Hours × loaded cost × 52. Then subtract everything you were hoping to forget.
Automation ROI is hours saved a week × the loaded hourly cost of the person doing the work × 52, minus the build, the running cost and the attention it takes to keep alive. Then compare it with doing nothing, which is not free. Most of the arithmetic below is uncomfortable, which is why calculators skip it.
Start with the loaded hourly cost, not the wage
The wage is the smallest honest number. Full-time adult average weekly ordinary time earnings were $2,083.70 in May 2026 (ABS, Average Weekly Earnings, Australia, May 2026, released 13 August 2026) — $54.83 an hour over a 38-hour week. On top of that: super guarantee at 12% (ATO, Super guarantee, rate from 1 July 2025), and workers compensation at WorkCover Queensland’s average rate of $1.343 per $100 of wages for 2026–27 (WorkCover Queensland, 26 June 2026). Payroll tax in Queensland only starts above $1.3 million of annual taxable wages, at 4.75% (Queensland Revenue Office, payroll tax rates and thresholds, checked September 2026), so most small businesses can leave it out honestly.
Run the sum. $2,083.70 × 52 is $108,352 a year. Add 12% super ($13,002) and 1.343% WorkCover ($1,455) and the annual cost is $122,810. Divide by the 1,976 paid hours in a year (52 × 38) and the loaded hourly cost is $62.15. That is the number to use in the formula, and it is 13% higher than the wage most people start with.
There is a second, higher figure that is also honest: the cost per hour actually worked. Full-time staff get four weeks of annual leave and ten days of personal leave under the National Employment Standards (Fair Work Ombudsman, Fair Work Information Statement, current edition, checked September 2026), and Brisbane had about nine weekday public holidays in 2026 (Queensland Government, public holidays, 2026 calendar). Take out roughly 7.8 weeks and the same $122,810 lands on 1,680 worked hours — $73 an hour. Use it if you are costing hours in worked weeks; use $62.15 if you are multiplying by 52. Both arrive at the same annual figure. Mixing them is how vendor calculators get a bigger number.
Count the hours on one real job, not the business in general
Most people undercount by half, because they count the re-typing and miss the chasing, the checking and the fixing. Follow one job from enquiry to paid invoice and write down every time a human moves the same information from one place to another, waits for information that has not arrived, checks something was done, or corrects it when it was not.
There is no Australian government figure for how many hours a week a small business spends on admin. Every version of that statistic you have seen traces back to a survey run by a company selling the fix, and we are not going to borrow one. The number that matters is yours, and it takes an afternoon to count — the benchmark page says exactly what is and is not published.
What is published: 32% of Australian businesses received orders online in 2024–25 (ABS, Characteristics of Australian Business, 2024–25, released 25 June 2026), so for two thirds every order still arrives by phone, email or paper and is re-keyed by someone. If that someone is you, the hourly cost is whatever an hour of the owner is worth, which is higher than any wage figure above.
The worked example, all the way through
Say the count comes back at six hours a week of re-keying between a job management system and an accounting package, done by a full-time office administrator on the average wage above. Six hours × $62.15 × 52 is $19,391 a year. That is the gross saving, and it is the only number most calculators show you.
Now the costs. Suppose the connection is quoted at $12,000 — an illustration, not our price; our minimum engagement is $5,000 and the real number depends on the systems. Running cost: hosting and API usage, call it $60 a month. If it were built on Zapier instead, the Professional plan starts at US$19.99 a month (Zapier pricing, checked September 2026, USD) and the bill rises with the tasks it completes.
Payback in months is build ÷ (monthly saving − monthly running cost). $19,391 ÷ 12 is $1,616 a month; less $60 is $1,556; $12,000 ÷ $1,556 is 7.7 months. Year one nets $7,391 after the build and running costs; year two nets $18,671. On those inputs the case holds comfortably.
Change the person and the case changes. An administrator on the Clerks — Private Sector Award, Level 2, is paid $29.45 an hour from 1 July 2026 (Fair Work Ombudsman, Clerks — Private Sector Award pay guide, rates from 1 July 2026) — loaded to $33.38 with super and WorkCover. The same six hours saves $10,414 a year and the same $12,000 build pays back in 14.9 months. Still positive; twice as slow. The wage of the person doing the work is the single most sensitive input, and it is the one calculators default for you.
The do-nothing option has a cost curve too
Doing nothing is not free; it is the current cost, rising. Award minimum wages rose 4.75% from 1 July 2026 and the national minimum wage 5.97%, to $26.44 an hour (Fair Work Ombudsman, Annual Wage Review 2026, from 1 July 2026). Average weekly earnings rose 3.7% over the year to May 2026 (ABS, Average Weekly Earnings, Australia, May 2026, released 13 August 2026). Six hours a week of re-keying costs more every July, whether or not anything else changes.
Doing nothing also carries error cost. Just over half of small businesses reported their tax correctly and 35% tried to comply but made errors (ATO, Small business random enquiry program findings, 2022 bundled sample covering 2019–20 to 2021–22, published 4 November 2025); simple mistakes doubled from 11% to 22% between the 2017 and 2022 samples (ATO, Small business random enquiry program findings, 2022 bundled sample, published 4 November 2025). Re-keying is where simple mistakes are made. On the debtor side, Australian small businesses were paid 6.6 days late on average in the December 2025 quarter (Xero Small Business Insights, published March 2026) — chasing that by hand is a cost most businesses never write down.
