Loading tutorials…
Loading tutorials…
Lead scoring is the most-built and least-used feature in HubSpot. Most setups generate scores that nobody references because they don't match how sales actually qualifies. Here's the model that earns sales trust — and the decay rules most owners skip.
Who this is forMarketing ops leads at companies generating 200+ leads/month where sales can't work all of them. If your sales team currently ignores the MQL flag, that's a lead-scoring problem (or a definition problem).
What you'll need
Step 1
Lead scoring is a translation of 'qualified' into points. If sales and marketing disagree on qualified, the score will be wrong forever.
Schedule a 45-min call with sales leadership. Bring the lifecycle stage definitions you wrote (see related tutorial).
Ask sales: 'When a lead becomes an MQL, what do they need to do or be for you to call them within 24 hours?'
Capture the answer in two buckets: Fit criteria (company size, industry, role, geography) and Behavior criteria (specific actions like booked demo, visited pricing 3x, replied to email).
Write the criteria down. This becomes the spec for your scoring model.
Common trap: sales says "anyone good." Push back. Force specificity. Without specificity, lead scoring will not work.
Step 2
HubSpot has two lead scoring tools. Pick based on complexity needs and tier.
Settings (gear icon, top-right) → Properties → Contact properties → search "HubSpot Score" (default scoring) or "Create new score property" (custom).
HubSpot Score (default): one positive + one negative score, additive. Good for most teams under 5K leads/month.
Custom score properties (Marketing Hub Pro+): create multiple scores for different scenarios (e.g., 'Enterprise score' vs 'SMB score'). Use when one model can't capture your segments.
Predictive lead scoring (Pro+, requires 100+ customers and 6+ months data): HubSpot AI/Breeze trains a model on what your historical conversions look like. Best for mature programs.
Start with HubSpot Score (default). Move to custom or predictive only when you outgrow it.
Step 3
Assign points for actions and attributes that signal qualification. Calibrate weights so the highest-value action gives the most points.
Click HubSpot Score → "Add a positive attribute."
Behavior weights (typical): Booked a demo = 50, Viewed pricing page = 15, Submitted contact form = 25, Clicked a marketing email = 5, Visited 5+ pages in a session = 10.
Fit weights (typical for B2B SaaS): Company size 100-500 = 20, Role contains "Director/VP/CEO" = 15, Industry = your target = 10, Country in your target list = 5.
Total a "fully qualified" lead should score in the 80-120 range. A "minimum MQL" should be at 50-70.
Avoid weights that compound illogically (e.g., 100 points for visiting any blog post). Sales will see leads ranked high that have done nothing meaningful.
Step 4
Negative score handles disqualifying signals. Decay handles stale activity. Most DIY models skip both and end up with permanent inflated scores.
Add negative attributes: Lifecycle stage = Customer (we already have them, -20), Role = student / intern (-20), Country not in target list (-10), Email contains "test" or "spam" (-50), Unsubscribed (-30).
For decay, HubSpot Score doesn't natively decay — you must build it via workflow. Create a workflow: enrollment = HubSpot Score is known, action = wait 30 days, then "subtract from HubSpot Score" by 10-20 points. Re-enroll every 30 days.
Without decay, a lead who scored 75 in March still scores 75 in October even if they have done nothing since. Sales will keep getting "MQL" alerts for stale contacts.
Test the decay: pick a contact, manually set score to 100, wait 31 days, verify it dropped to ~80.
Step 5
MQL threshold is the score at which a contact becomes an MQL. Set too low = sales gets flooded with bad leads. Set too high = real opportunities ignored.
Pull 30 days of historical leads. For each, manually rate "would I call this today" (yes/no).
Score each in your model. Plot the score vs the yes/no rating.
The MQL threshold should be the score at which 70%+ of leads above it would be "yes." Typically lands in the 50-80 range for first-pass models.
In the "Promote to MQL" workflow (see Lifecycle Stages tutorial), set enrollment criteria = HubSpot Score >= [threshold] AND Lifecycle Stage = Lead.
Don't set the threshold below 50 in the first 30 days. Sales gets overwhelmed and trust crashes. Always err high on the first calibration.
Step 6
Lead scoring drifts. Buying behavior changes. The model needs quarterly recalibration based on what's actually closing.
Every quarter, pull a sample: 20 MQLs that closed-won, 20 MQLs that closed-lost, 20 leads that scored under threshold but converted.
Look for patterns: do closed-wons share traits the model missed? Do losses share traits the model rewards?
Adjust weights based on findings: increase points for the strongest closed-won signal, add a negative score for the strongest loss signal.
Document each change in a model-version log. "Q2 2026: increased weight for pricing page view from 15 → 20 because 80% of closed-wons hit pricing 2+ times pre-call."
Without calibration, the model becomes folklore. With calibration, it gets sharper every quarter.
Step 7
Beyond MQL, build an instant-alert workflow for the top 5% of scores or specific behavior combos.
Workflow: enrollment criteria = HubSpot Score increased by 30+ in last 7 days OR (booked demo + visited pricing in same session).
Action: in-app notification + Slack to the contact owner + create a high-priority task due same day.
