Name: Behavioural Banking
Brand: Digitata

The premise

Segments describe. Behaviour predicts.

Two customers with the same age, income and product holding can behave nothing alike. One saves on a fixed schedule and never checks a rate. The other holds a cash buffer, panics on a failed payment and leaves after one bad experience.

Demographic segmentation cannot tell them apart - so banks treat them identically and lose both. Behavioural Banking scores the difference from transaction data the institution already holds, and makes it usable by the people who run retention, pricing and product.

Undifferentiated service

The same message, the same offer, the same tone to customers with opposite motivations.

Late churn detection

Attrition is visible in monthly reporting only once the customer has already gone.

Campaign guesswork

Blanket discounting buys volume from customers who would have stayed anyway.

Unused data

The richest behavioural dataset in the institution sits inside the core, unexploited.

Feature 01 - The engine

The CBS Engine - five dimensions from transactions alone

The Customer Behaviour Scoring engine is an XGBoost multi-output regressor trained across 16 engineered features. It scores five behavioural dimensions per customer, achieves R2 = 0.674 on core banking data, and produces an actionable profile within a 90-day transaction window.

Dimension 01

Planner

Detected by: fixed-schedule flows, low spend variance, standing orders, high savings consistency.

What it means: highest loyalty potential. Responds to scheduled products, rate locks and tier upgrades. Very low churn risk unless service fails.

savings_cv - auto_transfer_ratio - savings_rate

Dimension 02

Spontaneous

Detected by: post-income bursts, end-of-month spikes, late-night activity, present bias.

What it means: responds to time-limited offers and urgency triggers. Volume is unpredictable - liquidity and limits must flex.

payday_decay_rate - night_spend_ratio - eom_spike_ratio

Dimension 03

Security Seeker

Detected by: large cash buffers, spending pullback after balance drops, insurance spend above actuarial value.

What it means: highest churn risk from any service failure. Needs guarantees, real-time status and visible service recovery.

balance_drop_response - emergency_buffer - insurance_excess

Dimension 04

Optimizer

Detected by: provider switching, cross-account transfers, active rate comparison, financial sophistication.

What it means: hardest to retain on price. Value must come from speed, limits, preferred rates and exclusive access.

provider_switch_rate - cross_account_xfer_rate

Dimension 05

Experiencer

Detected by: hedonic spend, subscription density, premium brand affinity, broad category entropy.

What it means: high ARPU potential. Responds to premium tiers, concierge service and relationship management.

experience_spend_ratio - subscription_density - brand_premium_index

Why it matters

Actionable in one quarter

Behavioural patterns emerge within 90 days of transaction history. A new customer can be classified and served on strategy before churn risk has a chance to solidify - and before a competitor's onboarding offer lands.

Feature 02 - The profile

Customer 360 - a living profile, not a CRM record

Every customer carries a profile that combines behavioural scores with lifecycle and engagement context, and updates continuously as new transactions arrive.

Behavioural scores across all five CBS dimensions, on a rolling window.
Transaction patterns - frequency, value, timing, payday alignment, counterparty consistency.
Life-stage signals - new customer, settling, anchor relationship, multi-product, disengaging.
Digital engagement - app usage, channel preference, digital adoption score, enquiry behaviour.
Psychographic axes - Confidence vs Anxiety - Stability vs Stress - Readiness vs Hesitation - Trust vs Doubt.
Churn signals - declining frequency, abandoned journeys, rate lookups without conversion, support contact rate.

Customer 360 - behavioural scores, lifecycle context and churn signals in a single profile.

Behavioural analytics - dimension distribution across the base and movement over time.

Feature 03 - The psychology

Four psychographic axes behind the numbers

Scores tell you what a customer does. The psychographic layer tells you the disposition to speak to - the difference between an offer that reassures and one that pressures.

Confidence vs Anxiety - how much certainty the customer needs before acting.
Stability vs Stress - whether the financial picture is steady or under strain.
Readiness vs Hesitation - proximity to a decision on a product or a switch.
Trust vs Doubt - the standing of the relationship, and how much a single failure will cost.

Each axis is scored from behaviour and feeds message tone, channel and offer construction - not just targeting.

Feature 04 - Early warning

Defection Monitor - risk surfaces in hours, not quarters

The churn risk monitor tracks defection signals across every segment in real time, isolates the trigger event behind a risk surge, and quantifies the revenue in motion before it leaves.

Segment risk scoring

Churn probability per segment with week-on-week movement, so you see which cohorts are shifting and how fast.

Trigger identification

Competitor promotion, service degradation, rate movement, payment or payout failure - the cause behind the surge, named.

Geographic heat maps

Where defection is concentrating - branch, region or market - for operational response.

Revenue at risk

Annual value in motion per segment, so intervention budget goes where the money is.

Priority classification

High, medium and low bands focus operations on the highest-value cohorts first.

Failure correlation

Churn risk spikes automatically correlated against incidents in the previous 72 hours.

