Not every enquiry deserves the same response. A visitor who wants a quote for a job this week is worth a phone call in minutes. Someone browsing prices for next year is worth an email. Someone outside your area is worth a polite reply and nothing more. Qualification is the practice of telling these apart, and doing it in a chat means asking the right questions without making the visitor feel interrogated. This guide gives you a simple framework for qualifying leads in a website chat: what to learn, how to learn it naturally, how to score and route leads, and how to check that your scoring reflects reality.
What qualification is for
Qualification serves two goals. For the business, it decides who gets attention first and how. For the visitor, it gets them to the right next step faster. Done well it feels like good service; done badly it feels like a form disguised as a chat. The framework below is about learning what you need with as few questions as possible.
The four things worth knowing
| Dimension | Question it answers | Example signals |
|---|---|---|
| Need | What do they want? | Service asked about, description of the problem |
| Fit | Can we serve them? | Location in your area, service you offer, price range acceptable |
| Timing | When do they want it? | 'This week', 'ASAP', 'sometime next year' |
| Ability and authority | Can they decide and pay? | Owner or tenant, budget mentioned, decision maker |
This resembles the classic BANT model (budget, authority, need, timeline) adapted to service businesses. You rarely need all four, and asking about budget directly often feels rude. Infer where you can.
Signals you can read without asking
- What they asked about: a specific service and price is stronger than a general question.
- Whether they asked about availability or how to book: strong intent.
- How specific the request is: a detailed job description beats 'do you do roofs?'.
- Whether they gave contact details unprompted.
- The page they started on: a pricing page is warmer than a blog article.
- How many questions they asked: several signals real evaluation.
- Whether they mention a deadline or an emergency.
See lead scoring from chat behaviour for how these signals combine.
A simple scoring model
Use a small scale you can act on. Points make it explicit.
| Signal | Points |
|---|---|
| Asked about price or availability | +2 |
| Wants a service this week or urgent | +3 |
| In your service area | +1 |
| Gave phone number or email | +2 |
| Booked or asked to book | +4 |
| Detailed description of the job | +1 |
| Outside area or service not offered | -3 |
| Only browsing, no specifics | 0 |
| Total | Label | Action |
|---|---|---|
| 6 or more | Hot | Call or message within minutes; text alert to the owner |
| 3 to 5 | Warm | Reply the same day by email or phone |
| 0 to 2 | Cool | Automatic follow-up and nurture; no urgent alert |
| Negative | Not a fit | Polite response, no follow-up |
These numbers are illustrations. Adjust them to your business, then compare labels with what actually happened.
Ask questions that feel like help
| Instead of | Kokeile |
|---|---|
| What is your budget? | So I can point you to the right option, are you looking at something basic or more comprehensive? |
| When do you need this? | Is this something you would like done soon, or are you planning ahead? |
| Are you the decision maker? | Would anyone else be involved in choosing, so we can include them in the quote? |
| Give me your phone number. | What is the best number to reach you if we need to confirm details? |
Each question should be one the visitor could see the value of. If you cannot say why you are asking, drop it.
Ordering the questions
- Answer what they asked first.
- Offer a next step (booking, quote, callback).
- Ask for the minimum to take that step.
- Ask one qualifying question at a time.
- Stop asking once you know how to route the lead.
Three or four questions are usually enough. More reduces completion. See conversational lead capture.
Routing by score and type
| Liidi | Route | Nopeus |
|---|---|---|
| Hot, urgent | Text and email alert to the owner or on-call person | Minutes |
| Hot, booking made | Confirmation sent; team notified | Immediate |
| Warm | Email to the sales inbox; add to a follow-up list | Same day |
| Cool | Nurture email sequence, if consented | Days |
| Not a fit | Polite decline with a helpful suggestion | Immediate |
| Complex or sensitive | Handover to a person with context | As soon as possible |
Instant delivery matters. See instant lead routing in chat and speed to lead. In Crescelo, leads go to email with instant alerts, and can go to Slack, Google Sheets or webhooks.
