You see which customers are drifting long before the renewal call.
Gully Sales builds one health score for every account from signals you already hold — usage, service history, payments, engagement and relationship — and attaches an owner and an action to each score band.
- One score per account, built from records your business already keeps.
- A weekly at-risk list with a name and a next action against each account.
- Expansion signals, so growth conversations start from evidence.
Gully Sales Private Limited works with small and medium businesses across India, and builds the model on data your team can actually maintain.
In one paragraph
What is Customer Health Scoring for Indian SMBs?
Customer health scoring gives every account one explained score built from what customers actually do: usage, service history, payment behaviour, engagement and relationship depth. Gully Sales defines the signals, sets weights and thresholds with your team, builds the scorecard in the systems you already run, and attaches an owner and a response to each band.
The problem
Risk arrives in your renewal month, not in your reports.
Most owners can name their unhappy customers. The trouble is when they learn it. In a small team, account health lives in the head of whoever last spoke to the customer, and it surfaces at the renewal call, in a payment that stops, or in a polite email saying they are moving in a different direction. Nothing in your reports moved before that. Revenue looked steady, because a customer who has quietly stopped using you keeps paying until the contract ends. By the time the number falls, the decision was made months earlier.
You will recognise it as
- You find out an account is unhappy when they call to cancel, not weeks before it.
- Ask three people which accounts are at risk this quarter and you get three different lists.
- Renewal conversations start from scratch, because nobody kept a record of how the year actually went.
- The accounts that get attention are the loudest ones, while quiet customers drift away unnoticed.
- Usage, service visits, tickets, payments and meeting notes sit in five places and are never read together.
- You cannot say which customers are ready to buy more, so upsell conversations remain guesswork.
What it costs the business
- Churn is discovered rather than prevented, and winning a replacement customer costs far more than keeping the one you had.
- Your team spends its week on whoever complained loudest, not on the accounts where attention would change the outcome.
- Renewal forecasts are opinions, so cash planning and hiring decisions rest on numbers nobody can defend in a review.
- Expansion revenue sits unclaimed inside accounts that were ready to buy more and were never asked.
Why it persists. It persists because the signals are real but scattered. Usage data sits with operations, tickets with support, payment history with accounts, and the relationship in one person's memory. Nobody has decided which of those signals matter, what weight each carries, or who reads them on which day. So the instinct of the person closest to the account stays the only assessment there is, and it is only as available as that person is.
If it stays unresolved. Left alone, this gets more expensive as you grow. Ten accounts fit in one person's head; sixty do not. Each new customer adds signals nobody watches, each departure teaches you nothing you can reuse, and the business keeps paying to replace revenue it already had.
What changes
You can say which customers are safe, slipping and ready to grow.
In the first weeks
- A single scored list of every customer, in one place, that you can read in five minutes.
- Written definitions of what healthy, watch and at risk mean in your business, agreed by your team.
- An owner's name against every account that needs attention this month.
In how the work runs
- A weekly account review that runs to a fixed agenda instead of a general discussion.
- A defined response when a score falls, so action does not depend on who happened to notice.
- Onboarding and adoption milestones that are tracked by date rather than assumed.
In sales and marketing
- Renewal conversations that open earlier, with the evidence of the year behind them.
- Expansion prospects identified from behaviour rather than from a hunch about who might say yes.
- A renewal outlook leadership can question and defend account by account.
In what management can see
- Leadership sees the state of the customer base without asking three people for their impression.
- Service and sales work from one account picture instead of two versions of the truth.
Over the longer term
- A retention habit that survives staff changes, because the knowledge sits in the model, not in one head.
- Cohort learning: every renewal and every loss makes the next quarter's score more accurate.
Gully Sales controls the deliverables: the signals, the model, the scorecard, the playbooks and the review rhythm. Whether a particular customer renews depends on your product, service and commercial terms. Scoring makes risk visible early. Acting on it remains your team's work.
Who it is for
This is for businesses with more accounts than one person can hold in mind.
The businesses it suits
- Service, subscription and retainer businesses where most of next year's revenue already sits inside current accounts.
