Legacy modernisation

Why it matters

A bank’s core holds decades of products, rules and integrations. Replacing it means moving every account and open loan without losing a rupee, while the bank keeps trading. A “big bang” replacement, where everything switches on one weekend, concentrates all of that risk into one moment. Incremental approaches spread it out, so each step is small enough to test, reverse and learn from.

The model explained

The strangler fig

Named after a vine that grows around a tree until it can stand on its own, the strangler fig pattern puts a facade in front of the old system. At first the facade simply passes every request through. Then, one capability at a time, requests are routed to new services instead. When nothing is left routed to the old system, it can be switched off.

Supporting patterns

• Anti-corruption layer. New services talk to the old system only through a translator, so the old system’s data model and quirks do not leak into the new design.
• Wrapping the core. Before replacing anything, put APIs in front of the core. Channels then depend on the APIs, not the core, which makes every later step easier.
• Branch by abstraction. Inside a codebase, introduce an interface over the thing being replaced, move callers to the interface, then swap what sits behind it.
• Hollowing out. Move functions such as payments, customer data or lending out of the core one by one, leaving it as a thinner ledger.

Sequencing the moves

Order matters. Many programmes start with capabilities that are valuable but loosely coupled to the ledger, such as customer onboarding, notifications or a new digital product, because they can move without touching balances. Deposits and loans move later, product by product, once migration tooling and reconciliation have been proven on smaller waves. The general ledger and end-of-day processing usually move last, because everything depends on them.

Data migration and parallel runs

Data is usually the hardest part. Migration moves accounts in waves, by product, region or customer segment, with automated checks that every balance, rate and schedule arrived intact. A parallel run processes real transactions through both old and new systems and compares the results before customers depend on the new one. Differences found in a parallel run are cheap; differences found by customers are not.

Cutover

Cutover is the moment a wave of customers starts being served by the new system. It needs a rehearsed plan, a clear decision point, a tested way back and enough people on hand to resolve problems quickly. Most failed migrations fail here, usually because the rollback was never really tested.

How it plays out in Indian banking

Core changes at large Indian banks are multi-year programmes. SBI's core banking has run on TCS BaNCS since an implementation that began in 2001; in October 2021 TCS said the partnership was extended for five years.^{1Tata Consultancy Services. SBI deepens partnership with TCS to drive innovation and enhance customer experience. As of 6 October 2021.} Most are carried out with vendors and system integrators, which brings RBIReserve Bank of India. India's central bank. It regulates banks and payment systems, runs RTGS and NEFT, and holds the accounts in which banks settle with one another.’s outsourcing rules into scope: RBI issued the Master Direction on Outsourcing of Information Technology Services on 10 April 2023 (RBI/2023-24/102), effective 1 October 2023.^{2Reserve Bank of India. Master Direction on Outsourcing of Information Technology Services (RBI/2023-24/102). As of 10 April 2023.} A regulated entity's outsourcing policy must include termination processes and exit strategies, including business continuity if a service provider leaves the arrangement.^{2Reserve Bank of India. Master Direction on Outsourcing of Information Technology Services (RBI/2023-24/102). As of 10 April 2023.} A new core must also meet the same resilience tests as the old one before it goes live. A DR test must switch over to the DR or alternate site and run it as the primary site long enough to cover at least a full working day, from beginning of day to end of day.^{3Reserve Bank of India. Reserve Bank of India (Information Technology Governance, Risk, Controls and Assurance Practices) Directions, 2023 (full text). As of 7 November 2023.}

Trade-offs

Failure modes

• The permanent middle, where the old and new systems both keep running indefinitely because the last capabilities are never moved.
• Untested rollback, discovered only when a cutover goes wrong.
• Migration defects that surface weeks later as wrong interest or charges.
• Scope creep, where the migration becomes a chance to redesign every product at once.
• Knowledge loss, when the people who understood the old system leave before its rules are documented.

When to use it, and when to avoid it

Prefer incremental change whenever the old system can be wrapped and the business can tolerate running two systems for a while, which is most of the time. Consider a big-bang switch only for small, self-contained systems, or when the old one is failing so badly that running it longer is the greater risk. In every case, test the rollback as seriously as the rollout.

Questions for a design review

• What will be wrapped first, so that channels stop depending on the old system directly?
• In what order will capabilities move, and what is the first small move that proves the approach?
• How will migrated balances, rates and schedules be checked, account by account?
• What will the parallel run compare, and for how long?
• Has the rollback been rehearsed end to end, with real data volumes?