The Question on the Board Agenda: Rip Out the Core or Work Around It
Every bank, lender or wealth platform running a decades-old ledger eventually faces the same choice. Replace the core in one program, or modernize it in pieces until the old system no longer matters? Vendors and consultants tend to pitch whichever answer they're set up to sell. This guide doesn't.
For most mid-sized institutions, progressive core banking modernization is the lower-risk path. It isn't cheaper by default, and it fails in its own way: you run two cores indefinitely, pay for both, and never switch off the old one. Below: how the progressive approach works, where the money goes, and when a full replacement is the better decision.
Looking to streamline vendor RFP selection?
Access pre-negotiated tier discounts through our independent institutional network.
Three Mechanisms Behind "Hollowing Out the Core"
To hollow out the core is to move functions off the legacy system one by one until what remains is small enough to retire or freeze. Programs usually combine three mechanisms:
- An API and integration layer. A layer in front of the legacy core means channels, partners and new products call clean APIs and never touch the mainframe or old database directly. It decouples things. It doesn't modernize anything by itself.
- Moving logic outward. Pricing, product configuration, limits, fees and customer data move into separate services or specialist platforms. The old core shrinks toward a ledger of record.
- A sidecar core. A second, modern core runs next to the legacy one. New products launch on it first; existing books move later.
The integration layer almost always comes first. Without it, each product you move creates fresh point-to-point integrations, and those are the main reason progressive programs stall.
How a Sidecar Core Works Day to Day
With sidecar core banking, the new core holds the authoritative ledger for the products assigned to it, and the legacy core keeps everything else. Customers often hold products on both, so shared services have to span the two: the general ledger feed, customer master, payments hub, regulatory reporting and fraud monitoring.
Banks usually start with a product that has few dependencies, such as a digital savings account, a new SME lending proposition, or a brand for a new market. Cloud-native cores from vendors such as Thought Machine, Mambu and 10x Banking are often used this way. Temenos, Oracle FLEXCUBE, Infosys Finacle and Fiserv's Finxact also support coexistence models. Vendor availability, local partners and data-residency options differ by region, so check delivery capacity in your own country before shortlisting anyone.
The detail that matters most is ownership of the customer record and the general ledger. If you can't answer "where is the single customer record?" and "how do both cores post to finance?" before you sign, the sidecar will create reconciliation work that never goes away.
Sequencing a Product-by-Product Migration
Progressive core modernization works when products move in order of rising complexity and each move switches something off. The sequence below is common, but your order should follow product dependencies and regulatory deadlines.
| Wave | Typical products | Why this order | Indicative duration |
|---|---|---|---|
| Foundation | API layer, customer master, GL integration | Every later wave depends on it | 6–12 months |
| Wave 1 | New digital deposits or a greenfield brand | No legacy data to migrate | 6–9 months |
| Wave 2 | Retail savings and current accounts | High volume, relatively simple products | 9–15 months |
| Wave 3 | Consumer and SME lending | Collateral, collections and provisioning links | 12–18 months |
| Wave 4 | Mortgages, corporate, trade, Islamic products | Complex, long-dated, heavily customized | 12–24 months |
These durations are illustrative and overlap in practice. The rule that matters: each wave needs a decommissioning milestone, such as a batch job retired, a module licence dropped, or a legacy screen turned off. A wave that only adds things isn't hollowing out anything.
Looking to streamline vendor RFP selection?
Access pre-negotiated tier discounts through our independent institutional network.
Where the Costs Actually Pile Up
Progressive programs look cheaper because the spend is spread over years. The full bill is often comparable to a replacement; what differs is when it lands and what it buys. The figures below are typical 2026 market ranges for a mid-sized institution. They vary widely by contract, region and scope, so confirm them with vendors and integrators.
| Cost driver | Indicative range (USD) | What inflates it |
|---|---|---|
| New core licence or SaaS subscription | $0.3M–$3M per year | Account-based pricing as volumes migrate |
| Legacy core licence and support kept alive | $0.5M–$5M per year | Vendor minimums that don't shrink with volume |
| Integration and API layer build | $1M–$8M | Point-to-point legacy interfaces |
| Systems integrator fees per product wave | $1.5M–$10M | Customization and data-quality remediation |
| Dual-running operations and reconciliation | $0.5M–$3M per year | Manual breaks between two ledgers |
Watch the second line most closely. Legacy vendors often price on minimum commitments, so moving 60% of accounts may cut your bill by far less than 60%. Get step-down terms in writing before the program starts.
Red Flags That a Progressive Program Is Stalling
Most failed modernizations don't blow up. They stall, and the bank pays for two cores permanently. Watch for these signals:
- Two or more waves delivered with no legacy module, batch job or licence retired.
- Reconciliation headcount rising rather than falling.
- Regulatory reports assembled by hand from both cores at every period end.
- New product launches still waiting on legacy changes, which means the API layer isn't decoupling anything.
- Migration of complex books (mortgages, corporate lending) postponed year after year.
- No named executive accountable for the switch-off date.
If three or more apply, re-baseline the program. Moving faster with the same plan won't fix it.
Regulatory and Data Rules That Shape the Choice
Supervisors don't generally prefer one approach over the other. They focus on operational resilience, outsourcing and data, and running two cores widens the scope of all three. In the EU, the Digital Operational Resilience Act generally treats a core vendor or cloud provider as an ICT third party, with register, testing and exit-plan obligations. UK rules on outsourcing and operational resilience expect important business services to be mapped across both systems. In India, the Reserve Bank's IT outsourcing and payments data localization requirements can affect where a SaaS core may run. Singapore's MAS technology risk guidelines, and regulators such as SAMA in Saudi Arabia and the Central Bank of the UAE, set their own cloud and residency expectations.
Budget time for supervisory engagement before each wave, not only the first. Requirements differ by jurisdiction and licence type, so confirm your obligations with qualified regulatory counsel and your supervisor before committing to a timeline.
Signs a Full Replacement Is the Better Call
The debate over core banking replacement vs modernization is often framed as caution against recklessness. That framing is wrong. In some situations a single, well-run replacement carries less risk than years of coexistence:
- The legacy vendor is ending support, or the skills needed to maintain the system are close to gone.
- Your product range is narrow. A lender, digital bank or building society with a handful of products gets little from multi-year coexistence.
- You're merging or demerging, so data migration and a new operating model are happening anyway.
- The legacy core has no realistic integration layer. Wrapping it would cost nearly as much as replacing it.
- The two-core period would run past five years. At that point, dual-running costs usually wipe out the advantage of the progressive route.
Replacements still fail on data migration and testing, so plan several full dress rehearsals and set clear rollback criteria.
A Decision Rule You Can Take to the Board
A hypothetical example: a regional bank with 1.2 million accounts runs a heavily customized core whose licence comes up in four years. It offers 40 products across retail and SME. Progressive modernization fits, but only if all retail products can move to a sidecar before that renewal, so the bank can renegotiate a much smaller legacy contract or exit it. If the plan can't show that, a phased replacement timed to the renewal is the more defensible case.
Apply the same test to your own situation. Choose progressive modernization if you can name a credible switch-off date for the legacy core within roughly three to five years and fund the integration layer first. If you can't, put a full replacement on the table. Next, ask shortlisted vendors and integrators for a decommissioning plan with dated milestones and legacy cost step-downs. Weigh those answers more heavily than their feature demos.