Working Out the ROI of Custom Software for a Small Business

The formula, without the spreadsheet theatre

ROI calculations for small-business software do not need discounted cash flows. Two numbers do the work. First, the full first-year cost: the fixed build price, plus twelve months of running costs (hosting, API fees), plus an honest allowance for your own hours during the project. Second, the monthly benefit: hours saved per month multiplied by the loaded hourly cost of whoever saves them, plus any hard cost the software removes, such as a subscription it replaces.

Divide cost by monthly benefit and you have payback in months. The threshold is a business judgement, but a usable rule: under twelve months, strong case; twelve to twenty-four, reasonable if the process is stable; beyond thirty-six, the money almost certainly has a better home. Loaded cost matters — a person's true hourly cost includes employer taxes, benefits and overheads, typically well above bare salary, and using bare salary quietly understates every saving.

A worked example with honest numbers

A five-person services firm produces quotes and invoices by hand: templates, copy-paste from a price list, chasing by memory. Call it six hours a week across the team — a deliberately modest figure; measure your own before trusting anyone's example. At a loaded cost of $35 an hour, that is roughly $910 a month in admin time.

How much of the work week is actually automatable

Before hunting for a project, it helps to know how big the hunting ground is. McKinsey Global Institute's automation research estimated that about half of all work activities globally could theoretically be automated by adapting currently demonstrated technologies — and, more usefully for a small business, that in roughly 60% of occupations at least 30% of the constituent activities are automatable. Very few whole jobs disappear into software; large slices of most jobs do.

Treat those numbers as a prompt, not a promise. The practical translation: almost every small business contains more automatable hours than it has budget to address, so the ROI exercise is a ranking problem. List the recurring manual processes, estimate weekly hours for each, and let the payback formula pick the first project. The winner is usually unglamorous — quoting, rekeying, chasing, reconciling — because frequency beats drama in this arithmetic every time.

Count the costs people forget

The build price is the visible cost; three others belong in the calculation. Running costs are small but real — hosting, transactional email, any API usage — typically $15–50 a month for a tool of this size. Your own time is the one everyone omits: answering questions during the build, testing, and rolling the tool out to the team. Ten to twenty hours is normal for a small project, and it is only honest to price them.

The third is the adoption dip. For the first couple of weeks, the team is slower with the new tool than they were with the old mess, and some processes run in parallel. It is temporary and normal, but a calculation that assumes full savings from day one will look wrong for a month and shake confidence in the whole exercise. Start the savings clock at week three or four.

Why build costs fell — and what that does to the formula

Payback thresholds that made sense at agency prices deserve rethinking, because the denominator changed. The production shift is thoroughly documented: Google Cloud's 2025 DORA report found 90% of software professionals now use AI at work, and Stack Overflow's 2025 Developer Survey found 51% of professional developers using AI tools daily. A controlled GitHub experiment measured developers completing a standardised task 55.8% faster with an AI assistant. AI-based software development is simply how small projects get built now — and the labour saving lands in the quote.

The first-party numbers show where that leaves prices: across PINCLER's 79 documented projects, every build landed between $500 and $2,500 fixed, with a median of $1,450 — the full dataset is at pincler.com/research/what-you-can-build. When the build cost drops by an order of magnitude, processes that failed the ROI test five years ago pass it comfortably today. The formula did not change; the inputs did, and it is worth re-running old 'no' decisions against current prices.

Revenue upside is real — but keep it out of the core case

Software often earns as well as saves: an enquiry answered in two minutes converts better than one answered tomorrow, an online booking page captures the customers who will not phone, a tidy follow-up sequence revives quotes that would have gone quiet. These effects are real, and for some builds they dwarf the time savings.

They are also hard to predict for your specific business, which is why the disciplined move is to build the case on time saved — measurable, defensible — and treat revenue effects as upside. If a project only makes sense when optimistic revenue assumptions hold, it does not make sense yet. If it pays for itself on hours alone, the revenue effects arrive as a bonus rather than a requirement.

Payback benchmarks from 79 fixed-price projects

To turn the formula into expectations, here are category medians from across PINCLER's 79 documented projects, paired with the monthly saving each would need to pay back inside twelve months. The saving column is pure arithmetic — median price divided by twelve — and most operational processes clear it easily once hours are counted at loaded cost.

Read the last column against your own numbers: $100 a month is under three hours of saved admin at $35 loaded. A process consuming six hours a week — roughly $910 a month — pays back the median automation in under two months. That gap between required saving and typical saving is why small fixed-price builds so often produce the best ROI a small business will see anywhere.

When the calculation says do not build

Run honestly, the formula returns no reasonably often, and the reasons repeat. The process happens too rarely — automating a quarterly task that takes an afternoon saves a few hours a year, and no build price justifies it. The process is still changing shape — software freezes a workflow, and freezing the wrong one costs more than the manual work did. Or an off-the-shelf tool already covers it — if a standard $30-a-month product does the job, the ROI of custom code is negative by exactly the build price.

There is also a scale threshold worth respecting: below roughly two or three hours saved a week, even a $500 build competes poorly with simply doing the work. The best candidates are boringly consistent: high-frequency, stable, rule-based processes that someone currently performs by hand every single week.

The three-year view: build, subscribe, or keep doing it by hand

Payback answers 'how fast'; the three-year comparison answers 'which option'. Take the quoting process from the worked example and price all three routes over thirty-six months. Doing nothing: roughly $910 a month in admin time, about $32,760. A SaaS subscription that fits the workflow, if one exists at say $60 a month: $2,160, plus whatever manual steps survive around its edges. The custom build: $1,400 once plus about $15 a month running, roughly $1,940 — with the workflow fitting exactly and no per-seat meter as the team grows.

The ranking will differ per process, and the SaaS option wins whenever a standard product genuinely fits — that is the honest answer more often than a development studio should admit. The point of the exercise is that 'keep doing it by hand' is also a purchase, made monthly, and it is routinely the most expensive of the three once it is priced.

Run your own numbers, then get a real quote

The exercise takes twenty minutes: pick the process, measure the weekly hours honestly for two weeks, multiply by loaded cost, and set the result against a realistic build price and a year of running costs. Do this before talking to any developer — including us — because walking in with your own payback number is the strongest negotiating position a buyer can have.

When you want the other side of the fraction filled in, a free 30-minute call gets you a written fixed quote within one working day, always between $500 and $2,500 per phase. Or browse the automation and internal-tool use cases first — each lists a fixed price band you can drop straight into the formula.