You cannot digitise a mess

Someone will say it in the meeting, usually after a bad week. "We need to digitise." The warehouse is chaotic. The spreadsheet is a folklore document. The 3PL portal and the marketplace portal and the finance system do not agree on how many orders exist. Digitise sounds like a clean verb. It sounds like progress.

What usually happens next is a project. A vendor. A mapping workshop that maps the mess in higher resolution. Then a go-live that makes the same mess faster, with login credentials.

I've implemented systems. I've inherited systems. I've watched good software lose to a process that was never willing to be specific. Digitisation is not a strategy. It is a multiplier. Multipliers do not care whether they are multiplying sense or nonsense.

You cannot digitise a mess. You can only make the mess searchable, timestamped, and harder to walk back from.

What people mean when they say digitise

They mean "I am tired of this being held in people's heads." Fair. Heads take leave. Heads argue. Heads remember a version of the SOP that was true in 2022.

They also mean "I want a screen so I don't have to go to the floor." Less fair. The floor is where the truth is. A screen that is not fed by a real event is a rumour with a UI. Digitising a rumour gives you a dashboard you will believe, which is worse than a WhatsApp group you at least knew was messy.

Sometimes they mean "finance wants an audit trail." Also fair. An audit trail of a broken handoff is still a broken handoff. You will just be able to prove, in a board pack, that you broke it at 14:07.

The mess has a shape

It is usually not that nobody works. It is that three processes are running in parallel and pretending to be one. The SOP. The way the night shift actually does it. The way the 3PL does it when your SOP and their WMS disagree. Digitise any one of those and the other two will fight the system until someone creates a workaround. The workaround becomes the real process. The system becomes the thing people click to make the workaround legal.

I have seen barcode processes where the scan happens after the carton is already on the truck, because the pack bench was never moved, and the system required a scan, and the truck would not wait. The data said packed. The parcel was gone. Digitisation succeeded. The operation did not.

The same thing happens in customs files, in COD recon, in returns. If the human path is "message someone who knows," the digital path will be "ticket someone who doesn't," unless you name the knowing and put it in the flow.

Clean the job, then put it on a screen

Walk one order all the way through. Not a workshop on stickies. An actual order. Who touches it. What they need. What they do when it fails. Write the minimum path that a new person could run without calling the hero.

Kill the parallel versions. There should not be a night-shift process and a day-shift process for the same SKU unless the work is genuinely different. If it is genuinely different, write both. Do not let the system discover the difference in production.

Then choose the smallest digital thing that enforces the path. Sometimes that is a WMS. Sometimes it is a decent OMS. Sometimes it is a checklist and scan discipline and you are not ready for a platform. Buying the platform first is how you spend a year configuring a mess you have now made official.

A system is not a brain. It is a very consistent intern. Brief it like one.

The temptation of the tower

Control towers, integration layers, "one pane of glass." Useful after you have events worth seeing. Dangerous before. You will integrate five lies into a sixth, prettier lie. Then you will staff a team to explain why the pane of glass and the warehouse disagree.

I've been in that team. It is not a technology career. It is a career in translating between a floor that is real and a screen that is political. Don't hire that team until the floor and the screen are trying to describe the same job.

At 500 parcels a day you can digitise later. At 5,000, the mess will digitise itself in the worst way, through tribal workarounds. The sequence is still the same. Name the process. Then give it a system. Never the other way around, no matter how tired the meeting is.

The goal was never a digital operation

The goal was an operation that is true, then a system that keeps it true when the person who remembers it is not in the building.

Digitise that. Not the pile. If your project plan starts with a vendor and a target architecture, you are already late. Start with the order, the exception, and the sentence a shift lead can say without lying. Then buy the software that will repeat the sentence, every time.