How to filter out bad leads with AI scoring
Set a buyer profile, add red flag words and choose a fit line, so students, competitors and tire-kickers never reach your LinkedIn outreach.
Get startedShort answer
Write a buyer profile as hard rules: titles that hold budget, company type and size, problem signals, and an exclusion list of phrases that mark non-buyers. Score every profile against it and pick a fit line, then drop anything below that score instead of queuing it.
Decide who you will never message
A search for head of marketing in one city returns a few hundred profiles. A few dozen of them can sign a contract and have a reason to talk to you. The rest are agency account managers working on someone else's budget, people between jobs, students padding a headline, and a couple of competitors who will read your message and forward it to their team.
The pull is to send to all of them because the search took an hour to build. That is how an account picks up a low acceptance rate, and how a founder ends up on a call with someone who wants a job.
Before you touch a filter, write down who you will not message. Name the titles, the company types and the phrases people put in their headlines when they are not buying.
Easy to mistake for a lead
- A founder whose company is one person selling the same service you sell
- A marketing manager at a 5,000-person company where procurement picks vendors
- A profile with open to work in the headline
- An agency owner whose posts are all about partnering with people like you
- A student with a class project written up as a case study
Actually a lead
- Head of growth at a 60-person software company hiring its first demand gen manager
- Owner of a 12-person home services business who answers the phone himself
- Someone who commented under a competitor's post about the exact problem you fix
- A sales director whose team doubled this year and whose budget line already exists
Write the buyer profile as rules, not a persona
A buyer profile is only useful if you can check it against a profile page in a few seconds. One sentence per rule. If a rule needs a guess about how someone thinks, it belongs in a conversation later, not in a filter today.
Put hard requirements first: seniority, company type, size, location. Then the signals that this person has the problem you solve. Then the exclusions. Exclusions do more work than most people expect, because noisy profiles all share the same handful of phrases.
Write it in the words you would use to brief a new salesperson on their first morning. If you would not say it out loud to a colleague, it is too abstract for a filter to apply.
- Keep the exclusion list to the phrases you actually see on profiles this week
- Ten entries taken from real profiles beat fifty copied from someone else's template
- Put the rules in the same order every time so you can spot which one changed your list
Paste this job
Companies: software or services businesses with 20 to 200 employees, based in the US or UK. Titles: founder, owner, head of growth, head of sales, sales director. Not account managers at agencies, not recruiters, not students. Signals: hiring salespeople, posting about pipeline or reply rates, recently funded, or asking outbound questions in comments. Exclude: open to work, looking for opportunities, student, aspiring, freelance, coach, and anyone whose company sells the same thing we sell.
Pick a fit line instead of deciding yes or no
Scoring turns a yes or no decision into a ranking. Every profile gets a number, and you choose the number where your list starts. That matters because the middle of any list is full of judgement calls you do not want to make one profile at a time.
Start strict. Sending to fewer people costs you a slower week. Sending to the wrong people costs a reply you have to handle, a place in someone's notifications, and a signal to LinkedIn that your invites get ignored.
The band between obvious yes and obvious no is where the line does its work. Read that band by hand for the first week, then trust the cut.
85 and up
- Score band
- Right title, right company type, clear problem signal
- What it usually contains
- Send first, inside your normal send window
60 to 84
- Score band
- Right title with a weak signal, or a strong signal at a company you are unsure about
- What it usually contains
- Read ten by hand before you decide where the line goes
40 to 59
- Score band
- Adjacent roles, agencies, consultancies, people who might refer but will not buy
- What it usually contains
- Keep only if referrals drive most of your pipeline
Below 40
- Score band
- Students, recruiters, competitors, wrong country, wrong company type
- What it usually contains
- Drop them from the group so no follow-up goes out
Audit the lowest-scoring leads that survived
Filters fail quietly. You find out three weeks later when a follow-up lands on a student. The fix is a short weekly check you can do in ten minutes, not another tool.
Do the same check at the top of the list. If your highest-scoring leads are not people you would want on a call, the profile itself is wrong and moving the line will not help.
- 1
Open the bottom ten
Sort the group by score and open the ten lowest profiles that made it in. That is where the rules are weakest.
- 2
Find what they share
Look for a phrase, a title variant or a company type that repeats. One shared pattern is worth more than ten individual deletions.
- 3
Fix the rule, not the lead
Deleting one profile changes nothing. Adding the phrase to the exclusion list changes every search after it.
- 4
Move the line ten points
If the bottom ten are still weak after two fixes, raise the fit line and watch what disappears from the count.
Handle a wrong-fit reply without a debate
Some bad fits get through. A student, a recruiter, an agency owner who wants to resell you. When they reply, answer once and close the thread. Explaining why they are not a buyer rarely changes their mind and always costs you the time the filter was meant to protect.
Example
- Prospect
Thanks for the note. I'm finishing an MBA right now but this looks interesting. Can we talk this week?
- You
Appreciate you saying that up front. You're not the buyer for this, so I'll skip the call. If we open a research role later this year I'll send it your way.
- Prospect
Fair enough. Good luck with the search.
When the first sentence tells you someone cannot buy, reply once, close the loop, and take them out of the group before the next follow-up is due.
Keep the filter current as the list grows
Titles change. Companies cross your size line. A phrase that meant nothing last quarter turns up in half the headlines you see. A profile written in January will let noise through by June.
Once a month, read twenty random profiles in your group and ask whether you would send the same message to each one. Every profile you would skip is either a missing exclusion or a line set too low.
It works in the other direction too. If your best customer came from a title you excluded, take that exclusion out. Treat the filter as a draft you keep editing.
- Check the exclusion list when a new job hunting badge or headline phrase shows up on several profiles
- Re-read your size and industry rules after you win or lose two deals in a row
- Keep a note of every reply that made you think the score was wrong, and revisit it that week
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 your fit line are not kept.
The leads you keep live in lead groups, and any campaign can point at any group. You can export a group to CSV, or import your own list of LinkedIn profile URLs when you already have names to check.
Frequently asked questions
Related
Get started with Kinlead
Prefer markdown? How to filter out bad leads with AI scoring.md