This catches the leads where time-to-call is the conversion variable. Calling within 5 minutes vs 1 hour can double connect rate.
Limit volume — only the top 5-10% of intent signals should trigger this alert. Otherwise, sales habituates and ignores.
Common mistakes
Building the score without sales in the room
What goes wrong: Marketing ships a 50-attribute scoring model. Sales sees the first MQL alert, the contact is a wrong fit, and from that point sales never trusts the score again. You're paying $890/mo for Marketing Hub Pro but using 15% of features because the scoring engine sits idle.
How to avoid: The sales alignment call is the first step, not the last. Without sales-side agreement on what qualified means, the model has nothing to translate into points.
No score decay
What goes wrong: A lead scores 80 in February by attending a webinar. They take no further action. In October they're still showing as a high-priority MQL. Sales calls them, the lead is cold, sales blames the scoring model.
How to avoid: Build a decay workflow: subtract 10-20 points every 30 days of inactivity. Re-enroll continuously. Without decay, scores are permanent records, not current signals.
Setting MQL threshold too low
What goes wrong: Threshold = 30. Sales gets 400 MQL alerts in week 1. They work the top 50, realize most are bad, and stop opening the rest. The model gets labeled 'broken' permanently.
How to avoid: Always err high on first calibration. 70-80 is safer than 30-50. You can lower the threshold once sales is converting MQLs above it consistently.
Giving points for low-signal actions
What goes wrong: 5 points for visiting the homepage. 5 points for any blog page. 10 points for opening an email. A scroll-from-Google visitor scores 25 in a session. Sales sees them as MQL, calls them, no idea who they are. Model loses credibility.
How to avoid: Reserve positive points for actions that distinguish qualified from unqualified. Pricing page, demo booking, contact form, multi-session engagement. Skip 'pageview of anything' as a scoring action.
Not testing the model on historical data before activating
What goes wrong: You build, activate, and ship MQL alerts immediately. Half of the first 100 MQLs are clearly not qualified. Sales loses trust on day 1 and you're rebuilding model perception for 3 months.
How to avoid: Always backfit on 30 days of historical leads. Manually rate "would I call this today" for 30 leads, score them in your model, verify high-score leads correlate with yes ratings. Activate only after backfit looks reasonable.
Never recalibrating
What goes wrong: Model built in 2024 still running unchanged in 2026. ICP shifted, buyer behavior changed, but the score doesn't know. Closed-won rate from MQL drops from 18% to 9%. Marketing blames sales; sales blames marketing.
How to avoid: Quarterly recalibration with sales. 90 minutes per quarter. Document changes in a model-version log.
Recap
Done — what's next
How to set up HubSpot lifecycle stages so marketing and sales actually agree
Read the next tutorial
Hand it off
Lead scoring is one of the highest-leverage HubSpot configurations when done well — and one of the most ignored when done DIY. A specialist will run the sales alignment, design the model, backfit on historical data, and recalibrate quarterly. EverestX HubSpot specialists typically run $400-1,200/mo at $14-16/hr including ongoing scoring management.
See specialist rates
Start with HubSpot Score (the default). Build custom score properties only when you have distinct segments needing distinct models (e.g., separate Enterprise vs SMB scoring), or when you want to keep an experimental score alongside the production one.
Predictive scoring (Marketing Hub Pro+) trains a model on your historical conversion data. It needs at least 100 customers and 6+ months of HubSpot data. Below that, the model has nothing to learn from. Above it, predictive can outperform manual rules.
HubSpot scoring runs on marketing/sales activity. Product-usage scoring (for PLG companies) lives in your product database and surfaces signals like 'invited 5 teammates,' 'connected Stripe,' 'sent 100 emails.' For PLG, integrate product usage as a custom property and add it to the HubSpot scoring model.
Yes if the qualification criteria genuinely differ. A SaaS product MQL and a services MQL likely have different signals. Build separate custom score properties, one per product. Don't try to make one score cover everything — it always picks one direction.
Anonymous visitors don't have a HubSpot contact yet, so no score accrues. Once they submit any form or get identified via _hsq.push(['identify', ...]), HubSpot retroactively attaches the prior session activity to the new contact and applies scoring. This is one of many reasons the identify() call matters (see tracking-code tutorial).
HubSpot Marketing Hub
Lifecycle Stage is the most misunderstood property in HubSpot. Set it up wrong and marketing reports show 'MQLs created' that sales never touches. Set it up right and the whole pipeline becomes legible. Here's the taxonomy and automation that holds up.
HubSpot Marketing Hub
Workflows are the engine under HubSpot's marketing automation. They are also where 80% of the silent breakage happens — wrong enrollment criteria, missing re-enrollment toggles, branch logic that loops. Here's how specialists build them so they hold up.
HubSpot Marketing Hub
Properties are the foundation under every workflow, segment, and report in HubSpot. Most owners create 80+ custom properties in the first month, then spend the next year cleaning up the mess. Here's the discipline to do it right the first time.
HubSpot Marketing Hub
HubSpot is one of the most powerful platforms in marketing — and one of the easiest to over-configure into uselessness. Here's the honest framework: when DIY costs you more than hiring, and how to tell which side of the line you're on.