Skip to content
GullySales

Banks · Customer health scoring

Spot the salary account going quiet while the relationship manager can still call.

Customer health scoring for a bank combines signals the bank already holds, such as salary credits, card use, app logins and open complaints, into a ranked view of relationships needing a conversation. GullySales designs the rules, tests them against customers you already know left or stayed, and agrees what each score band prompts.

Scope, fee, reporting and term are agreed in writing after we have read your brief and spoken to you.

Realistic editorial illustration of banks at work in India

What customer health scoring means

Customer health scoring combines signals from existing customer records into a score that shows where a relationship needs attention before an order stops or a renewal is lost. GullySales builds the scoring rules, tests them against known customer outcomes, and gives your team a ranked view with a clear response for each score band.

Last updated 14 Nov 2026. Rules for this trade differ by state and professional body, and they change. We describe restrictions in general terms, so check the current position with your own council or adviser before you publish anything.

For banks

Where it usually goes wrong, and what we would do.

  • “A bank sees the exit last”

    A customer who moves a salary account does not announce it. The credit stops arriving, the balance falls, the card goes quiet and the app is opened less. Each fact sits in a different report, so the relationship manager hears about it when the standing instruction is cancelled or the account is closed. By then a rival bank has had the conversation.

  • “One signal alone misleads”

    A low balance after a school fee payment means nothing. A low balance together with a stopped salary credit and a lapsed card tells another story. Scoring that reads one number flags the wrong people. The method here weighs several signals together and shows the reasons beside each flag, so a manager can judge it.

  • “A score is not a credit decision”

    Using customer behaviour to decide who is called is different from using it to decide who gets credit or what price they pay. Mixing the two creates fairness and compliance problems. This score is built only to prompt a human conversation. Your risk, privacy and compliance teams decide which data may be used and where the score must never be applied.

  • “The branch has no time for a long list”

    A cluster head who receives a spreadsheet of every customer with a changed pattern will not act on it. A relationship manager can take a handful of calls in a session. The scoring ranks accounts and attaches one suggested action to each band, so the list the branch receives is short enough to use.

What we do

What we deliver for banks.

The work available for this need. We agree the deliverables and responsibilities for your engagement in writing.

  1. Signal inventory from bank data

    A list of signals the bank can actually produce, such as salary or regular credits, card use, app and net banking logins, standing instruction cancellations, open complaints and product holding. Each has a source, a refresh point and a note on reliability, and your data team marks which are permitted for this use.

    Result: You can see every signal, its source system and whether the bank has approved it.

  2. Scoring rules with visible reasons

    A written model showing how each signal moves the score and why. A customer flagged as at risk has the reasons listed beside the name, such as salary credit absent and card unused, so the relationship manager knows what to raise on the call and does not guess.

    Result: Any flag can be traced to the specific signals that produced it.

  3. Comparison with each customer's own pattern

    Where enough history exists, a customer is compared with their own usual behaviour rather than a branch average. A trader whose balance swings at festival time is not flagged for a swing that always happens. Where history is thin, the rules say so and the flag is marked lower confidence.

    Result: Seasonal or irregular customers are not flagged by an average that does not fit them.

  4. Back-test on known outcomes

    The rules are run against customers who earlier moved their salary account or closed a product, and against those who stayed, using data your team prepares. We report how many times the rules flagged the leavers and how many stayers were also flagged, so you can judge usefulness before anyone is called.

    Result: A written test shows what the rules caught, what they missed and what they wrongly flagged.

  5. Action guide for each score band

    For every band, one agreed response: a call from the relationship manager, a check of an open complaint, a note to the branch head, or no action. The wording of the call opening is drafted for a service conversation, not a sales push, and your compliance team reviews it.

    Result: Each band has a named owner and a written response the branch can follow.

  6. Ranked view for branch and cluster heads

    A short list by branch or relationship manager, produced from the approved data extract in a form your team can open without special software. It shows the customer reference, the band, the reasons and the last contact logged, so nobody is called twice by different people.

    Result: A branch head can open one list and see whom to call, why, and who last spoke to them.

How the result is measured

  • Share of customers with sufficient reliable data to score
  • Warnings recorded before a confirmed lapse or non-renewal
  • False alarms and missed departures

Recorded as a baseline before work starts, so every later report has an honest comparison.

Who it is for

This is written for these banks.

  • Private sector bank branches
  • Public sector bank branches
  • Regional rural banks
  • Cooperative banks
  • Small finance banks
  • Business correspondent networks
  • NRI-focused branches
  • MSME-focused branches

Before you decide

Agree the work. Know how to judge it.

Customer health scoring for banks: use these checks to prepare, choose the scope, and review what changes.

Bring the right evidence

Review repeat orders, inactive customers, and reasons for leaving. Separate a normal buying interval from a customer who stopped because of an unresolved problem.

Make the scope clear

Choose the customer group and the reason to contact them. Agree how complaints, renewal decisions, and additional offers are handled, including when a sales message is inappropriate.

Check the work in use

Track the response through to a retained or repeat order, with discounts and service costs visible. Record why the customer returned rather than crediting every repeat purchase to the campaign.

Know when to change course

If customers leave because delivery is unreliable, fix that before offering a loyalty reward. Repeated promotions can hide the cause without improving the experience.

These are discussion points, not a fixed package. GullySales confirms the deliverables, responsibilities, fee, and term in writing after reading your brief and speaking with you.

Before we start

What to have ready before the work starts.

The records, access and decisions this work needs from banks. Anything missing is noted in the written scope, with who will supply it.

