Hyper-personalisation at IDFC First Bank

Improve cross-selling conversion and engagement through omni-channel personalisation

TL;DR:

My involvement with IDFC First Bank's personalisation project was one of the most challenging and satisfying experiences of my career. Initially focused on the data platform, my role expanded to encompass the entire personalisation journey.

Key highlights include:

IDFC First Bank, formed in 2018, focused on digital-first banking to compete with established players.
Goal: Maximise customer lifetime value through increased product ownership per customer.
Phase-1 start with us implementing the MVP of a cloud-based data platform for hyper-personalisation in 2019-2020.
After the success of phase 1, we developed a Customer Data Platform integrating multiple data sources for a 360° customer view.
We also implemented a sophisticated Next Best Action system for targeted cross-selling and up-selling.
We expanded the personalisation platform to include real-time offers, retargeting capabilities, and gamification campaigns.
We further advanced on the personalisation maturity curve and achieved omni-channel personalisation across digital and physical touch-points.
IDFC First Bank received multiple awards for digital banking innovation and customer experience.
We were powering personalised offer recommendations to ~20Mn+ customers using our personalisation engine.

This project showcased the power of data-driven personalisation in transforming a new bank's digital offerings and customer engagement strategies.

About IDFC First Bank:

IDFC First Bank, formed in 2018 by merging Capital First and IDFC Bank, is a leading Indian private bank. Post-merger, the bank aimed to differentiate itself by addressing gaps in competitors' retail banking services. Their strategy focused on revamping their digital platforms—net banking and mobile apps—to enhance customer experience. The bank's key objective was to leverage data analytics to drive innovation in their digital banking offerings.

Problem Statement:

The problem identified was to maximise the customer lifetime value through increase product ownership per existing customers.

Background:

IDFC First Bank faced stiff competition from established players like SBI, HDFC Bank, and ICICI Bank, which had extensive user bases and physical presence across India. As a new entity, IDFC First Bank recognised that traditional expansion through brick-and-mortar branches would be time-consuming and capital-intensive.

Instead, the bank identified a strategic opportunity in targeting millennials—a demographic seeking digital-first banking solutions that minimise the need for branch visits. This insight led IDFC First Bank to focus on revamping its digital offerings, aiming to provide comprehensive banking services via user-friendly mobile and online platforms. This platform would provide a unified experience to the then IDFC Bank and Capital First customers.

As a young institution, IDFC First Bank aimed to become the primary bank for its retail customers, focusing on building loyalty, higher engagement and maximising customer lifetime value. The bank's strategy was based on a key insight: increasing product ownership per customer would likely lead to stronger retention and primary bank status. By encouraging customers to adopt multiple banking products, IDFC First Bank sought to deepen relationships and secure its position as their preferred financial institution.

In 2019-20, IDFC First Bank pioneered the concept of hyper-personalisation in Indian banking, where most institutions were still at the lower end of the personalisation maturity curve. The bank's goal was to boost conversions for cross-selling and up-selling additional banking products to existing customers.

IDFC First Bank had access to significant transactional data of customers that were inherited from both IDFC Bank and Capital First post-merger. The challenge was to effectively leverage this vast data repository to enhance cross-selling offer conversions, setting a new standard in personalised banking services in India.

My Role:

In Aug 2019, IDFC First Bank approached Thoughtworks to build a cloud based data platform that would power the personalisation engine. The team comprised of data engineers, quality assurance engineers, product manager, data analysts, data scientists, ML engineers and UX designers. My role was to lead this cross-functional team to deliver the data platform and build customer 360 attributes that can be used by the machine learning model to curate personalised cross-sell and up-sell offer recommendations across digital channels.

What is Hyper- Personalisation?

Let me give a quick summary around hyper-personalisation and the personalisation maturity curve.

Hyper-personalisation is an advanced marketing strategy that leverages real time data, machine learning/artificial intelligence to deliver highly relevant experiences to the customers.

The personalisation maturity framework talks about the different levels of personalisations that are possible. As we go higher up the personalisation curve, higher is the lifetime value.