Churn signal	What it indicates
Activity frequency decline	Weekly to fortnightly to monthly - the classic pre-churn trajectory
Rate enquiry without conversion	The customer is checking and comparing with a competitor
Relationship shift	Balances or flows moving to a secondary institution without an obvious reason
Support contact spike	Query or complaint volume rising on a product line, ahead of segment-level churn
Failure correlation	Risk rising sharply among customers who experienced a failed payment or outage

Feature 05 - Targeting

Smart Segment Builder - precision without engineering

Retention and CX teams build cohorts themselves, combining static attributes, calculated behavioural metrics and ML scores, with live size and revenue-at-risk preview before anything launches.

High-value planners - Planner score > 70 with consistent high-value flows. Rate lock and loyalty tier.
New relationships - under 90 days, activity increasing, single product. Speed and reliability messaging.
Rate shoppers - Optimizer score > 70 with repeated enquiries and no conversion. Fee waiver and preferential programme.
At-risk dormant - frequency down more than 30% over 60 days. Personalised win-back.
Security-seeker churners - Security Seeker score > 65 with a recent service failure. Service recovery.
Multi-product anchors - two or more active products, tenure over 18 months. Premium tier and priority support.

Smart Segment Builder - rules, live cohort size and revenue exposure in one view.

Campaign engine - playbooks, delivery channel and performance per behavioural cohort.

Feature 06 - Intervention

Campaign engine with ready-built playbooks

Identifying risk is half a loop. The campaign engine closes it - personalised outreach generated from the behavioural profile, delivered on the channel the customer actually uses.

Rate lock offer - for Optimizers and Security Seekers showing risk.
Schedule activation - for Planners not yet on a recurring arrangement.
Service recovery - for Security Seekers after a failure, with a real gesture attached.
Seasonal pre-offer - ahead of known spending peaks for seasonal cohorts.
Win-back - for dormant customers, with an incentive sized to their value.
Onboarding series - for new relationships inside the critical first 90 days.
Premium tier upgrade - for high-value customers past a tenure threshold.

Feature 07 - Agentic AI

The autonomous retention loop

The agentic layer runs the full retention workflow continuously - scanning for emerging risk, generating a personalised intervention per customer, and submitting it for delivery, with an optional human gate before anything reaches a customer.

Continuous scan

Not a monthly batch. Risk is identified within hours of a behavioural shift, across the entire base.

Per-customer generation

Offer, channel, tone and timing are constructed against that customer's CBS profile and psychographic axes.

Human-in-the-loop gate

Operations review and approve AI-recommended campaigns before launch - or set thresholds below which the loop runs unattended.

Multi-channel delivery

App push, SMS, WhatsApp, email, USSD and in-branch prompts - whichever the customer's engagement profile favours.

Performance capture

Delivery, redemption, churn reduction and revenue preserved measured per campaign and fed back into the models.

Feature 08 - Lifecycle

Where the platform acts across the relationship

Stage	What it looks like	How Behavioural Banking acts
New - days 1-90	First transactions. Trust being built. Single product.	Onboarding series, reliability messaging. CBS profile forms and becomes actionable inside the quarter.
Active - months 3-12	Regular activity. Habits forming. Behaviour readable.	Habit reinforcement, rate programmes, loyalty. Segment assigned and strategy applied.
Anchor - 12 months+	Multi-product, high lifetime value, deepening relationship.	Premium tier, cross-product expansion, referral and relationship management.
At risk - signal detected	Frequency declining, comparison behaviour, support contact.	Immediate personalised intervention; service recovery where a failure triggered it.
Recovered	Reactivated and re-engaged.	Returned to the Active or Anchor track, with campaign ROI captured.

Feature 09 - Deployment

Built to sit alongside the core, not replace it

Core-agnostic ingestion

Reads transaction, account and product data from existing core banking systems, data warehouses or an open banking feed.

Open banking wedge

Where core integration is slow, a consented open banking feed produces behavioural scoring on a live customer base as a low-friction first step.

Zero development risk

The platform is productised and in production, not a bespoke build. Pilots run on real data inside a defined window.

Compliance-ready

Built for GDPR, DORA and EU AI Act obligations: explainable scoring, data residency, retention controls and auditable decisioning.

Explainable by design

Every score decomposes into the features that produced it - defensible to a risk committee and to a regulator.

Land, implement, standardise

A staged commercial model: paid pilot on real data, phased implementation, then platform standardisation across the base.

The business case

What a point of retention is worth

The value sits in the gap between a first-year customer and an anchor relationship - and in the cost of the intervention that converts one into the other.

"Personalised outreach is the most effective retention lever measured on the platform - 83% effectiveness, against 42% for generic educational content."
Platform intervention data

Intervention effectiveness by type: personalised outreach 83% - loyalty rewards 75% - discount offers 62% - tier upgrades 58% - generic content 42%. The 41-point gap is the value of knowing who the customer is before you speak to them.

83%

Personalised outreach effectiveness

42.7%

Typical win-back redemption rate

90 days

To an actionable behavioural profile

41pp

Gap between personalised and generic intervention

Your customers already told you who they are.