Put the do-nothing line on the same table as the automation line: current hours × loaded cost, growing at about 4–6% a year, plus an honest estimate of what errors cost you last year. If the automation still does not beat it, do not build.
Sensitivity: what happens when your inputs were wrong
They will be. Halve the hours and rerun: three hours a week saves $9,696 a year on the average wage, and the $12,000 build pays back in 16 months. Add a 50% build overrun on top and payback is two years. If the case only works when every input is right, it does not work.
The published research points the same way. In the Standish CHAOS sample, about one project in five failed outright and half were “challenged” — late, over budget or short on scope — while small projects succeeded 61% of the time against 6% for the largest (Standish Group, CHAOS Report 2015, projects sampled 2011–2015 — a US research firm and not an Australian dataset). It is not Australian data and we flag that; the lesson is still to keep the build small enough that a bad case is survivable.
Our rule: run the halved case. If payback on the halved case is still under 18 months, proceed. Between 18 and 36, tighten the count — go back and measure the hours for a fortnight rather than estimating them. Over 36, the answer is no, and it is a real answer. If the process itself is unstable or undocumented, the ROI is irrelevant and that page comes first.
- KNDR·01Base case6 hrs/week, $62.15 loaded, $12,000 build, $60/month running: payback 7.7 months.
- KNDR·02Hours halved3 hrs/week, same costs: payback 16 months.
- KNDR·03Award wage instead of average6 hrs/week at $33.38 loaded: payback 14.9 months.
- KNDR·04Hours halved and build overruns 50%3 hrs/week, $18,000 build: payback 24 months. This is the case to decide on.
- KNDR·05Hours saved but not redeployedAny of the above: cash payback never. Capacity only. Decide beforehand what the hours become.
What the vendor calculators leave out
Four things, every time. Attention: someone watches the automation, handles exceptions and updates it when a system changes its API — budget an hour a fortnight at the loaded rate. Change: the process will move, and the automation moves with it at a cost. Platform fees that rise with success — per-task and per-seat pricing scale with the thing you were making cheaper. Redeployment: hours only become cash as billable work, avoided hiring or a shorter day for the owner. Otherwise they are capacity, valuable but not bankable.
The honest version of the formula is therefore: (hours × loaded cost × 52) − build − (running + attention) × 12, compared against (current hours × loaded cost × 52 × growth) + error cost. If a calculator does not have a line for attention or for the do-nothing case, it was built to produce a yes.
Run it yourself before anyone quotes you
The leak calculator does this arithmetic in your browser. It uses the ABS average weekly earnings figure above for the wage default and the 12% super guarantee for on-costs, and it returns a range rather than one confident number. If the range comes back below our $5,000 minimum engagement, that is a real answer and you should stop there — we say the same on how we charge, where about 40% of Leak Checks end without a proposal.
If the sums do hold, the next question is whether to buy a product or build the join, and the build-vs-buy framework takes about an hour. The scorecard below is the shorter version for the ROI question alone.
Ten questions about the case you are building. Answer for one process. If you have not done the arithmetic yet, most of these will be “no”, and that is the honest starting point.
- Have you counted the hours on one real job rather than estimating them?
- Did the count include chasing, checking and correcting — not just re-typing?
- Are you using a loaded hourly cost that includes super and workers comp?
- Do you know the person’s actual wage rather than a default?
- Is every recurring cost listed — hosting, platform fees, API usage?
- Have you budgeted attention: someone’s time to watch and maintain it?
- Have you costed the do-nothing option, including wage growth and errors?
- Does the case still pay back inside 18 months with the hours halved?
- Have you decided what the saved hours will become?Billable work, avoided hiring, or a shorter day — named, not assumed.
- Is the process stable and documented enough to automate at all?
0 of 10 answered. The result appears once you have answered them all.
Nothing here is sent anywhere. If the halved case fails, so does the case — a build that only works when everything goes right is the kind we would rather not sell.
The questions people actually ask.
01Why do your numbers come out lower than the vendor calculator?
Because we subtract things. Attention, running cost, the do-nothing line and a halved-hours test all reduce the figure, and using the paid-hours rate with × 52 rather than a worked-hours rate stops the double count. A lower number you can defend beats a higher one you cannot.
02The saved hours won’t turn into cash. Is the automation still worth it?
Sometimes — as capacity. If the owner gets six hours a week back and spends them quoting, that is real. If nobody has decided what the hours become, they become nothing, and the ROI is a spreadsheet. Decide first.
03Should we include the owner’s time at a wage rate?
No — it is worth more than that, and pricing it at the average wage understates the case. Use what an hour of your time earns when you spend it quoting or on site. It is the one input where the honest number is higher than the default.
Stop bending.
Start shipping.
Bring the one flow that keeps costing you a Tuesday afternoon. We map it live and tell you what we'd build, what we wouldn't, and what it costs.