Digital Strategy & Business GrowthOctober 19, 20269 min readTika Aurora
How to Calculate ROI on Digital Investments Before You Spend
A usable ROI figure needs two numbers: a measured baseline of the manual process, and a fixed price for one defined stage of work.
Part of our complete guide: Measuring Digital ROI and Growth
If you're being asked to approve budget for a system, an app or a website, a defensible ROI calculation needs two numbers. The first is the size of the process being done by hand right now: hours per week, how often it goes wrong, how long a decision waits for data. The second is a fixed price for one stage of work with a written scope.
You can collect the first number yourself in two weeks. The second has to come from the vendor before you sign anything.
Why the ROI number usually arrives too late
In a lot of companies, ROI gets calculated after the final invoice is paid. At that point the sum can only justify a decision already made or regret it, because nothing is left to change.
The reason is usually that one of the two numbers didn't exist when the decision was taken. The denominator stays empty while a vendor offers a wide range and promises a real figure once the work is done. The numerator stays empty while nobody has ever timed the manual process.
So reverse the order: gather both numbers first, then approve the first stage. We cover judging digital spend more broadly in the guide to measuring digital ROI and growth, and this piece is the arithmetic branch of it.
Measure the process you're running now
Take two weeks and record four things about one process, not every process at once. The person doing the work keeps the record, not you guessing from across the office. Make the record as light as possible: start and finish times sent to the team chat, or one extra column on the spreadsheet already in daily use. Don't create a new form.
First, how many hours a week go into the manual step. An admin copies orders out of the chat group into the master spreadsheet, then copies them again into the accounting package.
Second, how often the same number gets typed in a second place, and how many times that fortnight something came out wrong and had to be corrected. Third, how long a decision waits for data. If the stock summary isn't ready until Monday afternoon, your purchasing decision slips a working day every week.
Fourth, write down which rules aren't recorded in any document, then write down what work stops for a week if the person who knows those rules is on leave. Without these four notes, the benefit figure in your proposal is invented. We go further into working from data you can check yourself in data-driven decision making.
The costs you count, and the ones you forget
The build price isn't the whole first-year cost, and the items people leave out aren't small.
Hosting, the domain and third-party services are billed monthly. Your own staff hours during the project count too: the admin who joins the weekly demo, the warehouse lead who digs out historical data, your own time in meetings. Cost those hours at the same rate as any other hour. Training and the transition belong in the column as well, when the old process and the new one run side by side for a few weeks.
Then there's running it after launch. At Arktik that's an optional monthly service covering hosting, monitoring, fixes and small changes. It's cancellable, and every account stays in your name. If you delete that line because you have technical people of your own, the same hosting, monitoring and fixes move to them, and their hours still belong in your cost column.
Choosing the wrong software carries a cost too, and it never appears on a quote. We treat that separately in the hidden cost of poor software choices.
Returns you can prove, and returns you can only feel
Keep the two kinds of return apart, and don't merge them into one figure.
The provable kind has a unit and can be re-checked three months after launch: hours returned to the admin, double entry removed, orders completed without a phone call to the warehouse, a summary ready Monday morning instead of Monday afternoon. Those belong in the calculation, because you can measure them again later the same way you measured the baseline.
The felt kind is real as well, and it has no number. A screen that's easier to look at, staff who aren't irritated every month end, the calm of having the data in one place. Write those into the proposal as notes and leave them as sentences. Convert them into money and your case looks better while becoming impossible to test, so when the result misses, nobody can tell which part was wrong.
A simple formula and a worked example
The formula is short: measurable annual return divided by total first-year cost, expressed in months, which is how long until the money comes back.
The numbers below only show the sequence. Replace all of them with your own. None of this is a result from anyone's project.
Start with the cost of an hour. Say salary and benefits for one admin come to $2,400 a month, divided by roughly 173 working hours, which gives about $14 an hour. From the two weeks of notes, two admins spend 15 hours a week on double entry. That's 15 × $14 × 52 = $10,920 a year. Add orders lost because the stock figure was wrong, say two a month at $200 margin, which is $4,800 a year. Measurable return is $15,720 a year, around $1,310 a month.
Now the cost side. Fixed price for the build stage $30,000, hosting and services $2,400 a year, 40 hours of your team's time at $560, training and transition $1,000. First-year total $33,960. Divided by $1,310 a month, the money comes back around month 26, which is usually the point where it's worth shrinking the first stage. Narrow the scope to one process, the double entry worth $10,920 a year or $910 a month, at a stage price of $14,000. First-year total drops to about $17,960, and payback lands around month 20.
Use the month figure rather than a percentage. An ROI percentage sounds convincing and hides when the cash returns, and cash is what decides whether you can carry it. Do the process that eats the most hours first, then calculate again before you add the next stage.
When the numbers say don't build it
If there's a ready-made tool sold monthly that already handles the process, your comparison becomes a subscription that costs far less, and the return on custom work disappears. We'll tell you that, and you spend nothing on a build. Check yourself first: open the reports menu in the software you already pay for and see whether the number you're after is sitting in there.
The second outcome shows up when the process isn't stable yet. Pricing rules that still change monthly, or a way of working that's been reorganised twice in six weeks, mean your baseline won't survive until the build finishes, and you end up with software for a process nobody uses any more. Tidy the sequence in a spreadsheet for a month or two, measure again, then calculate.
How to get a cost figure you can calculate with
The cost side stops being a guess once there's a fixed price attached to one stage with a written scope. At Arktik the first stage is discovery. It runs one to two weeks and produces a written plan with a fixed price. That document is yours whether you continue or not, and you can take it to another vendor for comparison.
So the cost of getting a number worth calculating with has a ceiling, and you know the ceiling up front. After that, the scope and price of every stage are agreed in writing before that stage starts. You can stop after any stage without an argument, and what you risk in trying us is one stage.
The code, servers, domain and every account are in your name from day one, with nothing waiting for handover. There's no vendor-exit cost to provision for in your first-year model. Alongside the account access you hold the written plan from discovery, and another vendor can pick the work up from it. What you paid for doesn't depend on us still being here next year.
Bring one process you want to cost out
You don't need a technical requirements list or a finished calculation. Describe one process you want to cost out, and if you have them, bring the two weeks of notes from your admin. We'll work out whether it's worth building and what a first stage would cost. The first call is free and there's no deck.
If an off-the-shelf tool already solves it, we'll say so. If something does need building, discovery gives you a written plan and a fixed price, and that document is yours whatever you decide afterwards. Use the contact form, which opens WhatsApp, or email hello@arktik.id.
roibudgetingfixed price stagesdiscovery