Lead Scoring Algorithms

I replaced my client's 3-person SDR team and saved 100+ hours monthly by automating lead research and scoring with Clay. We created a process that automatically researches, enriches, and scores leads based on 6 key data points. In this post, I'll show you exactly how we built this system that anyone can implement. 1. Industry targeting: Instead of settling for broad categories like "Software" or "Technology," given by LinkedIn or major data providers, we set up an AI enrichment in Clay that reads websites and LinkedIn data to output specific niches like "HealthTech," "Martech," etc., making targeting much more precise. 2. Seniority filtering: We went beyond basic titles like Director or VP. Using Clay's AI enrichment, we analyze complete LinkedIn profiles to categorize prospects into Tier 1, 2, or 3 based on actual decision-making authority. You could feed the AI model their complete LinkedIn profile like their work experience, summary, or any other data available. 3. Persona identification: For complex segmentation, we set up Clay to identify hyper-specific personas. For example, we could identify "sales leaders managing 10+ SDRs in cybersecurity companies,". 4. Headcount qualification: Clay provides accurate headcount data from company LinkedIn profiles. We use this in the lead-scoring process to prioritize accounts within the client's sweet spot. 5. Intent signals tracking: Clay's AI Agent or native integrations can get critical signals like: - Job changes/Champion movements - Recent relevant posts - Hiring activity - Expansion/funding events - Tech stack changes - Event/conference participation 6. Lead scoring: To score leads with 100% accuracy, we use all the data points above and assign scores: - We pick scoring criteria based on the client's ICP (industry, headcount, seniority) - Set up simple comparisons (ranges for company size, exact matches for industries) - Assign points based on importance (right industry = 10 points, Tier 1 decision-maker = 10 points) - Clay adds everything up automatically This gives instant clarity on which leads deserve attention first. 7. CRM integration & data enrichment: Clay pushes everything directly to the CRM: - All enriched data flows straight to HubSpot or Salesforce - Custom variables map additional research findings to correct fields - Leads get tagged by priority score - The sales team only works on qualified, high-scoring prospects - Everything stays updated automatically with scheduled runs We also set up Clay to pull existing contacts from their CRM: - Dedupe them automatically - Re-enrich and score them based on fresh data - Push back with updated priorities - Let the team focus only on prospects most likely to convert This system now handles the same workload that previously took 3 people, while also delivering higher quality leads that convert better.

🔥 The lead scoring blueprint you wish you had 3 quarters ago. Built on Clay’s internal prioritization model, and it’s the same system we apply internally at SalesCaptain and with our clients. At SalesCaptain, we work with go-to-market teams across industries. And this prioritization matrix consistently drives impact. Why? Because it aligns sales, marketing, and growth around the ONLY two questions that matter: 1. Is this account the right fit? 2. Are they showing meaningful engagement right now? We walked through this in our recent webinar with Clay, where we shared a practical 2x2 matrix that drives everything from outbound plays to PLG routing to paid campaigns. 👉 If you only update one thing in your GTM motion for 2026, make it this. Here is how the "2026 GTM Prioritization Matrix" works ✅ Account Fit Score We look at indicators like: - B2B vs B2C - GTM motion (PLG + SLG) - Stack: Salesforce, HubSpot, Snowflake, Clay...etc. - ICP signals: size, vertical, hiring patterns - Similarity to past closed-won accounts ➡️ This tells us if this account worth pursuing at all? ✅ Engagement Score We track behaviors like: - Pricing page visits - LinkedIn engagement - Webinar attendance - Product activation - Positive replies to outbound ➡️ This tells us: are they leaning in, right now? Then we tier every account accordingly: 🟥 Tier 4: De-prioritize → Low fit, low engagement → No sales effort. Light nurture via PLG motion 🟦 Tier 3: Opportunistic Sales → High engagement, low fit → Route to PLG. Sales steps in only when signals are strong 🟨 Tier 2: Marketing Nurture → High fit, low engagement → Warm up with content, events, and thought leadership 🟩 Tier 1: Target Accounts → High fit, high engagement → AE multi-threading, dinners, BOFU ads, the full pipeline play This matrix now powers every core GTM workflow we run: * Clay-based scoring + tiering * CRM enrichment * Real-time Slack alerts * Tier-specific outbound messaging * Dynamic paid campaigns * Internal dashboards * Client workflows No matter if you’re running outbound, PLG, ABM (or all of the above) this system adapts and scales. We’ve deployed versions of it for category leaders, high-velocity startups, and bootstrapped teams. It works, it scales, and it gets your entire GTM speaking the same language. These strategies separate good GTM from elite GTM. Save this post and share it with your team.

