Go Develop All articles
Product Engineering

The Museum of Old Decisions: Rebuilding UK Software Without Rebuilding the Past

Go Develop
The Museum of Old Decisions: Rebuilding UK Software Without Rebuilding the Past

Photo: Foreign, Commonwealth & Development Office, CC BY 2.0, via Wikimedia Commons

Parity is a comfort blanket, not a requirement

Every legacy system is a museum of old decisions. Some exhibits are important: the statutory calculation, the approval step that keeps a regulator satisfied, the customer reference that appears on every invoice. Others are fossils: the temporary field added during a merger, the report nobody has opened since 2019, the workaround introduced because a previous supplier could not integrate with a bank feed. When a UK SME decides to replace that system, the safest-sounding brief is usually feature parity. Match the old product screen for screen, field for field, report for report, and nobody can accuse the rebuild of losing anything.

That brief is a trap. Parity treats the current system as a specification when it is really a sediment of history. It preserves the wrong workflows, the duplicated data entry, the manual reconciliation and the shadow spreadsheets that grew up around the software rather than inside it. The result is a new platform with old constraints, a higher bill and a fresh decade of technical debt.

Why parity feels safe

Parity is attractive because it moves the argument away from outcomes and towards artefacts. A finance director can compare two screens. A operations manager can tick off a list of reports. A procurement team can point to a clause. None of those comparisons proves that the business will be better served. They simply prove that the old shape has been reproduced.

There is also an emotional dimension. Staff have learned the old system. They know which fields to ignore and which workarounds to trust. Replacing software is already disruptive; asking people to change their habits at the same time can feel like one risk too many. Parity appears to offer continuity. In practice, it often offers the illusion of continuity while carrying forward the cost.

For a UK SME, those costs are rarely visible in the business case. They appear later as duplicate administration, slow month-end closes, customer service queries caused by inconsistent records, and integration workarounds that need maintaining alongside the new product. The rebuild may be modern in technology but conservative in behaviour.

Separate the requirement from the artefact

A more useful exercise is to classify every requested feature into three groups: genuine business outcomes, current workarounds and accidental legacy behaviour.

A genuine outcome has a reason that survives scrutiny. It might be a statutory obligation, a contractual commitment, a customer promise or a control that prevents a known loss. A workaround exists because the old system could not do something properly. An accidental behaviour is simply how the system evolved: a field order, a compulsory step, a report layout that nobody designed but everybody now expects.

The test is not 'do users ask for it?' Users ask for what they know. The test is 'what decision does this support, what outcome does it protect and what happens if we change it?' A parity matrix can help. For each legacy feature, record the outcome it serves, the evidence for that outcome, the risk of changing it and the decision to keep, redesign or retire. That matrix becomes a governance tool, not a wish list.

Make continuity a product surface

Customers and staff do not experience continuity through identical buttons. They experience it through stable references, reliable data, familiar language and predictable service. A rebuild can preserve those things while changing the interface substantially.

The continuity layer should include data migration with reconciliation, stable customer and order identifiers, integration cut-over plans, clear communication and a period of parallel running where the risk justifies it. It should also include a deliberate retirement plan for features that are not coming back. Silence is not a communication strategy. If a report is being replaced by a dashboard, say so, show the equivalent and give people time to adapt.

This is where a rebuild earns trust. Continuity is not the absence of change; it is the absence of surprise.

Sequence the rebuild around evidence

A sensible sequence starts with an outcome map rather than a screen inventory. Identify the journeys that carry the most business risk and the most daily friction. Build a walking skeleton through those journeys, migrate real data early and test with the people who do the work. Then expand by outcome, not by module parity.

Retirement should be scheduled alongside delivery. Each release should make the old system less necessary and the new one more trusted. Measure the things that matter: time to complete a quote, number of manual reconciliations, support tickets per order, days to close the month. If the rebuild is not moving those numbers, parity has become an excuse for avoiding the real work.

Give the team permission to say no

Feature parity survives because nobody wants to be the person who removed something. A small governance group with representation from operations, finance, customer service and product can make retirement decisions explicit. It should ask for evidence, not sentiment, and it should record why a feature is being retired. That record protects the team later when someone asks why the old report disappeared.

The rebuild should be smaller than the system it replaces. That is not a failure of ambition; it is proof that the organisation has learned something. A new platform that reproduces every old compromise is not a transformation. It is a renovation of the museum, with the same exhibits and a newer roof.

All Articles

Related Articles

Paid Once, Recorded Once: Engineering Payment Flows That Withstand Repetition

Paid Once, Recorded Once: Engineering Payment Flows That Withstand Repetition

The Day After the Expert Leaves: Designing Continuity Into UK Business Software

The Day After the Expert Leaves: Designing Continuity Into UK Business Software

Reading the Queue: How UK SMEs Can Turn Support Logs into Product Evidence

Reading the Queue: How UK SMEs Can Turn Support Logs into Product Evidence