How Long Until Custom Software Pays for Itself? The Actual Maths
Every pitch for custom software ends the same way: "it will pay for itself." Almost none of them show the maths. Here is the actual formula, three worked examples with real Australian price bands, and the honest case where the answer is "do not build it".
The formula
Payback period = build cost ÷ (yearly savings − yearly running costs)
Yearly savings = hours saved per week × fully loaded hourly cost × 48 working weeks. Use the loaded cost of the person doing the work (wage plus super plus overhead), not their base pay rate. Running costs are hosting and any maintenance retainer.
That is the entire trick. Everything else is estimating the inputs honestly, which is where most pitches quietly cheat. Estimate hours saved by watching a week of the actual task, not by gut feel.
Example 1: admin automation that works
A trades business spends 6 hours a week re-keying job details between enquiry emails, a spreadsheet and Xero. Loaded cost of the office manager doing it: about $45 an hour.
- Yearly saving: 6 × $45 × 48 = $12,960
- Build cost (business automation band): $8,000; hosting $100 a month = $1,200 a year
- Payback: 8,000 ÷ (12,960 − 1,200) = about 8 months
Everything after month eight is roughly $980 a month back in the business, indefinitely.
Example 2: a quoting system, counted honestly
A manufacturer takes 4 hours of a senior estimator's week ($60 an hour loaded) to produce quotes by hand: 4 × $60 × 48 = $11,520 a year. A $15,000 quoting tool that halves quote turnaround also plausibly wins one extra job a month; call it $800 of margin each, but treat that as upside, not the case. On time savings alone: 15,000 ÷ (11,520 − 1,200) = about 17 months. With the extra jobs, closer to 9. Fund the decision on the 17-month number and let the upside be pleasant.
Example 3: the one you should not build
A client portal to reduce status-update phone calls: 3 hours a week saved at $45 an hour is $6,480 a year. At a $12,000 build with $1,200 a year of running costs, payback is about 27 months. That is past my own two-year line: at that horizon your business will have changed enough that the software might need to as well. Cheaper fix first: a weekly update email template, which costs nothing.
The two-year rule
If the honest inputs put payback past 24 months, do not build it yet. Shrink the scope until the first version pays back inside a year (automate the single worst task, not the whole workflow), or wait until volume grows into the maths. The best custom software projects are boring: a clear repetitive cost, removed, permanently.
Run your own numbers in the free payback calculator; it does this arithmetic with your figures and is deliberately conservative. If the result looks good, the usual next steps are automating the manual process or scoping a small business automation build. And if the result looks bad, you just saved yourself five figures, which is the cheapest consulting you will get this year.