Updated September 28, 2026 | 11 min read
Updated September 28, 2026 | 11 min read
Ready to launch?
Convert strategy into pipeline
Launch your first outbound campaign with verified leads, AI sequences, and multichannel outreach from one platform.
ICP Scoring: The 5-Step Framework That Tells You Which Leads Are Worth Your Time
<article>
Nearly four out of five marketing leads never become sales. That 79% figure comes from MarketingSherpa, and it’s been quoted for years because nobody has managed to disprove it. The same research found that 61% of B2B marketers pass every lead straight to sales, and only 27% of those leads turn out to be qualified. I’ve sat in enough pipeline reviews to recognize what that looks like in practice: a rep with 400 open accounts, half of which were never going to buy, working them in the order they landed in the CRM.
Nearly four out of five marketing leads never become sales. That 79% figure comes from MarketingSherpa, and it’s been quoted for years because nobody has managed to disprove it. The same research found that 61% of B2B marketers pass every lead straight to sales, and only 27% of those leads turn out to be qualified. I’ve sat in enough pipeline reviews to recognize what that looks like in practice: a rep with 400 open accounts, half of which were never going to buy, working them in the order they landed in the CRM.
In this post, I’ll show you how to build an ICP scoring model that fixes the order. Five steps, the criteria that actually correlate with closed-won deals, how to layer intent on top of fit, and how to keep the model from rotting six months from now.
What is ICP scoring and why your pipeline depends on it
Ranking leads by fit is the whole idea. ICP scoring assigns point values to the traits your best customers share (industry, company size, tech stack, geography) and totals them into one number that tells your team where to spend time first. Those traits come from your Ideal Customer Profile, which is the firmographic and behavioral pattern behind your closed-won deals.
Here’s why it matters beyond tidy CRM fields. Most pipeline gets burned on accounts that never had a real shot. A scoring system doesn’t fix a broken lead supply, but it does fix the routing, so reps stop working accounts that look promising on paper and never had the profile to buy. The effect shows up in the numbers you report to your board: win rate on sales-qualified opportunities, sales cycle length, and average deal size all move when the wrong accounts stop entering the funnel. RAIN Group’s research on top-performing sales organizations, based on a survey of 472 sellers and sales executives, puts the average win rate at 47% once a proposal is actually delivered. Getting to that stage with the right accounts is the entire game.
Think of ICP scoring as turning a static slide deck into a live filter inside your CRM. Once it’s built, every new lead gets scored automatically, and your team knows within seconds whether an account deserves a personalized sequence or a pass.
ICP scoring vs. lead scoring
These two terms get mixed up constantly, but they answer different questions. ICP scoring asks whether a company resembles your best customers, regardless of anything they’ve done. Lead scoring asks whether someone at that company is showing signs of active interest right now.
ICP scoring | Lead scoring |
|---|---|
What it measures | Company fit to your ideal profile |
Data used | Firmographic, technographic attributes |
When it’s set | At lead creation or enrichment |
Best for | Prioritizing which accounts to pursue |
ICP scoring is proactive. It grades an account before any engagement happens, which means you can define a good fit before you spend a dollar finding one. Lead scoring only becomes useful once someone has actually interacted with your content or site.
So build ICP scoring first, then layer lead scoring on top for timing. Together they give you fit and readiness in one view.
The 4 criteria that predict closed-won deals
Your scoring criteria should come from patterns in your own closed-won data, not from a template you found online (including this one). The categories below hold across most B2B companies. The weights won’t.
Firmographic fit covers the structural facts: industry, headcount, revenue range, geography, growth stage. Wrong company size or wrong industry usually means no deal, no matter how much interest a prospect shows.
Technographic fit looks at what tools a company already runs. A prospect using a CRM you integrate with (say, HubSpot or Salesforce) tends to be a warmer fit than one with no detectable stack at all.
Behavioral and intent signals change constantly, and they tell you whether an account is in-market right now. The events worth tracking are the ones that precede a purchase:
- A company posting for a Head of Outbound or a RevOps Manager is investing in growth infrastructure, and that role usually needs tooling soon.
