The migration took eight months. A year later, the headaches were the same.

How not to… · “Sound familiar?” · 06. The migration took eight months. A year on, the problems were the same. The bottleneck is seldom “the platform” alone — more often it is the missing answer to what we will actually do with it after go-live.

Eight months. That is how long the migration ran. The budget earmarked for two years ahead was blown on a single programme. At kick-off it looked like every other project: new stack, new interface, new possibilities. The board deck sounded convincing.

Twelve months later the new platform carried the same tensions as the old one. The same bespoke pricing exceptions handled by hand. The same logistics workarounds. Still no owner for product data quality. The same promotion policy that never lived in one system — only in three spreadsheets and two people who “just know how it works”.

You moved everything — including what was already broken.

Why “newer” is not automatically “better”

Organisations lift-and-shift the same compromises they spent years baking into the legacy stack. A new platform is not a fairy godmother for transformation; it is a faster front door to the same operational mess.

A tool amplifies whatever is already there. If what you have is chaos, you get faster chaos.

Hence the line on the card: the platform is rarely the real constraint. More often the constraint is no shared picture of what to do with it — which outcome we expect next quarter, who actually decides, and how we measure success beyond the word “deployed”.

Role boundaries stay put. How you calculate margin stays put. How data flows between departments stays put. Catalogue quality stays put. Only the UI is new — and a year later everyone is shocked that nothing feels different.

What you migrate to the new platform

Tick what already exists in your organisation before migration. Whatever you tick, you will migrate.

What to do differently before the next RFP

The IT project and the operational project are two different animals. The first ends at go-live. The second starts the day after and covers people, process, training, and data. Companies that conflate the two tend to pay twice for the same migration.

Swapping platforms without moving the organisation

The pattern resurfaces every few years under a new acronym. New ERP. New CRM. New e-commerce engine. New WMS. Each time the migration runs longer than planned and costs more than budgeted.

Each time, a year after go-live, the same tensions are back — because you moved data and features, not the wiring around them: who decides what, how margin is calculated, how information actually flows, which policies still live in people’s heads instead of in software.

Most AI debates sit in the same squeeze between **tool and organisation** — see who will deploy AI for us (sequence, what “deployment” even means, costs beyond the quote).

Part of the “How not to…” thread from the 70% of programmes with no operational target and Gemba, plus the dead end of “we will just train our own model” — when language and data are the issue, not the licence fee alone.