A banking personalisation engine receives transaction and customer events from the bank’s systems, enriches and categorises them, decides in real time which action is right for each customer, and delivers that action through the bank’s own app, web, notification and chatbot channels. Integration typically takes three to six months from kick-off to go-live. Most of that time goes on connecting transaction feeds, building the front end and testing. After go-live, business teams can launch new content in hours without another integration project.
Key takeaways
- A personalisation engine for banks ingests events, enriches them, decides the next best action, delivers it through existing channels and learns from how customers respond.
- It sits between core systems and digital channels, so it does not replace the core.
- A typical first release takes three to six months, covering integration, a first set of use cases and testing.
- Starting with a few use cases tied to one outcome, such as activation, card usage or deposits, gets a bank live faster than a big-bang programme.
- Marketing automation runs campaigns well. Deciding what each customer should see from their financial behaviour is the personalisation engine’s job, and the two work together.
Why banks add a personalisation engine
Banks are not short of data. They are short of a way to act on it while the moment still matters. In research by FinTech Futures and Moneythor, 61% of banking professionals said they cannot track the customer journey accurately, 55% said they cannot deliver personalised experiences at scale, and 49% said they cannot sustain engagement beyond onboarding.
A personalisation engine closes that gap. It turns events the bank already records, such as a salary credit, a card payment or a subscription renewal, into a personalised insight, nudge, offer or reward while it is still relevant.
How a banking personalisation engine works, step by step
1. A signal arrives
The engine receives transaction feeds for accounts and cards, together with customer and product data, from the bank’s systems of record. Feeds can arrive in real time through event streaming or an API, or in batches. Open banking data from other institutions can be added where regulation allows.
2. The data is enriched and understood
Raw transaction descriptions are cleaned and categorised, merchants are identified with names and logos, and recurring patterns such as salary, bills, premiums and digital subscriptions are detected. Moneythor’s financial data enrichment combines customer-specific learning, rules on merchant category codes, transaction codes and descriptions, optional machine learning models and third-party classification services.
3. The engine decides the next best action
Rules and AI models assess each signal against the customer’s context and the bank’s priorities, then pick one action, or decide to do nothing. The bank defines the logic: eligibility, thresholds and which products or rewards to offer. In Moneythor, business teams configure this in Moneythor Studio, starting from more than 150 pre-built use cases or designing their own.
4. A personalised experience is delivered
The action reaches the customer through the bank’s channels: an insight card in the app, a push notification, an email or SMS through the bank’s existing notification system, or a reply in a chatbot. Rewards and cashback can be fulfilled in real time.
5. The engine learns
Customer interactions with each insight and notification are tracked and fed back, so the bank can see what works and adjust the logic. Moneythor’s AI for banking teams covers how predictive, generative, conversational and agentic AI support each of these steps.
Where the engine sits in the bank’s architecture
A personalisation engine is a layer between core systems and digital channels. It does not replace either.
| Connects to | How |
|---|---|
| Core banking and card systems | Core-agnostic ingestion of accounts, cards and transactions via HTTP/JSON API, or Kafka, MQ or similar streaming, in real time or batch |
| Mobile and web apps | A documented front-end API returning JSON data and HTML5 Web Components, usable from iOS, Android, Flutter, React Native and web |
| Notification and CRM systems | Alerts and nudges delivered by push, email, SMS or in-app through the bank’s existing infrastructure |
| Open banking sources | Multi-bank account and card aggregation |
| Bank analytics | The bank’s own analytical and machine learning models can be called by the engine |
| Staff access | LDAP, Active Directory and SAML single sign-on |
Moneythor can be deployed as SaaS, in a private cloud, on AWS, Azure or GCP, or on-premise, and supports Kubernetes, Docker and OpenShift. All customer-facing features are white-labelled and built by the bank’s teams on the Moneythor API, so the bank keeps control of the experience. The Moneythor platform page sets out the integration and deployment options in full.
How long does it take a bank to integrate a personalisation engine?
