A digital banking platform costs considerably more than its licence. The real figure is the total cost of ownership (TCO): every one-off and recurring cost of acquiring, integrating, running, changing and eventually leaving the platform, counted over the same period for every option, normally five years. Compare quotes on that basis and the cheapest licence is frequently not the cheapest platform.
This guide contains no price ranges, because a range that fits a small bank with one retail app misleads a group with four segments and three markets. It gives you the model instead: twelve cost lines, what drives each, the question to put to a vendor, and a method for comparing quotes like for like.
Why is the licence only a minority of the real cost?
The licence is the one number every vendor is ready to quote, so it anchors the comparison. Whole-life costing exists to prevent exactly that. The UK government’s Should Cost Modelling guidance note, written for public buyers, states that the key factor is whole life cost and not the initial purchase price, and describes a cost model as an estimate of a specific set of costs under particular circumstances over a defined time period.
For a channel platform, the costs outside the licence come from three places: the work to connect it to your core, cards, payments and screening systems; the people and infrastructure needed to run it around the clock; and the cost of changing it every month for as long as it is live. As a rule of thumb, not a statistic, expect the licence to be a minority of the five-year total, and question any business case in which it is not.
What are the cost lines in a digital banking platform TCO?
There are twelve. A vendor proposal usually prices the first few and is silent on the rest, which is how two quotes for the same scope end up far apart. Put all twelve in your pricing template.
| Cost line | What drives it | Question to ask the vendor |
|---|---|---|
| Licence or subscription | Pricing metric, modules, channels, environments | What exactly is counted, and what is the price at twice our volume? |
| Implementation and configuration | Journeys, segments, languages, brands | Which journeys are standard product and which are project work? |
| Integration | Number and quality of core and third-party interfaces | Which connectors exist today, and who builds the others? |
| Customisation and change requests | Gap to requirements, change volume after launch | What can our team change without you, and at what rate otherwise? |
| Infrastructure and environments | Deployment model, resilience, data residency | What sizing do production, disaster recovery and test need? |
| Security and compliance | Penetration tests, audits, card-data scope | Who pays for testing and for fixing the findings? |
| Third-party services | Identity verification, SMS and OTP, push, maps | Which third-party fees are inside your price? |
| Internal team and operations | Product, engineering, testing, hours of cover | What team do comparable clients run? |
| Upgrades and regression testing | Release frequency, depth of customisation | Are upgrades included, and who retests our journeys? |
| Support and SLAs | Hours of cover, response times, severities | What does each support tier include and exclude? |
| Training and change management | Staff, branches, contact centre, customer migration | What training and documentation is included? |
| Exit and migration | Data export, transition period, parallel running | What exit help do you give, and at what rate? |
How does the pricing metric behave as your customer base grows?
Two quotes that match in year one can diverge sharply by year five, because the metric behind the price sets the shape of the curve.
- Per registered user: rises with every enrolment and keeps charging for dormant customers, so cost can climb faster than value.
- Per active customer: tracks real usage, but the contract’s definition of “active” (which actions, over which window) is worth as much as the unit price.
- Tiered bands: a staircase, flat inside a band and then a step, which may land in the middle of a budget year and may be applied retroactively.
- Per module: flat against customer numbers, but it steps up each time your roadmap adds a capability.
- Per transaction or API call: grows with engagement, which is the very thing you are trying to increase.
- Flat or enterprise: a level line, usually higher at the start and cheaper at scale.
- Perpetual licence with annual maintenance: a large one-off followed by a recurring charge calculated on the licence value.
One-off or recurring, capex or opex?
Split every line two ways. One-off costs are implementation, integration, initial customisation, data and customer migration, first-time training and, at the far end, exit. Recurring costs are the subscription or maintenance, infrastructure, third-party volumes, your own team, support, security testing and upgrades.
Whether a line is capitalised or expensed depends on your deployment model and accounting policy, and is a question for your finance team and auditors, not for the vendor. Compare options on cash over the full five years first, then show the accounting view separately, so that classification does not choose the platform for you.