- AMC, maintenance and contract-led firms with renewals spread across the calendar.
- Software and technology products holding usage data that nobody currently reads account by account.
- Distributors and B2B suppliers whose repeat order patterns say more than any survey would.
- Businesses with an account management or customer success function of two to twenty people.
- Owners who already sense which accounts are shaky and want that judgement written down and shared.
What usually prompts the call
- A large customer left recently and, looking back, every signal was already there.
- Renewal season arrives as a surprise each year, with no early view of what is at risk.
- You are adding customers faster than one person can keep track of them.
- The team is about to grow, and account knowledge must leave individual heads before it does.
- You want to sell more into existing accounts but cannot say which ones are ready to hear it.
What Gully Sales does
The work, component by component.
What a healthy customer looks like
We begin by writing down what a successful customer looks like in your business: what they should be using, how often they should be in contact, what result they should be seeing by month three and by month twelve, and what they should have bought again by then. Without that picture, a score is a number with no meaning behind it.
- Why it matters:
- Every later signal is judged against this definition, and your team has to agree on it before anything is measured.
- You receive:
- A one-page written definition of a healthy customer, by segment.
- Business value:
- Your team stops debating which accounts matter and starts working to one standard.
Onboarding health
The first ninety days decide most of what follows. We define the milestones a new customer must reach — activated, first result seen, people trained, first review held — and make each one a tracked field with a date and an owner rather than an assumption.
- Why it matters:
- Customers who stall in their first quarter rarely recover on their own, and the cost of that stall is paid at renewal.
- You receive:
- An onboarding milestone tracker with dates, owners and a stall alert.
- Business value:
- New customers reach value sooner, and a slow start is caught in week three rather than month ten.
Usage and milestone signals
We identify which behaviours genuinely indicate value in your business — logins, order frequency, service visits accepted, sites live, trained users still active, volume against contract — and turn each into a signal with a threshold. Signals that are interesting but not predictive are deliberately left out.
- Why it matters:
- Behaviour is the honest measure. Customers say they are fine and then stop using you; the usage record flatters nobody.
- You receive:
- A signal dictionary: each signal, its data source, its threshold and its weight.
- Business value:
- Engagement is measured from evidence rather than from the account manager's impression.
Risk signals and score bands
Falling usage, ageing tickets, repeat complaints, later payments, a champion who has left, silence since the last review — each becomes a weighted risk signal. We then set the bands that turn a combined score into a plain label: healthy, watch, at risk, critical.
- Why it matters:
- A number nobody can interpret changes nothing. A band with a name and a rule attached changes what happens on Monday.
- You receive:
- A weighted risk model with thresholds and four score bands defined in writing.
- Business value:
- Risk becomes visible weeks or months before the customer says anything to you.
Intervention playbooks
For each band and each common risk pattern we write what happens next: who contacts the customer, within how long, with what message, what is offered, and what the escalation is if there is no response. The playbook names an owner, so the action is not optional.
- Why it matters:
- Scoring without a response is reporting. The value sits in the intervention the score triggers.
- You receive:
- Playbooks per risk pattern, with owner, response time, outreach outline and escalation.
- Business value:
- Every falling account gets a defined response instead of whatever that week allowed.
The review rhythm
We install the meetings that keep the model alive: a short weekly triage of accounts that moved band, a monthly review of the whole base by segment, and a quarterly recalibration where the score is checked against what actually happened to those accounts.
- Why it matters:
- Models drift. Without a standing review, a scorecard becomes a spreadsheet nobody opens by the second quarter.
- You receive:
- Weekly, monthly and quarterly review agendas with the report each one needs.
- Business value:
- Account health stays on the leadership agenda instead of surfacing only in a crisis.
Expansion readiness
The same data that shows risk also shows readiness. Accounts using more than they bought, adding people, asking about adjacent needs, or scoring healthy for three straight months are flagged for a growth conversation, with the evidence attached to the flag.
- Why it matters:
- Most SMBs carry unclaimed revenue inside existing accounts and have no system for saying which ones to ask.