Most banks go live within three to six months of project kick-off. The table below shows an indicative plan. Actual schedules are agreed in a detailed planning session at the start of each project.
| Phase | What happens | Indicative timing |
|---|---|---|
| Planning and workshops | Agree scope, roles and the first use cases; refine requirements with the business team | Month 1 |
| Installation and configuration | Set up the engine in the bank’s test environment | Month 1 |
| Back-end and front-end integration | Bank’s teams connect transaction feeds and build the customer-facing features on the APIs, with vendor support | Months 1 to 4 |
| Data preparation and configuration | Analyse the bank’s test data, configure categorisation and the first recommendation rules | Months 2 to 4 |
| Testing and knowledge transfer | System testing, UAT, security testing, and train-the-trainer sessions so business staff can run Studio | Months 3 to 5 |
| Go-live and monitoring | Production set-up, readiness checks, roll-out, then a month of post-go-live monitoring | Months 4 to 6 |
A focused first release, such as a set of PFM insights, can be planned at about four months. Larger scopes with several channels or many use cases sit towards six.
What makes integration faster
- Start with a small number of use cases tied to one outcome. Activation after onboarding, card usage and savings deposits are common starting points.
- Prepare transaction feeds early. Categorisation and the first rules are configured from the bank’s test data, so the sooner that data is available, the sooner configuration starts.
- Use pre-built use cases where they fit, rather than designing every journey from scratch.
- Name an owner on the bank side for integration and another for the business use cases.
- Decide the deployment model early, so infrastructure and security reviews run in parallel.
What happens after go-live
The first integration is the slow part. Once the engine is connected to transaction feeds and channels, new use cases reuse the same connections. Business teams can launch new content in hours in Moneythor Studio, using Web Components, without going back to engineering.
Core banking module or independent personalisation layer?
Many core banking and digital banking vendors include some personalisation features. For a mid-sized or large bank, the question is whether those features can act on enriched transaction data in real time and be changed by the bank’s own teams.
A module inside the core or channel platform is convenient when the bank’s needs are simple and the vendor’s roadmap matches them. An independent personalisation layer makes more sense when the bank wants to:
- personalise across several channels and products, including ones outside the core vendor’s platform
- change insights, offers and rewards without waiting for the core vendor’s releases
- add loyalty, gamification and referrals on the same engine
- keep the option of changing its core or channel platform later without rebuilding personalisation
Because Moneythor is core-agnostic and sits between systems and channels, a bank can add it alongside the platforms it already runs. Our comparison of digital banking platforms looks at this choice in more detail.
Personalisation engine or marketing automation?
Marketing automation tools are good at campaign execution: scheduled workflows, audience lists and outbound communications. They typically rely on predefined rules and scheduled workflows.
A banking personalisation engine reads financial behaviour in real time and decides what each customer should see. The two work well together: the personalisation engine identifies the moment, the message and the action, and the marketing platform can handle downstream campaign execution. Moneythor’s article on why marketing automation alone falls short for financial institutions explains the difference in more detail.
What results look like
Banks see results when personalisation is tied to a commercial outcome from the start.
- Trust Bank in Singapore, which runs personalised engagement, rewards and referrals on Moneythor, acquired 70% of its customers through referrals. In the bank’s words, its Moneythor-powered real-time reward redemption programme “drives our groundbreaking referral program and keeps our acquisition costs 7x lower than the competition”. Trust Bank passed S$1 billion in deposits in May 2023, eight months after its September 2022 launch.
- At Chiba Bank in Japan, 79% of users who opened a campaign message completed the full engagement journey.
- A global tier 1 bank using Moneythor’s personalised insights reported 25% more active customers and 22% fewer support calls.
In Moneythor client reporting, personalised insights reach 15% to 30% engagement, against 1% to 2% for a traditional in-app banner.
Frequently Asked Questions
Typically three to six months from kick-off to go-live, depending on the number of use cases and channels. A focused first scope, such as PFM insights, can be planned at about four months. New use cases after go-live can be launched much faster because the integration is already in place.
The bank’s transaction data is enriched and categorised, a decision engine picks the next best action for each customer based on their behaviour and the bank’s rules, and the app displays the result as an insight, nudge, offer or reward through the engine’s front-end API.
No. It sits between the core and the digital channels, ingests data through APIs or event streams, and delivers actions back through the bank’s existing channels.
A core module suits simple, single-channel needs. An independent layer suits banks that want real-time decisions on enriched transaction data, changes made by their own business teams, and loyalty, gamification and referrals on one engine.