One size fits all messaging: This is where there is no personalisation at all. Every customer will get a standard digital experience.
Field insertion: At this level, organisations still provide the standard experience to all the customers. However they start using dynamic fields like the name of the customer to add a dash of personalisation.
Rule based segmentation: This is the level where there is some kind of personalisation in the digital experience. The orgs divide the existing customer base in to smaller segments based on a set of pre-defined attributes and rules
Behavioural Marketing/Triggers: This is where we started IDFC's journey. At this level, orgs will start leveraging data to identify some key behaviours and trigger personalised notifications.
Real-time personalisation: The difference between the previous stage and the current stage is the flow of data. At this level the marketing triggers are generated based on real time data and therefore real time behaviours.
Optimised Omni-channel: At this level, the organisations provide highly personalised experience across multiple channels like App/Web platforms, Email, SMS, Push notifications etc.
Predictive Personalisation: This is the most advanced level of personalisation which leverages predictive analytics and machine learning to determine behavioural triggers even before they occur. It is believed that Netflix has achieved this highest form of personalisation where they predict the kind of content a customer will like with very high accuracy.

Personalisation Maturity Curve, Source: Yieldify

As mentioned above, we started IDFC First's personalisation journey targeting the behavioural marketing/triggers. To reach the advance stages of personalisation, one needs to collect a lot of behavioural data. Since we were building the new digital banking platforms from scratch, we didn't have access to clickstream/product usage data.

The idea of the personalisation platform was split into multiple layers as mentioned below:

Data Layer: The first layer was known as the data layer that consisted of raw customer data from the internal and external sources. The raw data was converted into customer 360 attributes that form the foundation of behavioural data
Analytics Engine: The second layer was called the analytics engine. This was the brain of the personalisation. The jobs to be done for the analytics engine were to select eligible customers, curate offers and choose the best channel to deliver the offer via a process called as the 'Next Best Action' which I have covered in the later part of the article. The output from the analytics layer was used to trigger offer communication via the respective channels.
Interaction Layer: The third layer was the interaction layer where the customers could actually view, click and avail the offer delivered to them via selected channels. The interaction data for all of the offers were captured and sent back to the data layer for further processing. This is what we called as a feedback loop in personalisation.
Feedback Loop: The interaction data + conversion data formed the feedback loop that would be very crucial for the machine learning model to curate cross-sell, up-sell offers based on propensity to convert.

Layers powering personalisation platform

Approach:

Phase 1- Building the Customer Data Platform:

After multiple discussions, we had aligned stakeholders on the technical architecture. We had identified cloud partners and prepared cloud instances where the data would be ingested and processed.

As next steps, we ran a lean workshop with the relevant stakeholders to derive the following outcomes:

Understand the different existing data sources.
Identify the different data attributes and their context.
Identify the sensitive information present in these systems.
Analyse the integrations to ingest the data from these sources.
Understand the refresh frequencies of these data sources.
Identify key data owners in the bank to manage the approval processes like whitelisting of IP addresses and ports.

After the workshop, we identified a few challenges:

There were more than 100+ data sources that we would need to power the personalisation engine.
IDFC bank and Capital First both of these entities had different mechanisms to collect and store data.
Most of the data sources didn't have a readily available data dictionary.
The refresh frequencies of these data sources were completely different.
One integration method for all the source would not work as we would receive data in different formats: CSV/Excel files over SFTP, Query the MySQL/Oracle DBs and so on.
Handling lot of PII information before storing it on cloud.

The third step was to prepare the list of customer 360 attributes that would be needed to identify key customer behaviour to curate personalised offers. The bank's focus was on existing retail banking customers, there we analysed the existing retail banking product portfolio that the bank has to offer. They are listed as mentioned below:

Liabilities: savings accounts, fixed deposits, term deposits, salary accounts and others.
Assets: home loans, personal loans, automobile loans and others.
Wealth: mutual fund and insurance
Credit Cards: new product offering from the bank.

MVP of Personalisation Engine:

For each of these products we had to brainstorm with the banking SMEs, data analysts to identify:

Demographic data: Gender, marital status, income, age, location, age of the relationship with the bank etc.
Liabilities: Average weekly/monthly balance, spend categories like medical, housing, movies etc.
Assets: Active loan amount, tenure of the loan amount, credit score, other active loans, repayment history etc.
Wealth: Total amounts invested in mutual funds, total returns, total sum insured, total annual premium, types of policy covers etc.
Credit Cards: Spending patterns, available credit limits, average monthly credit utilisation, repayment history and others.

