How to set a lead fit line for AI scoring

Qualify B2B leads with AI scoring by defining a strict ICP and setting a numerical fit line to automatically filter out low-quality prospects.

Get started

Short answer

Qualify B2B leads by defining a strict ideal customer profile and setting a numerical cutoff, or fit line, that rejects leads who don't meet your core requirements. AI scores profiles from 0 to 100 against those criteria, so you only message the people above your threshold.

Define your non-negotiables

AI scoring fails when the instructions are vague. Tell it to find innovative leaders and it will score anyone who put the word innovative in their headline. You end up with a full lead group and nobody in it who buys.

Start with hard requirements you can check on a LinkedIn profile. These are binary. If someone fails one, they should score near zero no matter how good the rest of the profile looks. A VP of sales at a 12-person agency is not a fit for a product built for enterprise sales teams, even if every other signal lines up.

Pull up your last ten closed deals and look for patterns in job title, company size, industry and who they report to. Write down the pattern, not the names. If eight of the ten were heads of marketing at companies between 50 and 200 people, that's your core. The other two are exceptions you don't build a rubric around.

Keep the list short. Three or four requirements is enough. Every extra criterion narrows the pool and makes it harder to tell which one is doing the work when the scores look wrong.

Vague criteria

  • Growth minded founders
  • Innovative marketing teams
  • Companies that want to scale

Hard requirements

  • Founder of a B2B SaaS company with 11 to 50 employees
  • VP of sales at a fintech company in North America
  • Head of demand gen at an agency with more than 200 staff

Build your scoring rubric

Once the non-negotiables are written down, decide how the 100 points get distributed. A person with the exact title you sell to is worth more than someone in an adjacent role, and the rubric should say so out loud.

Split it into tiers. The core match carries most of the points. Secondary signals add nuance, like a related title or a company in a neighboring sector. Bonus signals break ties between two leads who both clear the core, such as recent posting activity or a shared connection.

The rubric is what stops the AI handing a 90 to someone in the right industry with a title that never buys. Without it, the model leans on whatever words appear most often in your description, which is rarely the same thing as who pays.

Write each line as something the model can check against a profile. 'Works at a company with 11 to 50 employees' is checkable. 'Ambitious' is not.

Core match

Tier
60
Points
Exact target title and target industry

Secondary match

Tier
25
Points
Related title, or the target title in an adjacent sector

Bonus signals

Tier
15
Points
Recent posts, shared connections, mutual groups

Choose your fit line

The fit line is the number where you stop keeping leads. Below it, the lead is dropped. Above it, the lead goes into a group and can be pushed into a campaign.

There's no universal right answer. A high-ticket service that takes weeks to close deserves a high line, because a bad meeting costs you an hour you could spend on a real one. A low-friction offer can afford a lower line, because the cost of a wasted message is small.

Move the line with your calendar. Empty pipeline, lower it and see what comes back. A week of meetings that went nowhere, raise it. The line is a dial you turn as things change.

  1. 1

    Start high

    Set the line at 80 and run it for a week. You'll get fewer leads and you'll read every one. The first week is about checking the rubric. Filling the calendar comes later.

  2. 2

    Drop to the middle

    Once the top-scoring leads look right, drop to 60 or 65 and compare. The leads that appear between the two lines tell you what the extra 20 points were filtering.

  3. 3

    Go low to test a new market

    A line around 45 is for testing a segment you haven't sold into. You're buying information about who responds.

Test and calibrate the score

Your first line will be wrong. Plan for it. After the first batch of leads, read the profiles that landed just above and just below the cutoff. The ones at the top will look fine either way.

If the ones just below look like buyers, the line is too high. If the ones above look wrong, either the rubric is too loose or the line is too low. Those are different fixes and it's worth knowing which one you have before you touch either.

Watch for keyword inflation. A word that shows up on half of LinkedIn will pull scores up without meaning anything. Penalize it in the rubric or take it out. The same goes for a title that has spread across industries, like account executive, which tells you almost nothing on its own.

Example

  1. You

    Twenty leads came in at 55 with the line at 60. I read all twenty.

  2. Prospect

    Any of them actual buyers?

  3. You

    Fourteen are VPs of sales, just at companies smaller than my range. The other six are wrong.

  4. Prospect

    So the title match is fine and company size is what's dragging them down.

  5. You

    Right. I'm dropping the line to 52 and adding points for company size so the ones at 55 sort themselves out.

Where the fit line goes wrong

A few failure modes show up again and again, and all of them are fixable in an afternoon.

Too many criteria is the common one. If your rubric needs twelve conditions to describe a buyer, the AI will find almost nobody and you'll blame the tool. Cut back to the three or four things that actually predict a purchase.

Setting the line to flatter you is the other. A line of 90 feels rigorous. In practice it means you get four leads a month, stop checking them, and never find out whether the rubric works.

  • A rubric with more than five conditions
  • A line above 85 that nobody reviews
  • Bonus signals worth more than 15 points
  • A rubric written once and never edited

Doing this with Kinlead

Kinlead finds people on LinkedIn who match the customer you describe, or your website, and scores each lead from 0 to 100 against that profile. Leads below the fit line are not kept.

Kept leads land in lead groups. Any campaign can point at any group, so a strict group can feed a high-touch sequence while a looser group feeds a volume test, and you can compare the two without rebuilding anything.

If you'd rather start from a competitor's audience than describe your own ICP, the Steal Customers agent watches competitor company pages and optional founder or employee profiles, reads the comments on their recent posts, and turns commenters who look like buyers into leads. Each lead keeps the post URL, what the post was about and what they commented.

Pro has a minimum booking guarantee: after a two-week warm-up, any week with fewer than three qualifying bookings, with at least one agent active all week, can be claimed as a refund minus transaction charges of about 3.5% to 5%. Terms are at minimum booking guarantee.

Frequently asked questions

Related

Get started with Kinlead

Prefer markdown? How to set a lead fit line for AI scoring.md