How to exclude wrong leads with AI scoring
Stop wasting LinkedIn credits on bad leads. Use negative constraints and scoring thresholds to filter out students, freelancers, and competitors.
Get startedShort answer
Stop getting wrong leads by defining negative constraints in your lead profile. List who you don't want (like students or freelancers) and set a scoring threshold that automatically removes anyone who matches those traits.
Identify your lead leaks
Most people start LinkedIn outreach by listing who they want. They target titles like CEO or Marketing Director. The problem is that LinkedIn search is broad. If you search for CEO, you'll find the head of a 500 person company, but you'll also find a student who calls themselves the CEO of a one person project.
This gap happens because standard search tools look for keywords but ignore the context. They don't see that a person is actually a job seeker or a freelancer unless you tell them to look for those red flags. If you rely only on positive targeting, you'll spend your day messaging people who can't buy your product.
To fix this, you have to think about the opposite of your ideal customer. Think about the last ten leads that wasted your time. What did their profiles have in common? Usually, it's a specific set of keywords in their headline or about section that signaled they weren't a fit, but your search tool didn't know to ignore them.
When you define what a bad lead looks like, you create a safety net. This ensures that your outreach only hits people who actually have the budget and authority to make a decision. It's the difference between a list of 1,000 random names and a list of 100 high-intent prospects.
Positive targeting only
- Target: Founder
- Target: VP of Sales
- Target: Director of Operations
Positive and negative targeting
- Target: Founder (Exclude: Student, Aspiring, Freelancer)
- Target: VP of Sales (Exclude: Open to Work, Consultant)
- Target: Director of Ops (Exclude: Bootcamp, University)
Build a negative constraint list
To stop the wrong people from entering your pipeline, you need a list of terms that act as automatic disqualifiers. These are negative constraints. If a profile contains any of these words in the headline or about section, it's discarded regardless of the job title.
Categorize these by role, company type, and intent. This ensures you don't just block a word, but a specific type of person who doesn't buy your product. For example, blocking the word 'Consultant' might be too aggressive if you sell to consultants, but blocking 'Freelance' might be exactly what you need to find full-time employees.
You should also include your competitors in this list. There's nothing more embarrassing than sending a sales pitch to the founder of the company that does exactly what you do. Add their company names and the names of their key employees to your exclusion list.
Intent keywords are the most important. People who are 'Open to Work' or 'Seeking new opportunities' are often in a transition phase. While they might be a great lead in three months, they aren't in a position to buy software or services for a company today.
Role
- Category
- Student, Aspiring, Looking for opportunities, Freelance, Intern
Company
- Category
- Self-employed, University, Bootcamp, Academic, Research
Intent
- Category
- Open to work, Seeking new roles, Actively searching
Create a scoring rubric
A binary yes or no filter is often too blunt. Use a scoring rubric instead. This lets you weigh different attributes of a profile to see if they're a high quality lead or just a marginal fit.
Assign positive points for traits that indicate they'll buy and heavy negative points for red flags. The goal is to create a fit line. Anyone who falls below that line is automatically removed from the campaign. This allows you to be flexible. A CEO of a tiny company might be a bad fit, but a VP of a medium company is a great fit.
The rubric should be based on your actual sales data. If you notice that leads from companies with 11 to 50 employees close faster than those with 500 plus, give the smaller companies more points. This aligns your outreach with your most successful customer segments.
When you set your fit line, start high. It's better to have a smaller list of perfect leads than a huge list of mediocre ones. You can always lower the threshold later if you run out of people to message.
- 1
Assign positive points
Give points for must-have traits. For example, +20 points if they're a VP of Sales at a company with 50 to 200 employees.
- 2
Subtract for red flags
Apply heavy penalties for disqualifiers. Give -50 points if the word Student or Aspiring appears in their headline.
- 3
Set the fit line
Determine a minimum passing score. If a lead scores 40 or below, they're discarded.
Test your filters with a small sample
Before running a campaign to thousands of people, validate your rubric. If filters are too strict, you'll miss great leads. If they're too loose, you'll waste time on bad ones.
The best way to test is to run a raw search and manually score a small batch. This gives you a baseline of what the AI sees and how it reacts to your constraints. If you find that the AI is letting through 'Consultants' despite your negative constraint, you know you need to strengthen that specific rule.
Pay close attention to the 'edge cases'. These are the people who just barely pass your fit line. If they look like bad leads, your fit line is too low. If they look like goldmines, your fit line is exactly where it needs to be.
Once you've validated a sample of 50 to 100 leads, you can trust the system to handle the next 5,000. This prevents the common mistake of launching a massive campaign and realizing three days later that you've messaged 200 interns.
- 1
Pull a raw sample
Run a search for 50 profiles without any AI filters or scoring applied.
- 2
Manual audit
Apply your rubric to each profile. Note how many would have passed and how many would have been cut.
- 3
Adjust the threshold
If 90% of your sample was cut, lower the penalty for some terms. If 90% passed but half are bad, raise the fit line.
Always check the profiles that barely passed the fit line. They tell you where your rubric is weakest.
Refining your outreach based on fit
Once you've filtered out the noise, your messaging can be much more specific. When you know a lead is a high-score fit, you don't have to use generic language. You can speak directly to the pain points of a VP of Sales at a 100-person company because you've excluded everyone else.
High-scoring leads should get a different approach than medium-scoring leads. For your top 10% of leads, you might want to spend more time on the personalized note. For the middle tier, a standard but professional script works well.
If a lead is a perfect fit but doesn't reply, don't assume the filter was wrong. Even the best leads have bad timing. The filter ensures the person is capable of buying, not that they are buying this second.
Keep updating your negative constraints as you get replies. If a specific type of person keeps replying but saying 'I'm not the right person for this', find the keyword in their profile that identifies them and add it to your block list.
Example
- You
Hi Sarah, saw you're leading sales at Northwind. Most VPs at your stage struggle with lead quality. We fix that by filtering out the noise. Open to a quick chat?
- Prospect
Actually, I'm just starting a new role here and cleaning up our pipeline. This is timely. Let's talk.
Doing this with Kinlead
Kinlead automates this scoring process. You describe your ideal customer or provide your website, and the AI builds the profile. It then scores every lead from 0 to 100 and removes anyone who falls below the fit line.
Beyond standard search, you can use the Steal Customers agent. This agent watches competitor company pages and profiles, reads comments on recent posts, and turns commenters who look like buyers into leads. This ensures you're targeting people already engaging with your niche.
Outreach runs from your own account through Unipile, meaning Kinlead never stores your password. It sends at a human pace, with at most one action every 5 minutes, and stays within the daily limits and send windows you set for the lead's time zone.
- Leads are scored 0 to 100 and filtered by a fit line.
- The Steal Customers agent identifies buyers from competitor interactions.
- Outreach is sent at a human pace to protect your account.
- You can book a demo to see the scoring in action.
Frequently asked questions
Related
Get started with Kinlead
Prefer markdown? How to exclude wrong leads with AI scoring.md