Lead scoring is one of the first things businesses ask for when a CRM goes in, and one of the first things quietly abandoned afterwards. The pattern is familiar: someone assigns points to a list of activities in a workshop, the scores start appearing on records, and within a few weeks the sales team is sorting by date again because the numbers do not match their experience of which leads are real.
The scoring tools in Zoho CRM and HubSpot are capable enough. What fails is the model put into them. This is how we build one that a sales team keeps using, regardless of platform.
Start with what the score is for
A score that does not trigger anything is a number on a record. Before deciding a single point value, agree what happens at each band, with the people who will act on it.
| Band | What it means | What happens |
|---|---|---|
| Hot | Good fit and showing clear buying intent now | Assigned to a named rep with a short response deadline and an alert on WhatsApp or mobile |
| Warm | Good fit, some intent, not ready yet | Enters a nurture sequence; rep notified if the score crosses into hot |
| Cold | Weak fit or no recent activity | Stays in marketing; no sales time spent |
| Disqualified | Outside what you sell or who you sell to | Closed with a reason, so it stops appearing in reports |
Once the bands are agreed, the point values become a means to an end: putting leads into the band the sales team would have put them in by hand, faster and more consistently.
Fit and intent are two different scores
Most models fail because they add everything into one number. A procurement manager at exactly the right size of company who has never opened an email gets the same score as a student who downloaded every guide on the site. Neither is a hot lead, for opposite reasons.
| Fit: is this who we sell to? | Intent: are they buying now? |
|---|---|
| Company size or type | Asked for a quote, a call or a demo |
| Role or seniority of the contact | Replied to a WhatsApp qualifying question |
| Location or emirate you serve | Visited the pricing or service page more than once |
| Budget band they selected | Opened or clicked recent follow-up messages |
| Timeline they gave | Returned after a period of silence |
Keep them as two fields. Both Zoho and HubSpot can hold more than one score on a record, and the combination reads far more clearly than a sum: high fit and high intent is a call today, high fit and low intent is nurture, low fit and high intent is a polite qualification question before anyone spends time on it.
Build the first version from your own deals
Generic scoring templates assign points to things that predict a sale for somebody else's business. The faster route to a model that works is to look backwards at your own pipeline.
- 1Export the deals closed in a recent period you consider normal, both won and lost, with the lead's source, the answers they gave and the activity before the deal was created.
- 2Compare the two groups. Look for attributes and actions that are common among won deals and rare among lost ones. Those are your signals.
- 3Give the strongest signals the most points, and leave out anything that appears equally in both groups, however much it feels like it should matter.
- 4Draft the thresholds so that, applied to those historical leads, most of the eventual wins would have landed in hot or warm.
- 5Show the sales team a sample of leads with their draft scores and ask where they disagree. Their objections are usually a missing signal, not a wrong one.
Negative scores and decay keep it honest
A model that only adds points drifts upwards until everything is hot. Two mechanisms stop that.
- Negative scores for signals that predict a loss or a mismatch: a competitor's domain, a job seeker using the contact form, a location you do not serve, an unsubscribe.
- Decay for intent. A pricing page visit last week means something; the same visit eight months ago does not. Reduce intent points over time, or score only activity within a recent window.
- Fit does not decay the same way. A company that was the right size last quarter usually still is, so keep fit stable unless the underlying data changes.
Make the data reliable first
Scoring runs on fields, and a model is only as good as the data reaching them. If budget is a free-text field, it cannot be scored. If WhatsApp replies never reach the CRM, the strongest intent signal many UAE businesses have is invisible to the model. If the same person exists as three contacts, their activity is split three ways and none of the records crosses the threshold.
- Turn qualifying answers into picklists on the form and in the WhatsApp flow, so they arrive as values the CRM can score.
- Connect every channel that carries intent into the CRM, including WhatsApp conversations and ad lead forms, not just the website.
- Deduplicate before scoring, so activity accumulates on one record per person.
Check it against outcomes, then adjust
A scoring model is a prediction, and predictions get checked. On a regular schedule, monthly at first and quarterly once it settles, compare how leads in each band actually turned out.
- Win rate by band. Hot leads should close noticeably more often than warm, and warm more often than cold. If not, the model is not separating anything.
- Wins that came from cold. Each one is a signal the model missed. Look at what those leads had in common.
- Hot leads that went nowhere. If many are the same type, one of your signals is weighted too heavily.
- Sales feedback. Reps who ignore the score usually have a reason, and it is worth hearing before changing any points.
Change a few things at a time and note what changed and when, so the next review can tell whether it helped.
Common questions
A way of ranking leads automatically by how likely they are to buy, using points for attributes such as company size or budget and for actions such as requesting a quote. The score is useful only when it changes what happens next, such as assigning a hot lead to a rep immediately or placing a warm one into nurture.
Fit measures whether the lead matches who you sell to: company type, role, location, budget. Intent measures whether they are showing buying behaviour now: asking for a quote, replying to qualifying questions, returning to pricing pages. Keeping them as two scores makes it clear whether a lead needs a call, nurture or disqualification.
There is no universal answer. Build the first version from your own won and lost deals, giving the most points to signals that are common among wins and rare among losses, then adjust after comparing win rates by score band. Templates assign points based on someone else's business.
Both include native scoring, with the features available depending on your edition or tier. Check what your plan includes before designing the model, since the limit on the number of scores or rules can shape how fit and intent are split.
Usually because the scores do not match the sales team's experience, the model only ever adds points so everything ends up hot, or nothing happens when a lead crosses a threshold. Recalibrate against closed deals, add negative scores and decay, and tie each band to a specific action.