The first process to digitise, and the ones to leave alone

The first process to digitise is the one that already annoys someone: it happens often, it exists on paper or in a spreadsheet, and a named person complains weekly. Starting with the biggest process is how these projects die — the scope cannot be measured, the first version arrives late, and the people who would have defended it have moved on. The answer flips only when the big process has an owner who will sit in the room and change how the team works.

Which three tests pick the first candidate?

It happens often. Daily or weekly, with a predictable volume. Frequency is what makes a first version worth building: you learn from the tenth run, not the first, and a monthly process takes a year to prove anything.

It is written down somewhere. A paper form, a spreadsheet with columns, a message template people copy. A process already written has a shape: you replace a medium instead of inventing a process while building software for it. If the only record is in someone's head, the first job is a conversation.

Somebody already complains about it. Not "we could be more efficient" from a consultant, but a person who says the current way wastes their Thursday. That complaint is the requirements document, and it is worth more than a specification written by someone who has never done the task.

A candidate that passes all three is small, frequent, visible and owned. If it passes one, it is a project with a sponsor and no user.

Why does an unhappy owner beat a happy one?

Because the person who complains can describe the failure precisely and will test what you build. They know which field gets skipped, which customer always causes the exception, and which step exists only because someone left in 2019. They also have the standing to change the way the work is done, and the motive to defend the change when a colleague resists it.

A happy process is a trap for a different reason: nobody complains because nobody looks. The work is done quietly, with workarounds that have become invisible, and the users will not tell you when your version breaks one of them. Silence is not satisfaction; it is usually absence of attention.

What happens when you digitise a bad process faithfully?

You get the waste, faster, with a receipt. Every unnecessary step becomes a screen, every approval nobody needs becomes a button, and the report now proves that the waste happened on time. Software makes a bad process harder to argue with, because it looks deliberate.

So the step question comes before the build: what does this step prevent? Three copies of a form prevent a lost page; a manager's approval prevents a discount nobody authorised. If nobody can name the incident a step prevents, the step is a habit, and a habit should be dropped before it is automated.

What should be left alone for now?

Anything that changes monthly. The process has not settled, so you will build it twice: once now, and once when it stops moving. Wait for a shape.

Anything with one user. There is no leverage in it, and a spreadsheet is already the correct tool. A system adds hosting, logins and maintenance to a task one person does in ten minutes a week.

Anything the business is about to stop doing. A product line being discontinued, a channel being closed, a supplier being dropped. Digitising it is a monument to a decision already taken.

Anything whose owner refuses to change the way they work. You would be building against a person rather than for a process, and the version that fits their resistance is the current one.

Signal	Digitise	Leave alone
Frequency	Daily or weekly	Monthly, or unpredictable
The record	A form, or a sheet with columns	One person's memory
The users	Several, one complaining	One, none complaining
Lifetime	Still done next year	Ending this year

How do you know it worked?

Not hours saved, and certainly not a percentage. The measure is a question somebody can now answer that they could not answer before: which deliveries are late and by how many days; how much stock sits at the second shop; who approved this discount and when.

That question has to be answerable by the owner in under a minute, from the screen, without calling you. If it is, the project produced information and not just a faster form. If no new question is answerable, the system replaced paper with a screen, and the only change is that the paper is now harder to find.

A decision procedure

1. Sit with the person who complains most, not with management, and list the processes they touch.
2. Score each on the three tests. Drop anything that passes fewer than two.
3. Watch one full cycle without helping. Count the steps, the handoffs and the times someone waits for someone else.
4. Ask of each step what it prevents. Delete the steps with no answer before anything is built.
5. Build only what remains, and name the one question the owner will be able to answer afterwards.
6. Run paper and system in parallel for one cycle, compare the counts, then switch on a date you chose.
7. Keep the old path documented for a month, then retire it deliberately instead of leaving both alive.

Where this goes wrong

• The biggest process is chosen for its visibility. It is the one management talks about, with six owners, four of whom disagree.
• The pilot is chosen politically. A director's favourite gets built, nobody uses it, and the lesson drawn is that staff resist change.
• Only the form is digitised. Paper becomes a screen: no queue, no counts, no question answered, and the same exceptions handled by the same phone calls.
• "Hours saved" is the measure, self-reported. The person whose job changed is asked whether it improved, which is a survey rather than a measurement.
• There is no owner after launch. The system works until the first exception, which arrives in week three.
• The old way survives quietly. Two parallel records, and the paper one wins because it is the one people trust.

What this does not cover

This is not a method for a transformation programme with a budget and a steering committee, and there are no figures here, because the numbers that matter are counted in your own business. It assumes a few dozen people, visible processes, and one person who can decide. Where a process crosses companies — customs, a courier, a distributor — the first candidate is the part inside your own walls. And digitising is not always the answer: sometimes the fix is one rule, one signature removed, and a phone call.

We start with the process somebody already complains about, remove the steps nobody can justify, and measure it by the question the owner can finally answer.