaded hourly cost: wage, benefits, payroll overhead and the supervision that goes with them. The number is larger than an hourly wage and smaller than a consultant's day rate, and it is the one a finance reviewer will check first.
Then separate two claims that are usually mixed together. Hours removed means the work no longer happens. Hours redeployed means the same people spend the hour on something else. Both can be valuable. Only one of them is automatically a cost saving.
Convert hours into cash
This is the judgement call the whole calculation turns on, so make it a visible input rather than a hidden assumption. Call it the cash-conversion share: the proportion of saved hours that becomes cash in the period, through one of three routes.
- Redeployed to revenue. The hour produces billable output that would otherwise have required another person or another shift.
- Overtime avoided. The hour was being paid at a premium and now is not.
- A hire avoided. The growth that would have required a new position does not.
If none of these applies, the honest conversion share is low. That does not make the investment worthless: capacity, quality and lead time have value. It makes the payback longer, and it makes the argument honest.
A saved hour is capacity. A converted hour is cash. The distance between them is the business case.
The arithmetic, including no
The second line only applies when monthly net savings are positive. Dividing a setup cost by zero or a negative number does not produce a payback period; it produces a spreadsheet error with a confident formatting.
Recurring cost belongs in the same line as the savings, not in a footnote. Licences, service contracts, consumables, hosting, support and the internal time to run the thing all continue after the purchase order clears. A project that saves $756 a month and costs $250 a month to operate has a payback based on $506, not $756.
Using the worksheet's illustrative defaults — 30 hours a month, $42 an hour, 60 per cent conversion, $250 recurring, $12,000 setup — monthly net savings are $506 and simple payback is about 23.7 months. Those numbers are an example, not a benchmark, and they are the ones the worksheet starts from so that the arithmetic can be checked by hand.
Now change only the conversion share. At 25 per cent, monthly net savings fall to $65 and payback stretches past fifteen years. At zero per cent, savings never cover the recurring cost, and there is no payback at any horizon. The machine did not change. The assumption did.
What simple payback cannot see
Simple payback is popular because it is legible. It is also deliberately incomplete. Before a decision is made on it, the following belong in the meeting, not in the appendix.
- Timing inside the payback period. Cash that arrives in month twenty is worth less than cash that arrives in month two. Simple payback treats them identically.
- Ramp-up. Most installations are slower in the first quarter than in the fifth. A flat monthly saving flatters early numbers.
- Risk. A payback of 24 months computed on a 60 per cent conversion share is a different proposition from the same payback computed on 95 per cent. The sensitivity table is more informative than the point estimate.
- Residual value and reuse. Equipment has a second life; licences may not. The payback calculation ignores both.
- Everything after payback. Payback says when the cost is recovered. It says nothing about what happens for the following five years.
- The cost of not deciding. A delayed decision has a price too, usually paid in overtime and in work that does not get done.
A five-step sequence
- Name the constraint. One step, described in hours per month, that the investment is supposed to relieve.
- Measure the current process. Time the work as it is done, not as the process diagram says it is done. Include waiting.
- Price the effect. Hours saved, loaded hourly cost, and a conversion share you can defend to the person who signs.
- Subtract what recurs. Put the recurring cost in the monthly line before dividing. If the result is zero or negative, say so.
- Set a review date and a falsifier. Write down what would have to be true in six months for the investment to have worked, and what would prove it did not.
Keeping the answer honest
Two habits prevent most of the damage. The first is counting a saved hour once. The same hour cannot pay back a machine, a software subscription and a training programme in the same budget cycle. The second is re-measuring after installation. A business case that is never checked becomes a story the organisation tells about itself.
None of this requires a finance department. It requires one column for the assumption that everyone privately doubts, and the willingness to publish the answer when it is no payback.
The worksheet implements exactly this arithmetic, including the negative case, and states its assumptions next to the result. If you want the measurement vocabulary first, read Explainer 01 — How to read a productivity number.
Sources
Method and citationLabour productivity, hourly compensation and unit labour cost, second quarter 2026
Published 3 September 2026. Read the source. Source of the quarterly business-sector labour productivity figure discussed in this issue.
Canadian ICT Sector Profile 2024
Data year 2024. Read the source. Source of the sector size, growth, employment, salary and research spending figures cited here.
Statistics are reproduced from the publications listed above, with their publication dates. They are fixed reference points, not live indicators. Nothing on this page is a forecast or investment advice.