Your lead scoring is broken. Here's the model that predicts revenue with 87% accuracy. Most B2B companies score leads like it's 2015. ┣ Downloaded whitepaper: +10 points ┣ Attended webinar: +15 points ┗ Opened email: +5 points Meanwhile, 73% of these "hot" leads never convert. Here's what we discovered after analyzing 10,000+ B2B leads: The leads scoring highest in traditional systems aren't buyers. They're information collectors. They download everything. Open every email. Click every link. But when sales calls? ↳ "Just doing research." ↳ "Not ready yet." ↳ "Send me more info." The leads that DO convert show completely different signals: They don't just visit your pricing page. They spend 8 minutes there, come back twice more that week, then search "[competitor] vs [your company]." They're not reading blog posts. They're calculating ROI and researching implementation. Activity doesn't equal intent. And that's where most scoring models fall apart. We rebuilt lead scoring from the ground up. Instead of rewarding every action equally, we weighted four factors based on what actually predicts revenue: ┣ Intent signals (40%) - someone searching "implementation" is closer to buying than someone downloading an ebook ┣ Behavioral depth (30%) - how someone engages tells you more than what they engage with ┣ Firmographic fit (20%) - perfect ICP match or bust ┗ Engagement quality (10%) - quality of interaction matters The framework is simple. The impact isn't. We map every lead to one of four tiers: ┣ 90-100 points → Sales gets them same-day ┣ 70-89 points → Automated nurture + retargeting ┣ 50-69 points → Educational content track ┗ Below 50 → Long-term relationship building No more dumping mediocre leads on sales and wondering why they don't follow up. Results after 6 months: ┣ Sales acceptance rate: +156% ┣ Sales cycle length: -41% ┗ Lead-to-customer rate: +73% The biggest shift wasn't the scoring model. It was the mindset. 🛑 Stop measuring marketing by MQL volume. ✔️ Start measuring it by how many MQLs sales actually wants to talk to. Your automation platform will happily score 500 leads as "hot" this month. But if sales only accepts 50, you don't have a volume problem. You have a scoring problem. Traditional scoring optimizes for activity. And fills your pipeline with noise. Revenue-predictive scoring optimizes for intent and fills it with buyers. If you'd like help with assessing your current lead scoring logic, comment "SCORING" and I'll get in touch to schedule a FREE consultation.

About 2-3 months back, I found out that one of my client’s page had around 570 people visiting the pricing page, but barely 45 booked a demo. Not necessarily a bad stat but that means more than 500 high-intent prospects just 'vanished' 🫤 . That didn’t make sense to me because people don’t randomly stumble on pricing pages. So in a few back-and-forth with the team, I finally traced the issue to their current lead scoring model: ❌ The system treated all engagement as equal, and couldn’t distinguish explorers from buyers. ➡️ To give you an idea: A prospect who hit the pricing page five times in one week had the same score as someone who opened a webinar email two months ago. It’s like giving the same grade to someone who Googled “how to buy a house” and someone who showed up to tour the same property three times. 😏 While the RevOps team worked to fix the scoring system, I went back to work with sales and CS to track patterns from their closed-won deals. 💡The goal here was to understand what high-intent behavior looked like right before conversion. Here’s what we uncovered: 🚨 Tier 1 Buying Signals These were signals from buyers who were actively in decision-making mode: ‣ 3+ pricing page visits in 10–14 days ‣ Clicked into “Compare us vs. Competitor” pages ‣ Spent >5 mins on implementation/onboarding content 🧠 Tier 2 Signals These weren’t as hot, but showed growing interest: ‣ Multiple team members from the same domain viewing pages ‣ Return visits to demo replays ‣ Reading case studies specific to their industry ‣ Checking out integration documentation (esp. Salesforce, Okta, HubSpot) Took that and built content triggers that matched those behaviors. Here’s what that looks like: 1️⃣ Pricing Page Repeat Visitors → Triggered content: ”Hidden Costs to Watch Out for When Buying [Category] Software” ‣ We offered insight they could use to build a business case. So we broke down implementation costs, estimated onboarding time, required internal resources, timeline to ROI. 📌 This helped our champion sell internally, and framed the pricing conversation around value, not cost. 2️⃣ Competitor Comparison Viewers → Triggered: “Why [Customer] Switched from [Competitor] After 18 Months” ‣ We didn’t downplay the competitor’s product or try to push hard on ours. We simply shared what didn’t work for that customer, why the switch made sense for them, and what changed after they moved over. 📌 It gave buyers a quick to view their own struggles, and a story they could relate to. And our whole shebang worked. Demo conversions from high-intent behaviors are up 3x and the average deal value from these flows is 41% higher than our baseline. One thing to note is, we didn’t put these content pieces into a nurture sequence. Instead, they were triggered within 1–2 hours of the signal. I’m big on timing 🙃. I’ll be replicating this approach across the board, and see if anything changes. You can try it and let me know what you think.

