AI lead scoring vs manual profile research

How to define a customer profile, set a fit line and test AI lead scoring against your own judgement so you stop reading profiles that were never going to buy.

Get started

Short answer

AI lead scoring compares each LinkedIn profile against a customer profile you write, then gives that person a score from 0 to 100. You set a fit line and keep only the leads above it. The work moves from reading every profile to writing one good description and checking the scores against your own judgement.

The cost of manual profile research

Most founders and sales teams start by clicking through LinkedIn profiles one at a time. You open a page, read the about section, scan recent posts for a signal, then check whether the job title matches the person you want to reach. Two hours later you have forty names and a sore wrist.

The slow part is the deciding. Every profile asks the same question and gives a slightly different answer, so your judgement drifts. By the fiftieth profile you're approving people you would have skipped at the start, because the list has to be finished and nobody is checking your work.

Volume makes it worse. A list of 500 names means hours spent on people who were never going to buy: a vice president at a company with four employees, a manager in a country you don't sell to, a consultant who lists every industry on their profile.

Scoring flips the order. You write the rules once, apply them to the whole list, and open only the profiles that came out on top. Your attention goes to conversations instead of filtering.

Manual vetting

  • Open each profile
  • Read the about section and recent posts
  • Guess at company size and industry
  • Decide, then decide again on the next one
  • Paste the name into a spreadsheet

Scored list

  • Write the customer profile once
  • Score the whole list against it
  • Drop everyone below the fit line
  • Open only the profiles that passed

How to define a high-fit lead

Scoring is only as good as the description you give it. Ask for marketing managers and you get thousands of people, many of whom can't approve a purchase. The description has to carry everything you'd check by hand.

Start with hard boundaries: titles you want, titles you don't, industry, company size, location. These are the filters you'd apply in LinkedIn search anyway. Then add soft signals, which are phrases that suggest the person is dealing with the problem you solve. A head of operations at a company that just hired its first salesperson is a different lead from one at a company that has been the same size for six years.

The quickest way to start is your own website. Point the system at it and it reads what you sell and who buys it, which gives you a draft profile to edit. Editing a draft is faster than staring at a blank field.

  1. 1

    List the titles you want and the ones you don't

    Head of Growth, VP Marketing, founder. Exclude interns, freelancers, students and anyone whose headline says aspiring.

  2. 2

    Set the company boundaries

    Industry, size band and the countries you actually serve. B2B software companies with 11 to 50 employees is a usable boundary. Tech is not.

  3. 3

    Write down the buying signals

    Phrases that show the problem is live: hiring for a role you support, opening a second office, replacing a tool, complaining about a process in a post.

  4. 4

    Start from your website

    Paste your URL and let the system draft the profile, then cut anything that describes your product instead of your buyer.

Paste this job

Customer profile

Titles to include: Head of Growth, VP Marketing, founder, co-founder
Titles to exclude: intern, student, freelancer, recruiter, consultant
Industry: B2B software, agencies that serve software companies
Company size: 11 to 50 employees
Location: United Kingdom, Ireland, Netherlands
Signals: hired a sales lead recently, opened a second office,
posted about outbound or pipeline in the last month
Disqualifiers: sells to consumers only, fewer than 5 employees

Setting your fit line

The fit line is the score below which a lead isn't kept. It's the number you'll change most often, so change it slowly.

A high line gives you a short list of obvious matches. A low line gives you a long list you still have to read. Neither is wrong. It depends on how many messages you can send this week and how well the profile is written.

Start at 70 and leave it there for the first few hundred leads. Then look at what came through and what got cut, and move the line once. Moving it every day tells you nothing.

90 to 100

Score band
Matches every hard boundary and most signals
What it usually means
Open first, write the opening line from the signal

70 to 89

Score band
Matches the boundaries, thinner on signals
What it usually means
Normal sequence, read the profile before sending

50 to 69

Score band
One boundary is off, or the profile is thin
What it usually means
Read it yourself before it goes anywhere

Below 50

Score band
Wrong role, wrong company or wrong market
What it usually means
Leave it out

Testing the scores against your own judgement

Don't trust the scores for the first batch. Read the top of the list and the bottom of the list yourself and check that the system picks what you would pick.

The top is where errors cost you. If someone scored 95 and you can't see why, the description is too loose and the system is rewarding something you didn't mean to ask for. Find the phrase it latched onto and tighten it.

The bottom is where errors hide. Pull ten leads that fell below the line and read them. If two of them are people you'd have contacted, the profile is too narrow or the line is too high.

  1. 1

    Read ten from the top

    Check each one against your own list of boundaries. Note anyone who scored high for a reason you didn't intend.

  2. 2

    Read ten from the cut pile

    Look for people who fit but describe their job in different words. Those are the ones a rigid profile loses.

  3. 3

    Change one rule

    Edit the title list or the signals, rescore the same batch and compare. One change at a time is the only way to know what fixed it.

  4. 4

    Lock the profile

    Once the top and the cut pile both look right, stop editing and start sending.

Rescore the same batch every time you edit the profile. A new list hides whether the change helped.

Using the reason behind the score in your opening line

A score on its own doesn't help you write a message. What helps is the reason: the title that matched, the signal in the profile, the post they wrote last week. Keep that next to the lead so the first line can reference it.

Generic openers get ignored because they could have been sent to anyone. An opener that mentions the thing that made the person a lead reads like you looked before you typed.

Example

  1. You

    Hi Priya, saw you hired your first account executive last month while the team is still five people. Guessing pipeline is the thing you're building right now. We help founders get outbound running before they hire an SDR. Worth twenty minutes?

  2. Prospect

    Timing is decent. Everything comes through referrals at the moment and it's not predictable.

  3. You

    Same story I hear from most founders your size. Thursday afternoon work for a short call?

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 aren't kept, so the list you work from has already been through the filter described above.

Leads sit in lead groups, and any campaign can point at any group. You can import a CSV of LinkedIn profile URLs if you already have a list to score, or export a group to CSV. Connection notes can open with the reason that person became a lead, which is the signal you spent this guide learning to write down.

Frequently asked questions

Related

Get started with Kinlead

Prefer markdown? AI lead scoring vs manual profile research.md