- You receive:
- An expansion-ready account list with the trigger and evidence behind each name.
- Business value:
- Upsell and cross-sell conversations open with a reason the customer recognises as true.
What you will have at the end.
- A written definition of a healthy customer for each of your segments.
- A signal dictionary naming every input, its data source, its threshold and its weight.
- The scoring model built in your CRM, spreadsheet or reporting tool, with the calculation visible.
- A scored list covering your whole customer base, with band, trend and named owner.
- An anonymised sample extract of the scored list, so you see the output before we build it.
- An onboarding milestone tracker with stall alerts for a customer's first ninety days.
- Intervention playbooks per risk pattern, with owner, response time and escalation path.
- An expansion-ready account list with the behaviour that triggered each name on it.
- Weekly, monthly and quarterly review agendas, with the report each meeting reads.
- A one-page leadership view: accounts by band, and movement since the previous month.
- A working session that trains your team to read, update and challenge the score.
- A recalibration note after the first quarter, recording what the model read right and wrong.
How it runs
The engagement, step by step.
- 1
Discovery and success definition
We meet your customer-facing team and, where you allow it, two or three customers. We establish what value looks like in your business, how customers come back, why the ones who left actually left, and how account decisions are taken today.
- You provide:
- Access to your account, service and sales people, and a list of customers won and lost over the last two years.
- We produce:
- A written definition of customer success by segment, and the decisions it should drive.
- Done when:
- You agree in writing what a healthy customer looks like.
- 2
Signal and data audit
We examine what you already record — CRM notes, invoices, service logs, support tickets, product usage, delivery records — and test which of it is complete enough to score with. Where a needed signal does not exist, we say so and propose the smallest way to start capturing it.
- You provide:
- Read access to the systems and exports you keep, and the person who knows what each field really means.
- We produce:
- A data audit listing usable signals, gaps, and the effort each gap would take to close.
- Done when:
- The signal list is fixed and every signal has a named source.
- 3
Score design
We design the model: which signals carry which weight, how recent behaviour counts against older behaviour, where the band boundaries sit, and how segments differ. Weighting is decided with your team, because they know which behaviour has predicted trouble before.
- You provide:
- Two working sessions with the people who own accounts, and their view on which signals matter.
- We produce:
- A documented scoring model with weights, thresholds, bands and the reasoning behind each.
- Done when:
- The model is signed off and anyone in the room can explain it.
- 4
Build and backtest
We build the score inside the system you already use and run it backwards over the last year or two, checking whether it would have flagged the customers you actually lost and left the loyal ones alone. Where it fails that test, the weights are corrected before anyone relies on it.
- You provide:
- Historic data for the backtest period and the list of accounts that were lost or renewed.
- We produce:
- A working scorecard, the backtest result, and the adjustments made because of it.
- Done when:
- The score agrees with what you already know about your past customers.
- 5
Playbooks and ownership
Each band and each pattern gets a defined response, an owner and a response time. We write the outreach outlines, the escalation ladder and the note that must be recorded afterwards, so an intervention leaves a trace somebody else can read.
- You provide:
- Agreement on who owns which accounts, and how much time each owner can commit weekly.
- We produce:
- Intervention playbooks, an owner map, and the recording standard for each action.
- Done when:
- Every band has a named response and a named person behind it.
- 6
Pilot and handover
The model runs live for a full cycle with your team using it in the weekly review. We sit in those reviews, correct the misreadings, fix the fields nobody updates, and train the team to maintain the model without us.
- You provide:
- Attendance at the weekly reviews, and honest feedback about what the score gets wrong.
- We produce:
- Training, a maintenance guide, and corrections drawn from real use of the model.
- Done when:
- Your team runs a full account review without us in the room.
- 7
Recalibration
After a quarter we compare what the score predicted with what happened: which flagged accounts were saved, which healthy ones surprised you, which signal proved noisy. Weights and thresholds are adjusted, and every change is documented with its reason.
- You provide:
- The quarter's outcomes: renewals, losses, expansions, and the interventions that were run.
- We produce:
- A recalibration report with the model changes and the evidence behind them.