The days of MQLs and SQLs are over. Say hello to PQLs. In Product-Led Growth (PLG) strategies, the good old traditional metrics like MQLs (Marketing-Qualified Leads) and SQLs (Sales-Qualified Leads) don’t cut it anymore. For PLG SaaS companies, Product-Qualified Leads (PQLs) are way more effective, especially if you add a sales motion to your self-serve funnel. Why? Because PQLs are users who: ✅ Fit your ICP ✅ Have experienced product value ✅ Show buying intent Unlike MQLs/SQLs, PQLs don’t need to be convinced. They’ve already experienced your product’s value. Your job? Help them take the next step. The key to a successful sales motion for a PLG company is scoring these leads to focus your sales efforts on the most promising ones. To do so, there are 3 types of criteria you can focus on: 1️⃣ 𝗗𝗲𝗺𝗼𝗴𝗿𝗮𝗽𝗵𝗶𝗰/𝗙𝗶𝗿𝗺𝗼𝗴𝗿𝗮𝗽𝗵𝗶𝗰 𝗦𝗶𝗴𝗻𝗮𝗹𝘀 (𝗪𝗵𝗼 𝘁𝗵𝗲𝘆 𝗮𝗿𝗲) - Job title → Within your ICP? - Team size → Bigger teams = bigger revenue potential. - Email type → Business email = higher intent. 2️⃣ 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗨𝘀𝗮𝗴𝗲 𝗦𝗶𝗴𝗻𝗮𝗹𝘀 (𝗛𝗼𝘄 𝘁𝗵𝗲𝘆 𝘂𝘀𝗲 𝘁𝗵𝗲 𝗽𝗿𝗼𝗱𝘂𝗰𝘁) - Have they reached an activation milestone? - Do they use key features regularly? - Are they inviting colleagues to collaborate? 3️⃣ 𝗕𝘂𝘆𝗶𝗻𝗴 𝗜𝗻𝘁𝗲𝗻𝘁 𝗦𝗶𝗴𝗻𝗮𝗹𝘀 (𝗔𝗿𝗲 𝘁𝗵𝗲𝘆 𝗿𝗲𝗮𝗱𝘆 𝘁𝗼 𝗯𝘂𝘆?) - Viewed pricing page - Asked pricing questions in support - Booked a demo (strong intent) To target your PQLs, score each signal based on its impact. The higher the score, the hotter the lead. Sales can then prioritize the right outreach, targeting people who are already convinced of the value of your product but need a human touch to fully upgrade. 🛠 𝗧𝗼𝗼𝗹𝘀: CRMs like Hubspot, ActiveCampaign, or Customer.io allow you to create a custom scoring system. Just make sure your product data is properly synced, as it’s the cornerstone of a good PQL scoring. How are you identifying and scoring your PQLs? Let’s chat below! 👇

Managing $20M+ in media buying taught us that bad leads kill ROAS faster than bad creative. The old way was guesswork: → Basic CRM rules ("opened 3 emails = qualified") → Manual scoring that never updated → Sales chasing leads that never close For high-ticket verticals one garbage lead can wreck your month. Here's what we rebuilt: Dynamic scoring that learns daily: Our AI model ingests conversion data, campaign performance, and intent signals. No more static if/then rules. Full-funnel visibility: It tracks from first click to closed deal across ad platforms, CRM, and analytics. Real journey scoring, not single-touch guesses. Predictive weighting. The system discovers which behaviors actually predict revenue, scroll depth, session time, creative engagement, not just form completions. The impact: → Lower CAC (we're not bidding on junk traffic) → Sharper lookalike audiences → Sales teams chase only 80%+ close probability leads AI lead scoring became our quality gate between ad spend and wasted budget. If you're running serious paid media with static lead rules, you're leaving money on the table. Are you tracking which scored leads actually convert to revenue? #ads #metaads #marketing #marketingagency

