Lead Scoring Models That Work: A Practical Framework
How to build lead scoring models that work: fit and intent signals, point thresholds, and a review loop that keeps scores honest over time.
- Score fit and intent separately; blended scores misroute effort.
- Start with a transparent point model you can explain in one page.
- Set the MQL threshold jointly with sales using real lead samples.
- Grade the model quarterly against actual pipeline outcomes and add decay.
Separate Fit From Intent or the Model Lies
A single blended score hides the two questions that matter: could this company buy, and are they showing signs of buying now. Fit covers firmographics like industry, company size, and role. Intent covers behavior like pricing page visits, demo requests, and email engagement.
Keep them as two scores, or at minimum two clearly separated point pools. A perfect-fit account with zero activity needs outbound, not a sales-ready flag. A tiny company binge-reading your docs needs nurture, not an SDR call. Blending the two produces confident scores that point reps at the wrong work.
Start With a Simple Point Model You Can Explain
Resist the urge to start with machine learning. A points-based model with ten to fifteen weighted signals is transparent, debuggable, and good enough for most B2B teams. Weight high-commitment actions like demo requests far above passive ones like email opens, and add negative points for disqualifiers like student email domains or competitor names.
Write the model down in one page that a new SDR can read. If you cannot explain why a lead scored 85, sales will not trust the number, and an untrusted score is just a column nobody sorts by.
Set Thresholds With Sales, Not for Sales
The MQL threshold is a contract, not a math result. Sit with sales leadership, pull twenty recent leads at various scores, and ask which ones they would genuinely want to call. Set the threshold where their yes rate gets high, then agree what happens to everything below it.
Add decay so intent scores fade over weeks of inactivity. A lead that was hot in January is not hot in June, and without decay your database slowly fills with permanently qualified ghosts.
Review the Model Like a Forecast
A scoring model is a prediction, so grade it like one. Each quarter, compare score bands against actual outcomes: what share of high scorers converted to opportunities, and which closed-won deals scored low at handoff. Low-scoring winners are your best source of missing signals.
Route the whole thing through one signal layer so scoring, routing, and reporting read the same fields. When each system computes its own version of engagement, the model can never be audited, and every debate about lead quality becomes a debate about whose data is right.
- Score fit and intent separately; blended scores misroute effort.
- Start with a transparent point model you can explain in one page.
- Set the MQL threshold jointly with sales using real lead samples.
- Grade the model quarterly against actual pipeline outcomes and add decay.
Frequently asked questions
What is a good lead scoring model for B2B?
A good B2B model scores fit and intent separately, uses ten to fifteen explainable signals, and sets thresholds jointly with sales. Fit reflects firmographics like size and industry, intent reflects recent behavior like demo requests and pricing views. Simplicity beats sophistication until you have enough volume to validate anything fancier.
Should we use predictive or points-based scoring?
Start points-based. Predictive models need meaningful conversion volume to train on and are hard to debug when sales disputes a score. Most teams get the majority of the value from a well-maintained point model, and can layer predictive scoring on later once the underlying data is clean and trusted.
How often should lead scores be reviewed?
Review the model quarterly and the thresholds whenever conversion patterns shift. The quarterly review should compare score bands to actual opportunity and win outcomes, then adjust weights for signals that over or underperform. Skipping reviews is how scoring models drift into decoration.
Why does sales ignore our lead scores?
Usually because the score has burned them before, or nobody can explain it. Rebuild trust by making the model transparent, setting the threshold with sales using real examples, and publicly grading the model against outcomes each quarter. Trust follows accuracy plus explainability, not another training session.
Liked this? Get the next play in your inbox.
One signal-driven GTM play every week. No fluff, no spam, unsubscribe anytime.
Operator-built
Built by someone who runs the playbook, not an agency reselling labor.
You own it
Your data, your CRM, your infrastructure. The system is yours.
No lock-in
Start with a free audit. No multi-month retainer to find out it works.
Privacy-first
Your data stays yours. We pen-test our own funnel before we touch yours.
▸ STOP READING. START PLAYING.
Don't just read about it. Drop your site below and see the revenue you're leaving on the table, live.