- A recent funding round often triggers new tool purchases within months.
- Repeated visits to your pricing or product pages.
- Likes, comments, or follows from key contacts on LinkedIn.
These are the same event types lemlist’s Intent Signal Agents monitor daily, which matters later when you connect scores to campaigns.
Negative scoring is the part most teams skip. A solid model subtracts points too. Define what automatically disqualifies a lead: wrong industry, company too small to use your product, competitor employees, or no verified contact data.
How to build an ICP scoring model in 5 steps
1. Analyze your closed-won deals
Pull your last 20 to 30 closed-won deals from your CRM. Look for the traits that repeat across them, and note how often each one shows up. Skip this step and you’re guessing at weights later.
2. Select and weight your criteria
Pick the attributes that correlated most strongly with wins, then assign points. A model with 15 attributes sounds thorough. In practice it means reps score inconsistently and leads stall.
A simple starting point might look like this:
Criterion | Points |
|---|---|
Industry match | 25 exact, 10 adjacent |
Revenue range | 20 ideal, 10 close |
Employee count | 15 ideal, 5 close |
Tech stack overlap | 15 |
Geography | 10 primary market |
Negative signals | −15 |
Weighting matters more than the number of criteria. A factor that strongly predicts wins deserves more points than one that’s loosely correlated with them.
3. Collect and enrich the data
A scoring model is only as good as the data feeding it. An empty field scores as zero by default, so a genuinely good account gets quietly demoted for no real reason. I’ve watched a team dismiss an entire segment because their employee-count field was blank on 40% of records.
Before scoring, every record needs complete firmographic, technographic, and contact data. Tools like lemlist’s AI Agentic Enrichment pull structured data from LinkedIn, websites, and CRM records to fill in exactly the fields your model relies on.
4. Calculate scores and assign tiers
Raw numbers mean little to a rep in the middle of a workday. Turn them into tiers instead:
Tier | Score range | Action |
|---|---|---|
A | 80–100 | Priority accounts, immediate outreach |
B | 60–79 | SDR-led nurture or automated sequences |
C | Below 60 | Deprioritized or excluded from outbound |
Set a minimum score threshold for a lead to count as sales-qualified, and hold reps to it.
5. Connect scores to routing and campaigns
A score sitting in a spreadsheet helps nobody. It needs to live in CRM fields so every rep applies it the same way. In lemlist, you can filter leads by ICP criteria from a 600M+ lead database, enrich them, and push qualified accounts straight into multichannel sequences, so the score triggers outreach instead of sitting idle.
One more thing before you call it done. Measure the model by tier. Track win rate, average deal size, and cycle time for Tier A against Tier B over a full quarter. If your A accounts don’t close faster or bigger, the tiers are decorative, and you’ll know exactly which weights to revisit.
Choosing an ICP scoring model type
There’s more than one way to structure this, and the right choice depends on how much closed-won data you have.
Point-based scoring assigns fixed points per criterion. Simple, easy for reps to follow, and a good starting point.
Weighted scoring builds on that same structure but applies multipliers based on importance. It’s more accurate, though it needs enough historical data to justify the weights you choose.
Tiered scoring skips numbers entirely and places leads into A/B/C/D buckets based on must-have and nice-to-have criteria. Useful for teams that prefer visual categories over math.
Predictive, AI-driven scoring uses machine learning to surface patterns in historical deal data that humans tend to miss. It needs a larger dataset to work well. Tools like Aviso, Koala, and 6sense offer this, and enrichment agents (lemlist’s included) can feed them the structured data they depend on.
Tip: if you have fewer than 50 closed-won deals in your CRM, a point-based model is the practical starting point. Predictive scoring only works once there’s enough signal to learn from.
Combining ICP fit with intent signals
A high ICP score tells you an account fits. It doesn’t tell you the account is ready to buy this week. Fit is necessary, and fit alone still leaves you guessing on timing.
Combining fit with intent produces a composite score that ranks leads by quality and readiness together. Signals worth tracking include:
- Website visits identified through IP matching
- Job changes among key contacts
- Funding events or M&A activity
- Tech stack changes (new installs or removals)
Fit should carry more base weight, since it rarely changes. Intent acts more like a multiplier.