Hot take: Lead scoring kinda sucks. I just finished deep research into lead scoring effectiveness. 98% of marketing-qualified leads never result in closed business. And only 35% of salespeople have confidence in their companies lead scoring accuracy. Zendesk tested 800 leads: → 400 "high-score" MQLs  → 400 random leads Conversion difference? ZERO. 98% of MQLs never close. 65% of reps ignore lead scores. But here's what actually works. Scoring your TAM. And here’s how you can build this in Clay. Step 1: Define Your ICP Criteria Pull your top 20 closed-won accounts. Find the patterns: • Revenue: $10M-$100M • Employees: 50-500 • Industry: SaaS, Tech, FinTech • Location: US/Canada • Tech Stack: Uses Salesforce • Growth: Funded or 20%+ headcount growth Step 2: Build Your Scoring Model Simple binary scoring (1 = match, 0 = no match): Criteria → Points → Weight • Revenue match → 1 point × 2 = 2.0 • Employee match → 1 point × 1.5 = 1.5 • Industry match → 1 point × 2 = 2.0 • Location match → 1 point × 1 = 1.0 • Tech stack match → 1 point × 1.5 = 1.5 • Growth signals → 1 point × 2 = 2.0 Total possible: 10 points Step 3: Score Your Entire TAM in Clay Import 5,000-50,000 accounts. Example A - Perfect Fit (10/10): • $50M revenue ✓ (2.0 points) • 200 employees ✓ (1.5 points) • SaaS company ✓ (2.0 points) • US-based ✓ (1.0 points) • Has Salesforce ✓ (1.5 points) • Series B funding ✓ (2.0 points) Example B - Partial Fit (5/10): • $200M revenue ✗ (0 points) • 300 employees ✓ (1.5 points) • SaaS company ✓ (2.0 points) • UK-based ✗ (0 points) • Has Salesforce ✓ (1.5 points) • No growth signals ✗ (0 points) Step 4: Assign Tiers & Take Action • Tier 1 (8-10 points): Dedicated SDR, personalized outreach  • Tier 2 (5-7 points): Coordinated campaigns  • Tier 3 (3-4 points): Marketing automation only  • Tier 4 (0-2 points): Exclude from outbound Step 5: Layer Intent Data Add a 30% weighted Intent Score: • Website visits • Competitor research • LinkedIn content • Topic consumption Final Priority Score = (Fit × 70%) + (Intent × 30%) Most lead scoring waits for someone to download a whitepaper. TAM scoring identifies your best accounts on Day 1. Comment "TAM" and I'll send you the full report. ✌️ P.S. Even HubSpot (who sells lead scoring) admitted their own system didn't work and built something else. Mark Roberge, former CRO at HubSpot, said: "At HubSpot, we tried the lead scoring approach, but ran into [problems]. We evolved to implement an alternative approach." 

Is your lead scoring still stuck in the pre-AI era? Traditional lead scoring gives you a number: "This lead is a 7 out of 10." or a "Medium Fit". Clean. Deterministic. Easy to route and prioritize. But here's what I keep running into with clients: SDRs look at that "7" and have no idea what it actually means. The score works for sorting, but it fails at decision-making. -- The observation: Most scoring models combine database filters (headcount, industry) with some AI-generated attributes (intent signals, "strength of social media presence," engagement propensity). You get a weighted score. But the rationale for the score is abstracted away. Your SDR sees a 4 and a 7, knows they should call the 7 first, but has zero context for how to approach either conversation. What if lead scoring needs two layers instead of one? ↳ Quantitative score (the "7/10") - for routing and prioritization ↳ Qualitative context (the "why") - for understanding and action Keep the first layer mostly deterministic - company size, technographics, behavioral signals, AI-generated attributes, whatever your model weights. The second layer is where AI actually helps. Not by making the score "better," but by explaining it with real data: Example context block: Score: 7/10 Recent activity: - CRO posted on LinkedIn yesterday about "evaluating new sales tools" - Engineering lead attended our webinar 2 weeks ago Company signals: - Series B raised 6 months ago - Hiring 3 SDR roles in past 30 days Timing context: - Q4 budget cycle likely starts in 2 weeks - No demo requests but high research activity Override signals: - Engagement spike suggests urgency despite mid-tier score - Multi-department interest (sales + eng) suggests internal testing -- The shift this enables: 1. Agency - SDRs and agents can override when context reveals the score misses something 2. Transparency - Everyone sees the same reasoning 3. Better judgment calls - That 6-score lead who just posted about their pain point might be more valuable than the 7 who downloaded something 3 months ago -- Future state thinking: This context layer doesn't have to be static. Imagine the context is updated periodically and by real-time events. And then you give an agent decision rights based on context thresholds: "If a lead's engagement score spikes in a short period of time and they exhibit key buying signals, send personalized outreach." The agent isn't making the scoring decision. It's acting on the combination of deterministic score + contextual signals that suggest the timing is right. -- As we move to an era of abundant intelligence, we don't have to abstract away all the details and tokens. We have AI for that now. Ironically, we can now architect flows that feel less rigid and more human by removing humans from the process. Anyone else experimenting with this? What am I missing?

