Retrofit before you replace

Every modernisation proposal I have seen starts with the same instinct: replace the old thing. New hardware, new vendor, clean architecture, no legacy. It is an appealing story and it is usually the wrong first move.

The equipment on the ground is often mechanically fine. What it lacks is not durability — it is a data interface. And adding a data interface is a fundamentally different project from replacing an asset base.

The economics are not close

Replacement means capital expenditure across the entire estate, an installation programme, disposal of working assets, and a long period where half your sites run the old system and half run the new one. Retrofitting means a module, an installation visit measured in minutes, and an estate that becomes uniformly readable in months rather than years.

When I ran a programme to digitise thousands of legacy units, the retrofit approach unlocked new revenue inside the first year — revenue that would not have arrived until year three under a replacement plan, because the replacement plan spends year one on procurement.

The point is not that retrofitting is cheap. It is that retrofitting converts a capital project into an operational one, and operational projects can start on a Tuesday.

What you actually gain is measurement

The revenue does not come from the hardware. It comes from finally knowing things.

Before instrumentation, an operation runs on estimates, and estimates drift in one direction: toward under-billing, over-provisioning, and unexplained losses. In the programme above, the utilities balance moved from deeply negative to solidly positive without changing a single price. Nothing about the commercial model changed. We just stopped guessing.

That is the pattern almost everywhere:

1. Instrument the asset.
2. Discover the gap between what you assumed and what is true.
3. Close the gap. The gap *is* the business case.

Where retrofit genuinely fails

It is not a universal answer, and pretending otherwise burns credibility:

• Safety-critical certification. If the retrofit invalidates a certification, the argument is over.
• Mechanical end-of-life. Instrumenting something with two years left is a rounding error away from waste.
• No physical access. If the module cannot be installed without a crane, your installation-cost model is wrong.

Test all three honestly before you commit. A retrofit programme that hits one of these mid-rollout is worse than a replacement programme you never started.

The design constraint that matters most

Assume installation will be done by someone who has never seen your product, in bad weather, with the manual in their pocket rather than their hand. Every minute you remove from the install procedure multiplies across the entire estate — and every ambiguous step becomes a support ticket that arrives eighteen months later.

Retrofit projects are not won in the architecture diagram. They are won in the fifteen minutes a technician spends on a ladder.