AI lead scoring in India: score the enquiry, not the person

Short answer: Most small Indian businesses do not need to buy AI lead scoring. Their problem is a noisy queue from directory listings, WhatsApp and missed calls, and a points score in the CRM or sheet they already use sorts it. On the DPDP Act, our reading is that the score should use only what the person did and told you in that enquiry, not bought or enriched data.

This sits under our guide to AI lead generation in India. Here: whom your team calls first, and whether paying for AI to decide it is worth it.

Is your problem scoring, or a noisy queue?

Picture 10am. Your counsellor or sales executive opens a list of 60 new enquiries. Some are serious. Many are not. The buyer who wanted a quote today waits behind someone who was only checking a price.

That is a sorting problem, not a prediction problem. A model tries to say who will buy. A sorting rule only says who deserves the first call, based on what has already happened. A small team usually needs only the second.

Where Indian enquiries come from, and why that makes scoring triage

A typical small business gets enquiries from directory and marketplace listings (IndiaMART or JustDial style buy-leads, often shared with several sellers), WhatsApp chats, missed calls, website forms and Meta lead forms.

Many of these are low intent. A person who sent the same requirement to five suppliers is comparing prices, not choosing one. A person who tapped a lead form in a queue may not remember doing it. We put no figure on this.

So the score is triage of a noisy queue, not prediction. You do not need a model trained on thousands of records. You need a short list of things that separate a person who is ready from a person who is browsing, and a column to hold the total.

If WhatsApp is where most of this arrives, the capture side is in WhatsApp chatbots for lead capture. The score starts once the enquiry is in your list.

The points table, and what to take away

Here is a starting set. The weights are illustrative. They are a place to begin, adjusted against your own won and lost deals, not a finding from any study.

What happened in this enquiry	Points
Replied on WhatsApp after your first message	+15
Shared requirement details or a quantity	+20
Asked for a site visit, demo or counselling slot	+25
Named a timeline (“this month”, “before admissions close”)	+15
Answered your callback	+10
Enquired a second time	+10
Directory lead shared with other sellers, no reply from the person	-15
Number unreachable on two attempts	-20

Add the points, sort the list, call from the top. A column and a sort filter do this in a sheet, and most CRMs take the same thing as a number field with a rule.

Weight actions the person chose to take above things that merely happened to them, and let points expire so a lead silent for three weeks loses its place. Eight rows is enough. A score nobody can explain in two minutes will be ignored.

Why Diwali and the end of March make recency points lie

Most scoring advice gives extra points for a recent enquiry. In India that rule bends at certain times of year. Around Diwali and through wedding season, enquiries surge in many trades, and a score that rewards recency will push the whole list up together. At month-end and around 31 March, the financial year end, some buyers rush to close, and a quiet April is not necessarily lost interest.

Tier 2 and tier 3 enquirers may also reply on WhatsApp more slowly for reasons unrelated to intent, such as shared phones or a shop busy until evening, so a table that punishes slow replies may push down serious people. This is reasoning, not a statistic. Review the weights after each season.

Surname, religion, caste and pin code stay out of the score

Some signals look useful and are not. None of surname, pin code or language says whether this person wants to buy from you this month.

Our view: keep surname, religion and caste out of the score, and do not let language stand in for anything. Keep pin code out as a proxy for wealth or community. These add bias and make the score less accurate. This is a fairness and accuracy argument, not a legal one: the DPDP Act has no anti-discrimination clause.

Language still matters, for a different job. It should decide which person calls back. A parent writing in Hindi should get a Hindi-speaking counsellor, and a Tamil or Marathi enquiry a Tamil or Marathi speaker. That is routing, and it belongs in the assignment rule, not in the points.

What the DPDP Act points at: the enquiry in front of you

Parliament enacted the DPDP Act on 11 August 2023. The Rules were notified in November 2025. Law firm summaries put the start of the duties that touch a marketing list (notice, consent, withdrawal, rights, breach reporting, penalties) at 13 May 2027. For the timeline and notice detail, see our DPDP Act guide for marketing.

Section 7(a) lets a business process personal data “for the specified purpose for which the Data Principal has voluntarily provided her personal data to the Data Fiduciary, and in respect of which she has not indicated to the Data Fiduciary that she does not consent to the use of her personal data.”

The Act’s own illustration: a person messages a real estate broker asking for help finding rented accommodation and shares her data for that. The broker may process it to identify and tell her about available rentals. When she later says she no longer needs help, the broker must stop processing.

Our reading: ranking an enquiry to decide whom to call back first sits inside the purpose the person gave, so a score built only on that enquiry fits the pattern. Pulling in bought or enriched data, or reusing the score for unrelated marketing, steps outside it. When someone says they are no longer interested, stop scoring and contacting them.

Section 8(3) adds that where personal data is likely to be used to make a decision that affects the person, the business must ensure its completeness, accuracy and consistency. Our reading: a score deciding who gets a callback is arguably such a decision, so keep inputs current and correct a wrong field when told.