- Done when:
- The model is corrected and the review rhythm continues without support.
Ways to work with us
Start with the model alone, or have us embed it with your team.
Health scoring design
Discovery, the signal audit, the scoring model and the backtest, handed over as a documented scorecard your own team then builds and runs.
Design and embed
Everything in the design engagement, plus the build inside your systems, the intervention playbooks, the review rhythm and a live pilot cycle run with your team.
Ongoing review support
A recurring arrangement where we join your monthly account review, keep the model calibrated as the customer base changes, and coach the owners through interventions.
Scorecard health check
A short review of a scoring model you already run, tested against your actual renewals and losses, reporting where it misleads you and what to change.
Why Gully Sales
What you are actually choosing when you choose us.
We connect the score to sales, not only to service.
Scoring pays back when the renewal, the service response and the expansion conversation are joined up. Gully Sales builds systems across marketing, sales and customer success, so the score lands inside a working revenue process rather than in a report.
Built for the data you actually have.
Most Indian SMBs have no product analytics and a half-used CRM. We design around invoices, service logs, order patterns and contact records, and add sophistication only when a signal earns its place in the model.
Tested against your own history.
A model that cannot explain the customers you have already lost will never be trusted. We run it backwards over your past renewals and losses before go-live, and show you plainly where it failed.
Every band carries an action.
Each score band comes with a named owner, a response time and an outreach outline, because a dashboard nobody acts on is an expense rather than an asset.
Your team can maintain it.
We build inside the tools you already pay for and train the people who will keep it running, so the model does not stop working the month our engagement ends.
Plain about what a score cannot see.
Behaviour flags risk. It will not explain a customer who leaves because of price, a change in their own business, or a new decision-maker with old loyalties. We say which is which.
Where it applies
The same service, in different businesses.
Software and technology products
- The situation:
- A growing subscriber base where the founder still personally senses which accounts are shaky, and cannot keep doing it past about fifty customers.
- How it applies:
- Login frequency, active users, support load, onboarding milestones and payment behaviour combined into one score per account, triaged weekly by the success team.
- Likely benefit:
- Risk is caught in the quarter it appears, and the founder stops being the only early warning system in the company.
Equipment and AMC providers
- The situation:
- Annual maintenance contracts renewing across the year, with the service history sitting in engineers' reports that nobody reads before the renewal call.
- How it applies:
- Service visit patterns, breakdown frequency, complaint history, spare-part spend and payment delays scored per contract, with an alert well before the renewal date.
- Likely benefit:
- Renewal conversations begin with the service record in hand instead of a cold call at expiry.
Professional services and agencies
- The situation:
- Retainer clients whose engagement quietly falls away: fewer meetings, slower approvals, unused scope, and a champion who has moved to another company.
- How it applies:
- Meeting attendance, approval turnaround, scope consumed against scope sold and stakeholder changes tracked as relationship signals, reviewed monthly.
- Likely benefit:
- A retainer at risk is discussed while there is still time to change the work being done.
Distribution and B2B supply
- The situation:
- Hundreds of repeat buyers, where a customer who has quietly moved half their volume to another supplier still looks like an active account on the ledger.
- How it applies:
- Order frequency, basket composition, volume trend against the same period last year and payment behaviour scored per buyer, with reactivation triggers.
- Likely benefit:
- A falling buyer is called in the month the pattern changes rather than a year later.
Education and training institutes
- The situation:
- Enrolled learners or corporate batches where dropout only becomes visible at the end of a term, when the fee for the next one does not arrive.
- How it applies:
- Attendance, assignment completion, portal activity and fee payment timing combined into a learner and corporate account health view with an intervention list.
- Likely benefit:
- Support reaches the learners who are slipping while completion is still realistic.
Healthcare and clinic groups
- The situation:
- Patients on treatment plans or corporate wellness contracts where follow-up visits simply stop, and nobody sees the pattern until the schedule is empty.
- How it applies:
- Visit intervals, plan adherence and recall response scored at the account level, within the privacy rules and consent your practice already follows.
