Skip to content
TechFabric

Fix the foundation

Modernise the estate that outgrew itself

The systems that got you here are now the reason things take a quarter. We take them apart in an order that keeps the business running.

Everything works, and nothing can change

Nothing is on fire. That is the problem, because a system nobody can safely modify still costs a quarter every time somebody asks for something.

A small change takes a quarter

Not because it is hard, but because four systems have to agree and nobody is certain what depends on what.

The integrations are the estate

A decade of acquisitions and point-to-point connections. Most of the expensive failures are integration failures wearing another name.

The move happened and nothing changed

The workloads run in the cloud now. They run the same way, cost more, and the reasons anybody wanted to move are still on the list.

The estate is not badly built. It was built for a company that was smaller, and nobody has been funded to redesign it since.

The shift

Three ways to modernise

One of them is the reason people distrust the word.

Lift and shift

Moved, and unchanged

Inventory
Rehost
Bill goes up
Same constraints
Nothing improved

Rewrite it

Correct, and it does not land

Greenfield started
Business keeps changing
Two systems to run
Scope grows
Cancelled

Strangle it in order

Incremental, and it finishes

Seams found
One capability extracted
Traffic moved
Old path retired
Repeat

How the work runs

Sequenced so that value arrives before the programme does, because a modernisation that pays nothing for a year gets cancelled in month nine.

Read what exists

Honestly, including telling you which parts are worth keeping. We would rather do that than sell a rewrite you may not need.

Find the seams

Where a capability can be extracted without the rest noticing. This is most of the skill and it is why the order matters.

Extract one thing

The first capability moved end to end, with traffic actually shifted, so the pattern is proven rather than designed.

Durable workflows at the joins

The multi-day, multi-system processes that must never half-complete get rebuilt on durable execution rather than on retries and a nightly job.

Retire the old path

Each extraction ends with the previous route turned off, because two live paths is how a modernisation becomes permanent.

Leave the map

The written record of what was decided, so the next extraction does not start by rediscovering the estate.

One at a time

Capabilities extracted, each with the old path retired

The failure mode we design against

The most common estate we are called into has already been through one of these. It moved, it cost more, and the constraints stayed. Sequencing for early value is the whole difference and it is a choice made in week one.

FAQ

Estate modernisation: common questions

We already migrated to the cloud. Was that wasted?

No, and it is a common starting point. The move solved hosting. What is usually left is the reason anybody wanted to move, and that work is separable from the hosting decision.

Do you rewrite or refactor?

Extract, mostly. A rewrite is occasionally right and it is almost never the first answer, because the business keeps changing while it happens and two live systems is a worse position than the one you started in.

How do we justify this internally?

By the first extraction paying for itself. If the sequence does not put value in front of the programme, it will be cancelled in month nine and that is a reasonable thing for a board to do.