Different businesses, different signals
| Business | Strong signals |
|---|---|
| Home services | Emergency wording, address in area, photos, this-week timing |
| Dental or clinic | New patient, urgent pain, insurance question, wants a time |
| Real estate | Specific property, pre-approval, viewing request |
| Professional services | Matter type, deadline, decision maker, consultation request |
| Salons | Specific service, date given, bridal or group booking |
See buyer versus seller lead qualification and AI lead capture for home service contractors for worked examples.
Check that your scoring works
- Record the label for each lead for a month.
- Note outcomes: booked, quoted, won, lost, no reply.
- Compare: did hot leads convert more than warm and cool ones?
- Adjust points where the labels mislead.
- Repeat quarterly.
| Metric | What it shows |
|---|---|
| Conversion by label | Whether the scoring predicts outcomes |
| Response time by label | Whether hot leads get fast attention |
| Leads lost with no follow-up | Process gaps |
| Share of leads out of area | Whether your content states your area clearly |
Privacy and honesty
- Tell visitors they are talking to an AI assistant.
- Say why you ask for contact details.
- Do not infer or store sensitive characteristics.
- Do not use scoring in ways that unfairly disadvantage people.
- Keep only what you need and delete what you do not.
See the GDPR checklist. This is general information, not legal advice.
Common mistakes
- Asking for budget in the first message.
- Making the chat a long form.
- Treating every lead as hot.
- Never checking whether the scores predict anything.
- Sending alerts for every lead so people ignore them.
- Not telling the team what the labels mean.
A worked example: three leads in one morning
| Liidi | What happened in chat | Score and route |
|---|---|---|
| Maria | Asked about a boiler that stopped working, gave an address in your area, left her mobile and said she is home all day | Need, fit, timing and contact: hot; text alert to the on-call engineer; call within ten minutes |
| Tom | Asked general prices for a new bathroom, no timing, no contact details | Warm or cool; assistant offers a brochure by email; follow-up if he leaves an address |
| An enquiry from 200 miles away | Asked for an emergency callout | Not a fit; polite reply explaining your area and suggesting a local search |
Each lead got a different response in proportion to its value and urgency. Maria was helped first, Tom was kept warm cheaply and the out-of-area visitor was not left hanging.
Rules for keeping scoring fair
- Score behaviour and stated needs, not personal characteristics.
- Do not use names, accents or writing style to infer anything.
- Give every visitor a courteous response, whatever their score.
- Review the model periodically for unintended bias, such as penalising leads who write in a second language.
- Let a person override a score at any time.
Qualification should make service better for everyone, not gatekeep. A low score means lower priority, not a rude reply.
Handing a qualified lead to a person
When a lead is hot, the handover matters as much as the score. The person who follows up should see a short summary at the top: what the visitor wants, where they are, when they need it and how to reach them, followed by the full conversation. They should also see what the assistant already told the visitor, so they do not contradict it or repeat questions. A good summary takes ten seconds to read and lets the caller start with 'I hear your boiler stopped this morning', which builds trust before the call has really begun.
Frequently asked questions
What is lead qualification?
It is deciding how promising an enquiry is and how to respond, based on need, fit, timing and ability to buy.
Can a chatbot qualify leads?
Yes. It can ask a few natural questions, read signals such as timing and specificity, and route the lead by score.
How many questions should it ask?
Usually three or four. Enough to route the lead, not so many that visitors leave.
Should I ask about budget?
Often indirectly. Ask about the scope or option they are considering rather than demanding a figure.
What do I do with cool leads?
Send helpful follow-up if they have consented, and check them again later.
How do I know my scoring is right?
Compare labels with outcomes each month and adjust.
See the lead capture feature, test a lead conversation in the playground and start a free trial.
Further reading
- Google Search Central: creating helpful, people-first content: What Google looks for in useful pages.
VASTAA JOKAISELLE KÄVIJÄLLE. VARAA AIKA NIILLE, JOTKA OVAT VALMIITA.
Ota Crescelo käyttöön verkkosivustollasi 15 minuutissa
Opeta avustajasi hinnastoillasi, asiakirjoillasi ja palveluehdoillasi. Ota käyttöön lähteisiin perustuvat vastaukset ympäri vuorokauden ja päällekkäisyydettömät ajanvaraukset suoraan chatissa.