How a startup drove 3,000% lift in sales conversions for enterprise bank customers. I met with the Viktoria Izdebska, CEO of Octrace, a startup that finds and prioritizes lead through real trigger events that actually drive sales conversions. Most GTM teams are drowning in correlated signals that feel meaningful but don’t actually cause conversions. Octrace did something a bit different. Viktoria came from the hedge fund industry so she knew that correlation did not indicate causation and was in search for causal triggers. She applied that same learning to lead scoring in B2B. Octrace built a system to identify causal trigger events — the kind of things with enough explanatory power that a human seller would say: “Yeah… if that happened, I’d absolutely call this lead today because it means they have a real pain.” Their identification pipeline was: 1. Identify the right signals Viktoria worked with the bank’s head of sales to determine the exact real-world events that actually matter. She also used an LLM and data from previous customers to assist in the discovery and creation of which events to track Not just “job postings” or “web visits” but things like: - A CEO turning 60 (succession triggers) - Keywords in financial statements that imply asset liquidation - A company opening a new manufacturing plant Signals grounded in reality 2. Collect those signals at scale Public, semi-public, and scraped sources across structured + unstructured data 3. Run each signal through an LLM agent to determine if it’s a “hit” Each incoming data point was evaluated in real time: “Is this the thing we care about? Does it match the trigger condition?” 4. Let another LLM score the combination of signals Not classical ML. Not random forest. Not feature engineering. Just a smart, explainable LLM evaluating causation. 5. Process the signals in real-time for the model to compute 6. Compare outcomes vs a control list Because they had access to CRM conversion data, they could backtest and refine signal selection and weighting. The result was a lead list that was explainable and outperformed their own lead list to reps by 3,000%. Customers loved them. Viktoria and I both came from a finance background. She was at a hedge fund prior to her company and I was on the trading floor. We both realized that models can, over time and with human guidance, discover and weight signals better than humans, and outperform intuition through backtesting - a concept finance traders have been using since the 1990s.

I think I know why a lot of signal models fail. And no, Clay or Seam AI don't inherently fix this (much as I love those teams) Simply put, no one tests scoring models. What is 100 lead score/account score SUPPOSED to predict? Theoretically, that score should mean you are very likely to book a meeting or create an opportunity. But in practice they get created because someone in marketing or revops looks at the data and says "I feel like having Salesforce should be +5 to account score." And that's it. We tested a couple scoring models recently. The first didn't show signals for 60% of the client's customers. The second predicted wins and losses directionally, but our R squared was 10%. The entire model only explained 10% of the win rate. Not great. But we want more. Here's how we're solving it. Step 1: Get a 90 day snapshot of your GTM data into a Claude Code repo. CRM, sales calls, email activity, everything. Side note: just from having 90 days of sales call conversations in one place, you can probably rewrite your entire GTM playbook with Claude Code. Cold call scripts, discovery frameworks, objection handling, competitive positioning. If you get everything in there, all of your playbooks can get updated fast. That alone is worth the exercise. But for the signal model specifically: Group it into cohorts. Contacts touched, responded, booked a meeting, created an opp, won, renewed, expanded. Step 2: Create samples. Edge cases (long cold calls, high or low ACV, fast or slow sales cycles) plus stratified samples across titles, industries, company sizes. Step 3: Build a detailed corpus for each sample. Website, LinkedIn, job posts, company news. Run qualitative analysis across each cohort to find specific signals from unstructured data. What do winners and losers look like at each stage? Positive and negative signals. Step 4: Turn those signals into an enrichment pipeline and test at scale across the full data snapshot. Look at p values (do these signals actually predict movement in the customer journey), R squared (how much movement do they explain), and coverage (how many accounts that advance or don't have this signal). If anything scores too low, you need more or better signals. Step 5: Build the scoring model. Automate the enrichment and scoring pipeline so it runs without you. Push signal summaries and sources to reps so they can personalize outreach. Roll it out with a deck showing scored data and real examples of winners and losers at each stage. Now you have a signal model your reps understand, you can prove it works, and since all your data is already in Claude Code you can rewrite any playbook that doesn't perform. We'll put a guide out for this eventually. But if you want to build a signal model, fix your playbooks, or deploy Claude Code in your org, DM me "code." Drop questions in the comments, I'll do my best to answer them.