Short answer: a lead scoring model for a small SaaS company needs four or five signals, not forty. Give 50 points for fit (is this the kind of company and person you sell to?) and 50 for intent (are they trying to buy now?), send 80+ to sales the same hour, and check the weights against your last 50 closed deals before you trust them.
Most lead scoring guides are written for enterprise teams with marketing ops staff and thousands of leads a month. This one is for SaaS companies with a small sales team, a few hundred leads a month at most, and no time to babysit a model. It uses the same 100-point structure we built and stress-tested in our AI lead scoring case study.
A lead scoring model is a set of rules that gives each new lead points for who they are and what they have done, then adds them into one number. Sales uses the number to decide who to contact first. A good model is short, written down, tested against past deals and owned by one person.
For the wider plan, see our SaaS marketing agency page; for the short definition, see lead scoring in our glossary.
Only if your team cannot reach every good lead quickly. If sales can call or email every inbound lead within an hour, a score adds little: fix routing and response speed first. Scoring starts to pay when leads arrive faster than people can read them, or when a few big opportunities are hidden among many small sign-ups.
Our rule of thumb: under about 50 inbound leads a month, skip the score and set a response time target instead. Our post on lead response time and the 5-minute rule explains why speed usually matters more than ranking at that volume, and our speed-to-lead calculator shows what slow replies cost. Between 50 and a few hundred leads a month, a simple model like the one below is worth it. Above that, add product usage data and consider your CRM's built-in or AI scoring.
Split the score into two halves so one strong signal cannot carry a weak lead. Fit tells you whether the company could ever buy. Intent tells you whether they are buying now. Here is a starting template; change the labels to match your own customers.
| Signal | Half | How to score it | Max |
|---|---|---|---|
| Company size band | Fit | Your best-selling size band 20, next band 10, outside your range 0 | 20 |
| Role of the person | Fit | Budget owner or founder 15, team lead who uses the product 10, student or job seeker 0 | 15 |
| Use case match | Fit | A problem you solve well 15, a partial match 5, not a fit 0 | 15 |
| Action taken | Intent | Demo or pricing request 25, trial sign-up 15, content download 5 | 25 |
| What they wrote or did in the product | Intent | Clear problem and timeline in the message, or the key "aha" action done in the trial: up to 25 | 25 |
| Negative signals | Either | Competitor domain, test or spam entry, region you cannot serve: subtract 30 or more | – |
Notice what is missing: email opens, page views and job titles parsed from LinkedIn. They are noisy, easy to game and rarely change the order of your top ten leads. Leave them out until the basic model is working.
One choice that matters for smaller SaaS: do not punish free email addresses hard. Many founders and owners of 5–50 person firms sign up with Gmail or Outlook. A small deduction is fine; a large one hides real buyers. Check how many of your won deals came from free email domains before you decide.
Start with three bands and a clear action for each. The action matters more than the exact number, because a score nobody acts on is just a column in the CRM.
| Band | Score | What happens |
|---|---|---|
| HOT | 80–100 | Assigned to a named rep at once; personal reply within the hour in working time |
| WARM | 50–79 | Rep reply the same day; short nurture if no response |
| COLD | 0–49 | Automated nurture; no sales time until the score rises |
These are the bands we use in our own Zapier workflow. Treat them as a first guess. If more than about a third of all leads land in HOT, the bar is too low and sales will stop trusting it. If almost none do, raise the weights for intent.
Score your last 50 closed deals by hand, wins and losses, in a spreadsheet. If most wins score above 70 and most losses score below 50, the model works. If wins and losses are mixed, find the signal that separates them and give it more points. This takes an afternoon and avoids months of wrong routing.
Fifty is a practical number, not a magic one. HubSpot's own AI scoring asks for at least 25 converted and 25 non-converted contacts before it will train, which is a useful reminder: with fewer examples, any model is mostly guesswork. If you have fewer than 50 closed deals, use the template as it is and review it after your next 50 leads.
Send us your current scoring rules and a list of recent won and lost deals (company size, source, first action). We will tell you which signals actually separate them. Tell us about your business →
Pick the simplest place your team already works. The model is the same everywhere; only the plumbing changes.
Use AI to read what a person wrote, and rules for everything with a fixed answer. A language model is good at judging whether "we get 300 leads a month and qualify them by hand" describes a real, urgent problem. It is a poor judge of company size or budget bands, which a lookup table scores the same way every time.
In our build, AI could add at most 35 of the 100 points, so it could never push a lead into HOT on its own. When we ran the same lead twice, it scored 68 and then 69, so small swings are normal. That is another reason to keep the AI share capped and the thresholds fixed in rules.
The biggest bug we found had nothing to do with AI. Our lookup table was keyed on the labels visitors saw, but the form sent different underlying values, so some answers silently scored zero. Check that every scoring rule matches the exact value your form or CRM sends.
Review it 90 days after launch, then every quarter. Look at three things: how many leads landed in each band, how many HOT leads became opportunities, and which lost deals scored high. Change one or two weights at a time and write down why, so the next review has something to compare against.
Add decay for intent points from the start. A pricing-page visit from last week means something; one from five months ago does not. Three months is a sensible default for most SMB sales cycles.
Four to six is enough for most small SaaS teams: two or three fit signals, one or two intent signals, and negative points for leads you never want. More criteria rarely change the order of your best leads, but they make the score harder to explain and harder to fix when it goes wrong.
There is no universal number. Start where your back-test puts most won deals, often around 50 on a 100-point scale for a marketing-qualified lead and 80 for a sales-ready one. Then adjust: if sales rejects many leads above the line, raise it; if good deals start below it, lower it or reweight.
Yes. A demo request is a clear buying signal on its own. A trial sign-up only becomes one when the user does the action that predicts conversion in your product, such as connecting data or inviting a teammate. Score that action, not the sign-up, and route trial users to sales only after it happens.
You can, but we would not for a small team. AI scores can drift between runs and are hard to explain to sales. Use rules for fixed facts like size, role and action, and let AI score only the free text, with a cap on how many points it can add.
No. HubSpot's rules-based lead scoring works on Marketing Hub or Sales Hub Professional. Enterprise adds AI contact scores. You can also build the same model in a spreadsheet, another CRM or an automation tool like Zapier, Make or n8n.
Scoring decides how good a lead is; routing decides who gets it and how fast. You usually need both, in that order: the score sets the band, and the routing rules send each band to the right person or sequence. Many small teams need routing first and scoring later.
Want help building it? See our AI marketing automation service, read how we built and broke a Zapier scoring workflow, or tell us how leads reach you today.