- Likely benefit:
- Follow-up outreach is aimed at those who have genuinely lapsed rather than at the whole list.
Questions buyers ask
Before you enquire, the answers you will want.
What is customer health scoring, in plain terms?
It is one number, with a name attached, for every customer you have. The number combines what the customer does — how much they use, how often they contact you, how they pay, how the relationship is holding up — into a band such as healthy, watch or at risk. It replaces the opinion of whoever spoke to them last with a shared, explained view that anybody in your team can read and question.
Which behaviours indicate value, risk and expansion potential?
Value shows as steady use: repeat orders, active logins, service visits accepted, trained people still working with what you sold. Risk shows as decline and friction: usage falling, tickets ageing, complaints repeating, invoices paid later, a champion who has left, silence since the last review. Expansion shows as pressure against limits: more users than purchased, questions about adjacent needs, rising volume, three healthy months in a row. Your exact signals are defined for your business, not imported from a template.
We have no CRM or product analytics. Can we still score health?
Usually yes. Most Indian SMBs hold more than they realise: invoices, order history, service and delivery logs, WhatsApp and email records, complaint registers. We audit what exists, build the first model from that, and name the two or three gaps worth closing. If the honest answer is that almost nothing is recorded, we say so and start with the smallest capture habit rather than a score built on air.
How long does the engagement take?
It depends on how many customers and systems are involved and how quickly your team can meet. Discovery, the signal audit and the model design usually run over a few weeks, the build and backtest follow, and the model then runs live for a full cycle with your team before handover. We do not commit to a fixed timeline in advance, because data access and internal availability set the pace more than our work does.
What inputs are required from us?
Access to your customer records, and to whatever holds usage, service, support and payment history. Time from the people who own accounts: two design sessions, plus attendance at the weekly review during the pilot. A list of customers won, renewed and lost over the last year or two, for the backtest. And a decision-maker who can settle what a healthy customer means in your business.
How is success measured?
Against the baseline recorded before we start. The direct measures are warning time, meaning how early a loss was flagged, and score accuracy against the customers you actually lost. The business measures are renewal rate, churn, expansion revenue from flagged accounts, time to value in onboarding, and lifetime value by cohort. Behavioural measures move within weeks; revenue measures need two to four renewal cycles to read fairly.
What is excluded from the scope?
We do not build your product, run your support desk, or take over ownership of your accounts. Buying new software is not part of this unless you ask separately; we build inside what you already run. Survey and NPS programmes, complaint handling systems and full customer success operating models are separate services. And we do not promise that a flagged account stays, because that depends on what your team does with the flag.
How is this different from NPS or a satisfaction survey?
A survey records what a customer says, on the day they answer, if they answer at all. Health scoring reads what they do, every week, whether or not they respond. Surveys often stay positive while behaviour is already falling, and quiet customers rarely fill them in. The two work together: run NPS for the reasons behind the behaviour, and use the score for the early warning.
3 more questions
How is this different from churn analysis?
Churn analysis looks backwards at customers who have already gone, to explain why they went. Health scoring looks forward at the customers you still have, to flag them while something can still be done. The two feed each other: what churn analysis learns about your past losses becomes weighting in the score, and the accounts the score misses become the next churn review's questions.
Who should own the customer health score in a small team?
One person, named. In most SMBs that is the account management or customer success lead; in a smaller firm it is often the founder or the sales head. Ownership means keeping the data current, running the weekly triage, and reporting movement each month. Interventions are shared across whoever holds each relationship, but a model owned by everybody stops being updated within a quarter.
What if the score is wrong about an account?
It will be, sometimes, and that is designed for. Every score is explained by the signals behind it, so an owner can see why an account was flagged and record a correction with a reason. Those corrections are read at the quarterly recalibration and change the weights. A model your team can argue with improves; a black box they cannot question gets ignored instead.
Talk to us
See what your own records already know about who is slipping.
It is a conversation, not a pitch. We look at what you record today, what your renewals look like, and whether scoring would genuinely help you — and we will tell you if it would not.
- No obligation and no sales script
- A reply from someone who does the work
- Your details are never sold or shared