A data extract your privacy lead has approved
The model needs customer-level records such as dates of credits, card use and logins. Your privacy or data protection lead decides what may be shared and how identifiers are masked. Without that approval we cannot test anything, and a score built on guesses would send managers to the wrong customers.
A list of customers who left or reduced activity
Closed salary accounts, lapsed fixed deposits and surrendered products give the model something real to learn from. If this list is missing, the rules cannot be checked against what happened, and the score would be an opinion. Even a short, accurate list is more use than a long uncertain one.
A statement of forbidden uses
Your compliance and risk teams should write down where customer behaviour data must not be used, for example in credit decisions or pricing. This keeps the score in its role. If the limits are not written, a useful list can drift into uses nobody approved, and a regulator or customer may later question it.
A relationship manager who will act
Someone must take the flagged list and make calls, then record what they heard. If no one owns that step, the score becomes a report that nobody reads. One or two managers willing to try it on a small group is enough to show whether the response works.
Records of complaints and contact logged
Open complaints and recent contact change how a flag should be read. A customer with an unresolved complaint needs a different response from one who simply went quiet. When these records are not available, we say so in the rules and lower the confidence of affected flags.

How it works

How the customer health scoring work runs

The work itself, in order. How an engagement starts, from your brief to a written scope, is set out on our method page.

  1. 1

    Agree the purpose and the limits

    We begin by writing down what the score is for: prompting a conversation about the relationship, nothing else. Your risk, privacy and compliance leads state which data may be used and which uses are forbidden, such as pricing or credit. These limits are recorded and attached to every later document.

  2. 2

    List the signals the bank can produce

    We go through your reports and extracts with the data and branch teams and list what exists: credits, card use, logins, instructions, complaints, product holdings. For each, we record the source and how reliable it is. A signal the bank cannot produce cleanly is left out rather than approximated.

  3. 3

    Choose a small set of customers to learn from

    Your team supplies a prepared set of customers who left or reduced activity, and another set who stayed, with personal identifiers removed or masked as your privacy lead requires. We study what the leavers' records showed beforehand, and what the stayers' records showed, so the rules rest on your own customers.

  4. 4

    Draft the rules and test them

    We write scoring rules with the reasons built in, then run them on the prepared set and report what was caught and what was wrongly flagged. Rules that flag a large number of steady customers are loosened or dropped. Your branch heads read worked examples and say whether the flags match what they know.

  5. 5

    Define the response for each band

    For each band we agree who acts, what they do and how the contact is logged. The opening line of a call is written as a service check. Your compliance team reviews it, and the branch tries it on a few accounts before the list is used more widely.

  6. 6

    Read the flags against what happened

    After the rules are in use, we compare flagged customers with what the bank recorded: who stayed, who left, which calls were made. Signals that did not predict anything are removed, and ones that did are given more weight. The comparison is written for your team and kept with the rules.

Where this fits

Three pieces of sales work, in this order.

This work sits inside sales process design, at stage 1, prospecting and lead generation. The audit comes before it, and sales management keeps it running.

  1. Step 1

    Sales audit

    Find where deals are being lost and what to fix first.

    Start here if you cannot say where orders are lost.

  2. Step 2 · you are here

    Sales process design

    One written way to sell, from first enquiry to repeat order.

    Start here if every salesperson sells differently.

  3. Step 3

    Sales management

    A target, a review routine and someone who owns the number.

    Start here if the team has targets and no routine.

Tell us what you need Your brief is read before we call. Scope, fee and term in writing.

FAQ

Questions owners ask before they call.

Not here? More answers, or ask on WhatsApp.

Is this a credit score?
No. It is a prompt for a service conversation about the relationship. It is not built to approve, reject or price any product, and the written limits we agree with your compliance team say so. Your own risk and compliance leads decide where behaviour data may and may not be used.
Which customer data do you need?
Only what your privacy lead approves: for example dates of regular credits, card use, login activity and open complaints, with identifiers masked as your policy requires. We do not need names, full account numbers or balances to design and test the rules, and we never ask for live system access.
Will the score predict who will leave?
It flags relationships that look like earlier ones that ended, and we show how it performed on your own past customers. It cannot know a customer's plans. Some flagged customers will be fine and some unflagged ones will leave, which is why the score prompts a conversation and not an assumption.
Is health scoring the same as customer retention?
Retention is the wider programme of keeping customers: offers, service fixes, follow-up. Health scoring is one tool inside it that tells you whom to look at first. You can use the score with your existing retention effort, or begin with it and add the rest later.
What if our data is patchy across systems?
We record exactly which signals are reliable and use only those. A score built on three clean signals is more useful than one built on eight uncertain ones. The inventory shows what is missing, and that list itself is useful to your data team for later clean-up.
Can branch staff see why a customer was flagged?
Yes, the reasons sit next to every flag, such as salary credit absent or card unused. A manager who can see the reasons can judge whether the flag is fair and can open the call properly. A number with no explanation would not be trusted and would not be used.
How much does it cost?
There is no price list, because the work differs by business. The fee is scoped after we have read your brief and spoken to you, and put in writing before anything starts, split into our time, your media spend and production.
How long is the contract?
The term is agreed in writing after we have spoken, along with the fee and the reporting. It is never a default twelve months, and renewal is decided on the numbers against the baseline recorded at the start.
How soon will we see results?
It depends on the trade, the channel and what is already in place. After we have read your brief and spoken to you, we say what we would expect to move first and how it will be measured against the baseline. Nothing here is guaranteed.
What do you need from us?
Your brief, and a conversation. After that, access to whatever this work touches, such as your website, your accounts or your enquiry record, and time with the people who handle your enquiries.

Your next practical step

Tell us what you are trying to grow, and we will read it before we call.

Eight short steps, mostly taps. Skip anything you would rather answer on the call.