Which hidden costs and contract terms should you watch?
Most of the difference between a quote and the eventual bill sits in clauses, not in the price table. Read these before you compare totals.
- Annual uplifts: the indexation formula, whether it is capped, and whether it also applies to service rates.
- Change-request rates: whether the rate card is fixed for the term, who estimates the effort, and any minimum billing.
- Upgrade obligations: how far behind the current version you may fall and stay supported, and who pays when an upgrade breaks your customisations.
- Environments: whether test, staging and disaster recovery environments are licensed separately.
- Volume true-ups: what happens, and from which date, when you cross a band.
- Source code escrow or access: the release conditions, who pays for the escrow, and whether your team could realistically build and run what is released.
Why are security testing and exit budget lines, not footnotes?
Security testing recurs for as long as the platform is live. The PCI Security Standards Council’s penetration testing guidance, an information supplement to PCI DSS, states that penetration testing must be performed at least annually and after any significant change, and gives an application upgrade as an example. Where your channels are in scope, a busy release calendar therefore generates testing and retesting cost.
Exit is the line most often left empty. The Basel Committee’s Principles for the sound management of third-party risk (December 2025) say banks should maintain exit plans for planned termination and exit strategies for unplanned termination, and that exit plans need to be regularly updated and tested for availability of budget, human resources, technical infrastructure, transfer of knowledge and access to data. SAMA’s Rules on Outsourcing for banks likewise say the contract should allow for renewal, renegotiation, default termination and early exit. If supervisors expect an exit budget, your TCO model should contain one.
How do you build a five-year TCO comparison, step by step?
The US Government Accountability Office’s Cost Estimating and Assessment Guide lists setting ground rules and assumptions, analysing sensitivity and risk, and updating estimates with actual costs among its key steps.
- Fix one baseline: channels, segments, journeys, markets, and customers, active customers and transactions for each of the five years.
- Write the assumptions in one log: growth, change volume per year, number of upgrades, team size.
- Issue one pricing template with the twelve lines by year, and decline quotes in a vendor’s own format.
- Fill in the lines vendors do not quote: your team, infrastructure, third-party volumes and testing.
- Normalise: same term, same currency, same inclusions, the contractual uplift applied to every year, taxes treated the same way.
- Add the in-house option as a comparator, costed on the same lines.
- Run sensitivities: growth at twice the plan, a heavy change year, a delayed go-live, and an exit in year five.
- Present the five-year total, the year-by-year profile and the cost per active customer, then update the model with actual costs after go-live.
How does build, buy or hybrid change the cost shape?
The sourcing choice moves cost between lines; it does not remove any. Building in-house has no licence, a high one-off cost and a permanent engineering team, and you carry upgrades, security testing and resilience as your own cost for the life of the system. Buying lowers the one-off and replaces it with a recurring fee, a change-request line and a real exit cost. A hybrid, in which you license a platform and your own team builds journeys on it, keeps a licence and a smaller team, and shrinks the change-request line only if the platform’s configuration tools genuinely let your people work without the vendor.
The UK guidance uses the same three shapes: an in-house comparator that includes the cost of acquiring and maintaining assets and the necessary capability, an expected market cost, and a mixed economy that combines the two.
How Eurisko fits
Eurisko has delivered banking software since 2009, with a team of 200+ engineers. Eurisko DXP is our licensed, versioned digital banking platform: it runs in front of any core, is deployed inside the bank’s own estate, and includes Journey Builder, which lets your product teams change journeys, products and rules through configuration and so addresses the change-request line directly. Releases keep APIs and configuration backward compatible and regression-test the journeys you built, which is the upgrade line, and what your teams build is yours and exportable in open formats, which is the exit line.
We are happy to go through the twelve lines against your scope. When we are not the right choice: DXP is not a core banking system, and because it is deployed on-premises or in your private cloud, a bank that wants to operate no infrastructure at all and simply rent a shared hosted service should look at a different model.