Checklist for setting up AI lead scoring profiles
Define an ideal customer profile for AI lead generation with observable markers, inclusion lists, exclusion filters and a fit line that holds.
Get startedShort answer
To define an ideal customer profile for AI lead generation, move from vague categories to specific, observable markers. AI scores leads by comparing a profile's LinkedIn data against your requirements. The most accurate profiles combine a target website for context with a list of must-have job titles and exclusion filters that remove people who look right but aren't.
Identify your high-value markers
AI cannot guess your intent. Tell it to find business owners and it will score every person with Founder in their headline as a match, including people running a one-person blog or a local landscaping company.
Observable markers fix this. These are words, titles and company details that actually appear on a LinkedIn profile. If a person never writes the attribute down, no scoring model can read it, so the profile has to be built from what shows up on the page.
Read five profiles of customers you already closed. Copy the exact phrasing they use about their role and their company. That phrasing is your raw material, not the way your team describes the same person in a sales deck.
Vague markers
- Marketing managers
- Tech companies
- Decision makers
Observable markers
- Head of Demand Generation at B2B SaaS companies
- Series A to C fintech startups with 50 to 200 employees
- VP of Sales or Chief Revenue Officer
Avoid internal jargon. Use the words your customers use about themselves on their own profiles.
Build your inclusion list
Your inclusion list describes a 100-score lead. The tighter the intersection of role, industry and company stage, the less guessing the model has to do.
Start with the titles that carry signing authority, then add the markers that separate your best customers from everyone else working in the same industry.
- 1
List the titles that carry signing authority
Write the exact titles that buy your product: VP of Sales, SVP of Revenue, Head of Partnerships. Skip 'Sales Leader', which matches managers, directors and people who sold something once in 2019.
- 2
Name the industry the way LinkedIn does
Use the values in the Industry field on a profile, such as Software Development or Financial Services, rather than 'the tech space'.
- 3
Set company size or stage
Size is the cheapest proxy for budget. Give a range like 11 to 50 employees, or a signal like a recent Series B.
- 4
Describe the situation they're in
One sentence about what's happening at their company right now: scaling an SDR team, opening a second market, moving upmarket. This separates a good-fit company from a good-fit person at a bad time.
- 5
Add your website as context
Your homepage gives the model the language your product uses for the problem. It also catches markers you forgot to write down.
Set your exclusion filters
Inclusions find the right people. Exclusions remove the ones who look right in a keyword search and then waste your sending capacity.
A fractional CMO has the title you want and none of the authority. A consultant lists five client logos and no budget. Someone at a competitor has every keyword and will never buy. None of that is visible unless you name it.
Only exclude agencies and consultancies if you don't sell to them. Plenty of teams write a clean profile and then wonder why every lead is a reseller.
Consultants
- Role group
- Freelance, Fractional, Advisor, Consultant, Interim
Non-buyers
- Role group
- Student, Intern, Assistant, Coordinator, Aspiring
Competitors
- Role group
- Your named competitors, plus 'we help companies like yours'
Agencies and resellers
- Role group
- Agency, Consulting Firm, Reseller, Partner
Write the profile in the order the model reads it
The order matters less than the words, but a fixed order stops you leaving out the part that does the work. Write it as something a person could read out loud, not as a spreadsheet row.
Leave out two things: adjectives about your product, and anything the model can't see on a profile. 'Ambitious' and 'fast-growing' don't appear anywhere LinkedIn stores, so they only add noise to the score.
Read the profile back and ask whether a stranger could sort ten profiles with it. If two people would disagree on a borderline case, it's still too loose.
Paste this job
Target titles: VP of Sales, Head of Sales, Chief Revenue Officer Industry: Software Development, IT Services Company size: 50 to 300 employees Signal: hiring SDRs, recently raised Series A or B Situation: sales team is growing faster than their outbound process Exclude: Student, Intern, Aspiring, Freelance, Fractional, [competitor names]
Test and refine the scoring
Your first profile will be too broad or too narrow. Expect two or three rounds before the scores line up with your own judgment.
Sort your scored leads by score and open the top twenty. Every one you'd never contact is a bug in the profile, not in the lead. Find the exact string that fooled the model and add it to the exclusions.
Change one thing per round. If you edit three markers at once you won't know which one moved the scores, and you'll be back to guessing.
Example
- You
Why did this person get a 95? They're a student.
- Suggested reply
Their headline says 'Aspiring Sales Leader', which matched your 'Sales Leader' title marker.
- You
Adding 'Aspiring' and 'Student' to the exclusions, then re-running the same search.
When you get too few leads, loosen company size before you loosen titles. A right-title person at a slightly wrong-size company is easier to qualify than a wrong-title person at a perfect company.
Decide where the fit line goes
The fit line is the score below which a lead isn't kept. Set it too high and you'll spend the week hunting for leads instead of talking to them. Set it too low and your list fills with people you'd never message.
Pick a line, run one week of outreach, then read the replies. If every interested reply came from the top of the list, raise the line. If you run out of leads by Wednesday, lower it.
Give each setting a full week. A line that looks wrong after two days usually looks fine after ten, because the first batch of leads is never a fair sample.
Loose (60 and up)
- Fit line
- People with the right role who may work at the wrong kind of company
- What lands in your list
- Your title list is narrow and you want volume to skim by hand
Middle (75 to 85)
- Fit line
- Mostly on-target leads with a few borderline ones
- What lands in your list
- You're running steady outreach and can check the top of the list each morning
Tight (90 and up)
- Fit line
- Only people who hit nearly every marker
- What lands in your list
- Your offer is expensive and a bad-fit call costs you an hour
Doing this with Kinlead
Kinlead takes the profile you just wrote, as a description or from your website, and scores each person it finds on LinkedIn from 0 to 100 against it. Leads below your fit line are not kept.
Scored leads sit in lead groups. Any campaign can point at any group, and you can export a group to CSV or import a CSV of LinkedIn profile URLs when you already have a list.
- 1
Enter the profile
Paste your written description or give Kinlead your website URL during setup.
- 2
Set the fit line
Choose the minimum score a lead needs to be kept. Anything below it is discarded before it reaches a group.
- 3
Point a campaign at the group
Build your outreach campaign against the lead group and adjust the line as replies come in.
Frequently asked questions
Related
Get started with Kinlead
Prefer markdown? Checklist for setting up AI lead scoring profiles.md