Based on multiple discussions with the stakeholders, we were able to prioritise liabilities product portfolio for the first pilot phase. As next steps, we identified the raw data sources required to be ingested in the cloud platform so that we could start building the customer 360 attributes for Demographic and Liabilities products.

The data ingestion module consisted of:

Scheduler
Data orchestration: the workflow related to data ingestion:
- data cleansing: rectify date formats
- validation: data quality checks to ensure not ingesting records with missing critical fields like date of birth, account number etc.
- encryption and anonymisation of sensitive customer information like account number, name, phone number, address etc. This was in reference to the
Storage in cloud: creating a data schema and data crawlers to ensure that the data can be queries using SQL queries.
Alerts and monitoring: data ingestion/validation failures.

We also identified the list of raw data sources that would have to be ingested in order to create the customer 360 attributes.

Further, we piloted the next best action journey to curate offers as mentioned below:

The campaign manager and the banking product leads identified a couple of pilot cross-selling offers that can be rolled out for the pilot phase.
We also randomly identified a control group and test group. The control group customers would receive the non-personalised standard communications. The test group would receive the personalised communication
We got a list of eligibility criteria that helped us select the eligible customer base from the test group using the customer 360 attributes.
Since the digital platforms were not production ready therefore using banners were not feasible, we chose to launch the pilots with email and SMS channels.
At the next step, we filtered out the customers who had either unsubscribed to email (from the previous campaigns) or their number was registered for do not disturb in TRAI database.
The campaign was a weekly campaign where we had scheduled the email and SMS communications for the customers. The communication were triggered via Salesforce marketing cloud
The interaction data was captured and sent back for processing.
We analysed the interaction data from test group against the interaction data from the control group.
The process was repeated for a couple of cycles.
At this stage, we were not using the machine learning models to curate the offers, we were collaborating with the data analyst and data science team to mock the process.

Outcome: The initial MVP results were positive and this is how we got the go ahead to build the sophisticated personalisation platform that we had envisioned.

Building the Personalisation Platform 2.0:

Customer Data Platform 2.0:

IDFC's Customer Data Platform 2.0 roadmap aimed to enhance personalisation through comprehensive data integration. The strategy identified five key data categories to feed behavioural triggers into the machine learning model:

Demographic: Personal information from account opening and CRM interactions
Psychographic: Spending patterns and active loans reflecting interests and lifestyle
Transactional: Credit and debit account activities
Behavioural: Clickstream, product usage, customer support interactions
External Sources: Credit bureau data on scores and inquiries with other banks

These diverse data streams were designed to be modelled into a holistic customer 360 view, encompassing attributes across Liabilities, Assets, Wealth, and Credit Cards. This approach aimed to enable more targeted and effective personalisation efforts, leveraging a comprehensive understanding of each customer's profile and activities.

Integrating more digital channels:

The success of the pilots also gave a go-ahead to integrate multiple channels, in the phase 2, we were looking at the below mentioned digital channels:

Whatsapp
Email
SMS
Rich Communication Service (RCS)
Push Notifications
Web Banners
App Banners

Next Best Action 2.0:

Next Best Action 2.0 at IDFC operated as a batch process, not utilizing real-time data. The system followed a structured workflow:

Campaign managers defined customer criteria
The platform filtered out do-not-disturb registrations
The platform also removed previously interacted offers
Further, the platform eliminated conflicting offers
The machine learning model within the platform applied propensity scoring using machine learning models
Implemented a governance layer to manage communication frequency across:
- Offer types (Cross-Sell/Upsell, Engagement, Third Party)
- Channels
The offers were prioritised based on propensity scores and predefined channel and offer limit quotas
The final payload was sent to Salesforce Marketing Cloud and Adobe Analytics for:
- Multi-channel distribution (SMS, Email, Push Notification, RCS, WhatsApp)
- Banner displays

This process ensured targeted, relevant communications while respecting customer preferences and channel limits.