Section 9 matters for coaching and school-admission businesses. For anyone under 18 you need verifiable parental consent, and a business “shall not undertake tracking or behavioural monitoring of children or targeted advertising directed at children.” Our view: score the parent’s enquiry, never a child’s behaviour.

On penalties, the Press Information Bureau lists up to ₹200 crore for violating obligations relating to children, and up to ₹50 crore for any other violation by a Data Fiduciary. A small business that scores only its own enquiries and stops when asked is a long way from either.

One more reading of ours: unlike UK GDPR, the DPDP Act has no separate rule on profiling or automated decisions. That does not make scoring unregulated, because purpose limits and accuracy still apply.

This is not legal advice. Check your own case with a lawyer.

A hypothetical: a Jaipur coaching institute in admission season

This is invented, not a client. A small institute in Jaipur prepares students for competitive exams. It has three counsellors. Enquiries arrive from a directory listing, Instagram lead forms and WhatsApp, a few hundred in admission season. Before the score, the sheet is in order of arrival, so a parent who messaged “fees?” at midnight ranks with one who asked for a counselling slot.

Add the points table above, with two rules. The first is parent, not student: the score is built on the parent’s or guardian’s enquiry (did they reply, share the target exam and class, ask for a slot, name a start date) and never records what the student browses or does. The second is language: each new row goes to the counsellor who speaks the parent’s language, and language is not in the points.

The top rows are now parents who replied, gave details and asked for a slot. The bottom holds directory leads with no reply and numbers that did not connect.

These are assumptions, not results. Suppose the first callback to a high-scoring parent moves from the next afternoon to the same morning, and each counsellor saves about two hours a day by not working the bottom of the list first. At an assumed internal cost of ₹500 an hour, three counsellors over a 25-day month would save about ₹75,000 in time. Check it against your own payroll.

When the AI layer earns its fee

Our view is that a points table stops being enough when your CRM holds enough closed deals, won and lost, recorded with the original enquiry details, for a pattern to show that you would not spot yourself. We give no number, because any threshold would be invented. Until then, a model has nothing to learn from.

Zoho CRM and HubSpot both have scoring features on some paid plans. Check your own edition before you buy anything separate. We quote no prices here.

At that stage, run the AI score beside your points score for a few weeks and compare which picked the buyers. If a tool cannot tell you which fields it used, it should not decide who gets a callback.

For tool options, see AI lead generation tools for India. If the score feeds calling, read AI voice agents for lead follow-up, and the marketing automation guide shows where a score fits among your other automations.

Cases where a score is just extra admin

Skip scoring if you get a handful of enquiries a week, or if each is a large one-off project that deserves a personal reply. Read each one and call. A score only adds a column to maintain.

Skip it if most rows hold only a name and a phone number; fix the enquiry form first. After a month, check whether top-scored leads convert better than the rest. If not, adjust the weights or drop it.

If you want help building the sheet or wiring it to your CRM, that is our AI automation services work.

Frequently asked questions

Do I need to buy an AI tool to score leads in India? Usually not. Our view: a small team with a noisy enquiry list gets most of the benefit from a points column in the CRM or sheet it already uses. A learning model needs a record of closed and lost deals to learn from. Until you have that record, a model has little to learn, and you are paying for a guess.

Can I buy a phone number list and score it under the DPDP Act? Our reading is that bought or enriched data steps outside the purpose a person gave when they enquired. Section 7(a) covers the specified purpose for which a person voluntarily provided her data. A score built on your own enquiry data fits that pattern better. Take legal advice before you build a score on purchased data.

Does the DPDP Act ban lead scoring or automated decisions? Our reading is that it does not. Unlike UK GDPR, the Act has no separate profiling or automated decision rule. That does not leave scoring unregulated. Purpose limits and the Section 8(3) duty of accuracy still apply when a score decides who gets a callback. This is not legal advice, so ask a lawyer about your own case.

Should a lead’s language or pin code change its score? No. Our view is that language should decide which person calls back, whether Hindi, Tamil or Marathi, and should never move the score. A pin code as a proxy for wealth or community is also a poor and unfair signal. Keep both out, and score only what the person did and told you during that enquiry.

How should a coaching institute score enquiries from students? Score the parent’s or guardian’s enquiry, never a child’s browsing or behaviour. Section 9 of the DPDP Act asks for verifiable parental consent for anyone under 18 and says a business shall not undertake tracking or behavioural monitoring of children. Our view is to keep the student’s own activity out of the score entirely.

How many closed deals do I need before AI scoring is worth it? We do not give a number, because any threshold would be invented. Our view is to ask a simpler question. Does your CRM hold enough won and lost deals, recorded with their enquiry details, for a pattern to show? If you cannot answer that from your own records, keep the points score.

---

Not sure which enquiries in your list deserve the first call? Get a free marketing audit and we will look at where your leads come from and how they are sorted.