Here’s what that looks like in practice. Say you have a 68-point account: right industry, right geography, headcount slightly under your ideal band. Tier B. It sits in a nurture sequence, and on any given Tuesday no rep thinks about it. Then the company announces a Series A and posts a job for a Demand Gen Lead in the same week. That account now has budget, a hiring mandate, and a reason to buy tooling in the next quarter. The fit score hasn’t moved a point, but the intent multiplier pushes it above the Tier A accounts that have shown nothing for three months, and it should be in a sequence that same day. Run the reverse case too: a 92-point account with zero activity stays on a steady cadence rather than getting hammered. And a lead with heavy intent but poor fit stays deprioritized, no matter how many pricing-page visits it racks up.
lemlist’s Intent Signal Agents track these signals daily and auto-route matching leads into campaigns with context-aware messaging, so the multiplier fires without anyone watching a dashboard.
ICP scoring tools compared
No single tool covers every part of this, so it’s worth knowing the landscape before picking a stack.
Category | Examples | Strength | Limitation |
|---|---|---|---|
CRM-native scoring | HubSpot, Salesforce | Built-in fields, easy if you’re already deep in the CRM | Limited on enrichment and intent data |
Standalone intent platforms | Bombora, 6sense, Demandbase | Strong intent and account identification | Often pricey, and usually paired with a separate outbound tool |
Data enrichment + outbound platforms | lemlist, Apollo, ZoomInfo | Lead data, enrichment, and outreach in one workflow | Predictive modeling is lighter than in dedicated AI tools |
AI scoring tools | Koala, Aviso, Madkudu | Predictive scoring on historical deal data | Needs large datasets and an ops team to maintain the models |
The real question is whether you want scoring and outreach in one place, or spread across tools that need to be stitched together.
Keeping your model accurate over time
Most scoring models fail quietly through decay: buyer behavior shifts, but the static score doesn’t move with it. Recalibrating once a quarter tends to catch this before it costs you pipeline.
The process is straightforward. Re-run your closed-won analysis, then compare scored tiers against actual conversion rates. If Tier A isn’t outperforming Tier B, the weights need adjusting.
A cadence I’d recommend: block 90 minutes in the first week of every quarter. Pull the deals closed since the last review, re-check which attributes showed up in the wins, and adjust one or two weights at most. Change six things at once and you’ll never know which one worked. Then re-enrich the records that have been sitting in your database since the last cycle, because a lead scored in January may be a different company by July. Enrichment agents can re-pull that data automatically, which saves someone the misery of refreshing records by hand.
Markets shift, products evolve, and the traits that defined your best customers six months ago might not hold today.
FAQs about ICP scoring
What is a good ICP score threshold for outbound sales?
There’s no universal number. Most teams treat 80+ on a 100-point scale as Tier A, 60-79 as Tier B, and anything lower as deprioritized, then adjust the cutoff quarterly based on actual conversion per tier.
Can ICP scoring work for both inbound and outbound leads?
Yes. It measures fit to your ideal profile, not how the lead arrived, so inbound and outbound leads get scored on the same criteria.
How many criteria are too many in an ICP scoring model?
Most working models use five to eight. Fewer than four won’t differentiate leads well, and more than ten tends to make scoring inconsistent across reps.
Is account scoring the same as ICP scoring?
They overlap without being identical. ICP scoring measures one dimension, fit, while account scoring often folds in intent and engagement signals alongside it to estimate how likely a company is to buy.
Over to you
ICP scoring turns your ideal customer profile from a slide into a live filter that drives every outbound decision. The model doesn’t have to be complicated: start with your closed-won data, pick five to seven criteria, weight them, enrich your leads, and route the best ones into campaigns that match the moment. Then measure by tier and adjust next quarter.
If you want to build this workflow in one place (lead database, enrichment, intent signals, and multichannel outreach), lemlist is rated 4.6/5 across more than 2,000 reviews on G2, Capterra, and Trustpilot. Start a 14-day free trial. No credit card required.
</article>
</article>
Share this post