Pre-Call Lead Scoring: Detect Intent Before You Dial

The best way to automate lead scoring and intent detection before a call is a two-part score: fit, which measures how well the company or person matches your buyer, and engagement, which measures what they did lately. Set a threshold, call the leads above it first, and let an AI call confirm intent and book only the ones who qualify.

This post covers the step before the call and the hand-off into it. Scoring during the conversation is a different job, and we cover it in AI lead scoring and intent detection on sales calls and in our guide to BANT lead qualification. Here we go one step earlier: which signals you collect, how you weigh them, where you set the line and what the AI call receives.

Key Takeaways

• A pre-call score has two parts: fit (who the lead is) and engagement (what the lead did), kept as separate numbers.
• HubSpot's scoring tool separates fit scores, engagement scores and combined scores for the same reason.
• Pick a call threshold from your own closed deals, not a template, and review it every month.
• The AI call confirms intent in conversation, so the score decides who gets called, not who gets booked.
• Topcalls returns a transcript, summary, extracted answers and an outcome for every call, so the score improves with each batch.

What is the best way to automate lead qualification scoring and intent detection before scheduling calls?

Score each lead on fit and on engagement, store both numbers on the lead record, and trigger a call when the combined score crosses a threshold you set from past wins. Then let a short qualifying call confirm intent, and book a meeting only when the lead meets your criteria.

Keep it simple at first. Five to seven signals, a clear weight for each and one threshold beat a model nobody can explain. You can add a statistical model later, once you have a few hundred outcomes to train it on. The score is an input to the call, not a verdict.

What goes into a pre-call lead score?

A pre-call lead score has four parts: firmographic fit, contact fit, behavior signals and timing. Fit tells you whether the lead could buy. Behavior and timing tell you whether the lead is likely to buy now. Each part gets a few points, and you adjust the weights as results come in.

HubSpot's own documentation describes the same split: a fit score uses property values such as region, company size and industry, and an engagement score uses events such as repeated interaction with your campaigns. Its example scores a company highly when it is in the target region, has 50 to 200 employees and works in hospitality. See the HubSpot guide to lead scores. You don't need HubSpot to use the idea.

• Firmographic fit. Industry, company size, region, tools in use. Source: your form, your CRM or a data provider.
• Contact fit. Job title, seniority, decision role. A buyer title earns more points than a student address.
• Behavior. Pricing page visits, a demo form, a reply to an email, a webinar sign-up, repeat visits within a week.
• Timing. How recent the action was. A form sent an hour ago beats the same form sent last month.

Take away points as well. A free-mail address, a competitor domain or an unsubscribe is a negative signal. Write each rule down in one line, so a salesperson can read it and agree.

How do you detect intent from behavior before you call?

Treat intent as a short list of actions that, in your history, came before a closed deal. Pull your last 50 to 100 won deals, list what each lead did in the 30 days before booking, and weight the actions that appear most often. Those actions are your intent signals.

Common signals in B2B are a pricing page visit, a demo or contact form, a reply to an outbound email, a second visit within a week and a download of a bottom-of-funnel asset such as a comparison page. Common weak signals are one blog visit or one email open, so give them few points.

Don't buy a "buyer intent" feed and trust it blindly. Third-party intent describes a whole company, while your own first-party behavior is a fact about one lead. Start with the facts you own. Our AI cold calling guide covers when outbound beats inbound.

How do you turn the score into a call threshold?

Set one threshold for "call now" and one for "nurture", based on the conversion rate of past leads at each score. Sort last quarter's leads by score, split them into bands, and look at how many in each band became meetings. Draw the line where a call is worth its cost.

A worked example with invented numbers, so replace them with yours: you score 1,000 leads. The top band holds 150 leads, 30 of which booked a meeting. The middle band holds 350 leads and 21 booked. The bottom band holds 500 and 5 booked. Calling the top band yields a booking rate of 20 percent, against 6 percent and 1 percent. You call the top band at once, call the middle band in a second wave and send the bottom band to email.

The cost side is simple with AI calls. Topcalls AI calls start from $0.25 a minute depending on the country, billed by the second, and recording, transcription and analytics are included. A 3-minute qualifying call costs from $0.75, so the threshold can sit lower than it would for a human rep's time. Put your own volume into the ROI calculator.

How do you hand a scored lead to an AI call?

Pass the score and its reasons to the call as fields on the lead, so the agent knows who it is talking to and why the lead was picked. On Topcalls, any field on a lead is visible to the agent in a lead data section, and custom fields work as well as name, email and notes.

A useful hand-off has four fields: the score, the top reason ("visited pricing twice this week"), the source and the time of the last action. The agent can then open with context: "I saw you were looking at pricing, so I wanted to check you got what you needed." Write instructions that refer to the lead, such as "mention the product they asked about when the lead has that field", rather than pasting values into the task.

Load the list by CSV, by API or from your CRM, and set the campaign's dispatch hours and time zone, so calls land in the lead's working day. The customer owns consent, so only load leads you have the right to call. Topcalls keeps no do-not-call list. See our guides to AI disclosure laws and GDPR for AI outbound calling.

What should the AI call confirm before it books?

The AI call should confirm the few facts a score can't know: the need, the decision role, the timeline and the budget range. This is the BANT idea of budget, authority, need and timeline, asked in conversation. The call books a meeting only when the answers meet your criteria, and it records why when they don't.

Define the questions and the pass rule before launch. Example: "has a named project, owns or influences the decision and expects to act within six months". The agent asks naturally and doesn't read a form. Topcalls extracts structured answers from each call through a schema you define, such as "did the lead agree to a demo?" or "what objections came up?", and maps them to an outcome.

For booking, the agent can call your tools during the call to book an appointment or update a record. For a hot lead, it can also transfer the call to a person while the lead is still on the line. See the lead qualification solution and the appointment setting solution.

How do call results feed back into the score?

Every finished call sends a webhook with the transcript, recording link, summary, extracted answers, outcome and your own metadata. Send that into your CRM, compare each outcome with the lead's pre-call score and adjust the weights each month. A score you never check against results drifts.

Topcalls also lets you subscribe to one outcome, so a "booked" call can trigger a task for the rep and a "not a fit" call can move the lead to nurture. Each outcome label becomes its own event, for example `call.completed.booked_callback`. Retry schedules handle no-answers: a lead who doesn't pick up can be tried again, 1 to 10 times, and voicemail counts as no contact.

A monthly review takes 30 minutes. Pull the leads that scored high and didn't qualify, and the ones that scored low and did. Ask what signal you missed and add or remove one rule. Keep a changelog, so you can tell which change moved the booking rate.

What mistakes should you avoid?

Five mistakes come up again and again. Scoring on email opens alone. Taking the threshold from a template. Hiding the reasons behind the score. Calling leads you have no right to call. Treating the score as the qualification. The score picks who to call, and the conversation decides who gets booked.

Also avoid a score that nobody on the sales team trusts. If a rep sees a lead marked hot and finds a student, the rep stops reading scores. Show the reasons next to the number and let reps flag mistakes.

Ready to test the hand-off on a small list? Start with 10 to 20 scored leads and read the transcripts before you scale. See AI voice agents for the call flow, then book a strategy call and we'll map your score to a first campaign.