Real-time Offers:

We enhanced IDFC's personalisation platform with real-time offer capabilities, utilizing dynamic data sources like customer transactions to identify behavioural triggers. Working with campaign managers, we developed targeted use cases, such as recommending travel insurance after an international flight ticket purchase. Our strategy focused on two key areas: activating push channels (email, SMS, push notifications, WhatsApp, RCS) for instant offer delivery, and implementing dynamic banner displays in mobile and net banking platforms. This real-time approach significantly improved our ability to provide timely, contextual offers, enhancing both customer experience and cross-selling opportunities in digital banking environments.

Retargeting Capabilities:

We developed a custom retargeting system for IDFC's personalisation platform, drawing inspiration from Google Ads but tailored for financial products. This system tracks customer interest in cross-sell offers and strategically reintroduces these offers at regular intervals. By maintaining engagement without overwhelming customers, it focuses on personalised, relevant retargeting. The goal is to increase conversion rates over time by allowing IDFC to re-engage customers with offers they've previously shown interest in, potentially improving cross-sell success rates in a targeted and efficient manner.

Gamification Campaigns:

IDFC First Bank's Personalisation 2.0 platform, enhanced with innovative gamification capabilities, became a game-changer in digital banking engagement during the COVID-19 lockdowns of August 2020. This cutting-edge feature enabled the bank to design and execute day-wise gamified campaigns, rewarding customers with vouchers upon completion of specific tasks.

Seizing the moment when physical branches were inaccessible, IDFC launched a groundbreaking 7-day campaign featuring Bollywood icon Amitabh Bachchan. The initiative challenged customers to perform daily digital banking tasks, from opening savings accounts to making UPI payments, with each milestone earning Amazon vouchers. This strategic move not only promoted IDFC's new mobile and net banking platforms but also significantly boosted digital savings account adoption. The campaign's resounding success catalysed its rapid replication across business and NRI banking sectors, marking a pivotal moment in IDFC's digital transformation journey and setting a new industry benchmark for customer engagement. This powerful demonstration of the platform's gamification capabilities underscored its potential to revolutionise digital banking interactions, proving particularly valuable during unprecedented times and paving the way for future innovative engagement strategies.

Outcome: With the one of the highly successful gamified savings account campaigns, the daily new savings account increased by ~70%.

Omni-channel Personalisation:

IDFC First Bank's next leap in personalisation maturity was the implementation of a sophisticated omni-channel strategy, designed to create a uniform experience across both digital and physical touchpoints. This comprehensive approach extended personalised offers beyond traditional digital channels like email, SMS, push notifications, WhatsApp, and rich communication services to encompass offline interactions with relationship managers, branch executives, customer support representatives, and even ATM machines. High net-worth clients managed by wealth managers received the same tailored experience as those engaging through mobile apps.

The innovation lay in enabling offline personnel to pitch personalised offers and capture responses, while ATM screens displayed targeted promotions to customers during routine transactions. Crucially, all interaction data, regardless of the channel, flowed back to the central customer data platform. This continuous feedback loop enriched the machine learning models, refining propensity-to-convert calculations and further enhancing offer relevance. By bridging the gap between online and offline channels, IDFC First Bank not only elevated customer experience to new heights but also created a data-driven ecosystem that continuously evolved and improved, setting a new standard for personalised banking in the digital age.

Achievements:

Awards:

Mentions:

Learnings:

Witnessed firsthand how a new entrant disrupted the traditional banking sector using technology.
Observed the effectiveness of identifying a core target segment, recognising gaps in existing offerings, and leveraging digital channels for acquisition and engagement.
Experienced the impact of data-driven personalised customer experiences in driving engagement and adoption.
Built a customer data platform capable of powering personalisation for 20+ million customers.
Implemented personalisation progressively, starting with rule-based triggers and advancing to omni-channel personalisation.
Collaborated with large, diverse stakeholder teams to align vision and roadmaps.
Built significant trust with key stakeholders, leading to independent decision-making authority. This is unlikely in a tech-consulting setup hence it was personally a huge achievement for me.
Gained practical experience in implementing omni-channel strategies, moving beyond theoretical knowledge
Gained insights into the underlying human psychology themes used in personalisation capabilities
Acquired comprehensive understanding of retail banking, including high-level journeys and data capture across multiple banking products
Observed the transformative power of technology in banking and the importance of data-driven, customer-centric approaches in creating competitive advantage.

Appreciation from one of my key stakeholders: