> Enrich.chat gives AI agents verified work emails and mobiles. To use it from an agent, read https://enrich.chat/agents.md . This is the Markdown version of https://enrich.chat/blog/lead-scoring-model/

# Lead scoring model: fit versus engagement, a worked scoring table, and the fit data to fill it

A scoring model ranks leads so sales works the likeliest buyers first. Here is how to split fit from engagement, set weights and decay, choose a threshold, test it, and fill the fit half from company and contact data.

By Enrich.chat Editorial. Updated October 16, 2026.

**Short answer:** A lead scoring model gives each lead points so sales works the likeliest buyers first. Keep two scores apart, fit (who the lead is) and engagement (what the lead does), as [HubSpot](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) and [Oracle Eloqua](https://docs.oracle.com/en/cloud/saas/marketing/eloqua-user/Help/LeadScoring/LeadScoring.htm) do. Let engagement points fade, and set the MQL threshold from your own won and lost deals.

## Key takeaways

- Score fit and engagement as two numbers, then combine them with a rule. HubSpot, Salesforce Account Engagement and Oracle Eloqua keep the two apart, and Eloqua maps each pair, from A1 to D4, to an action.
- A single summed score can send a quiet, well-matched lead to sales. In this page's example, a lead with 100 on fit and 16 on engagement totals 116 and passes a combined line of 100.
- Let engagement fade, not fit. HubSpot decays each event over 1, 3, 6 or 12 months, and Inbound 281 says a 30-day penalty can be too harsh for long sales cycles.
- No published number sets your MQL threshold. Base it on conversion history and the leads sales can follow up, and backtest it on won, lost and stalled leads before you route anyone.
- On Enrich.chat, fit data for one lead takes three lookups, 15 credits when all are found. It does not track engagement or compute scores.

## What is a lead scoring model?

A lead scoring model is a set of rules that gives each lead points, so sales works the likeliest buyers first. [HubSpot's documentation](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) describes scores built from record properties and actions, with points that can go down as well as up. Its examples subtract points when a contact unsubscribes or when a company sits in a region you do not serve.

Most models have two halves. Fit scoring rates who the lead is: job title, company size and industry. Engagement scoring rates what the lead does: visiting your site, clicking an email or booking a meeting. Four large tools keep the halves apart.

| Tool | Fit side | Engagement side | How they combine |
| --- | --- | --- | --- |
| HubSpot | Fit score | Engagement score | Combined score, labels A1 to C3 |
| Salesforce Account Engagement | Grade, F to A+ | Score | Two separate values |
| Oracle Eloqua | Profile score, A to D | Engagement score, 1 to 4 | Lead score, A1 to D4 |
| Adobe Marketo Engage | Demographic scoring | Behavioral scoring | Users suggest an A1 to D4 grid |

The split comes from the documentation of [HubSpot](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool), [Salesforce](https://trailhead.salesforce.com/content/learn/modules/lead-scoring-and-grading-in-account-engagement/get-started-with-lead-grading), [Oracle](https://docs.oracle.com/en/cloud/saas/marketing/eloqua-user/Help/LeadScoring/LeadScoring.htm) and [Adobe's Marketo session](https://experienceleague.adobe.com/en/on-demand-events/scoring). The Marketo grid is a suggestion from community users, not a product feature. A [lead qualification](https://enrich.chat/blog/lead-qualification/) framework covers the questions asked on the call. A scoring model decides who gets that call.

### When does predictive scoring make sense?

Predictive scoring trains a model on past outcomes, so you do not write the points. It needs history. [HubSpot](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) says an AI-built score needs a sample of at least 50 contacts, 25 converted and 25 not. With less than that, a points model is the practical start. You can swap in a trained model later and keep the same routing.

## Should you score fit and engagement together or separately?

Score them separately, then combine them with a rule. One summed number can hide the problem the model exists to catch.

Practitioners in [Adobe's Marketo community](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/lead-score-vs-behavior-demographic-score-why-both-196894) describe the failure: a tire kicker with a low demographic score and a high behavior score can total more than a better-matched lead. [Leadfeeder](https://www.leadfeeder.com/blog/sales-prospecting/sales-intent-score/) makes the mirror point. A prospect who ticks every fit box but shows little interest looks hot on paper, and the lack of engagement is a red flag. [Inbound 281](https://web.inbound281.com/blog/lead-scoring-model-guide), a HubSpot partner agency, suggests a minimum fit score plus enough intent points, so a poor match cannot reach sales just by consuming content.

Oracle Eloqua turns the pair into a grid with an action for some cells. Its [documentation](https://docs.oracle.com/en/cloud/saas/marketing/eloqua-user/Help/LeadScoring/LeadScoring.htm) grades profile A to D and engagement 1 to 4, then gives four examples. A4, the right prospect with no interest, gets "why now" messaging. B1, a good fit and very interested, goes to a sales queue for immediate follow-up. C1, not the ideal prospect but very interested, keeps being nurtured and profiled. D4, a wrong fit with no interest, is segmented out. Figure 1 shows the grid, with three example leads that the next section scores.

**Figure 1: Fit and engagement as a grid with an action in four cells.** A four by four grid. Rows are the fit grade, A best to D worst. Columns are the engagement grade, 1 most active to 4 least active. The six cells with fit A or B and engagement 1, 2 or 3 are shaded as an example sales-ready zone. Four outlined cells carry Oracle Eloqua actions: A4, a right fit that is quiet, gets why-now messages; B1, a good fit that is active, goes straight to sales; C1, a weak fit that is active, is nurtured and profiled; and D4, a wrong fit that is quiet, is segmented out. Three example leads sit in cells A4, B3 and D3.

*Fit (rows) and engagement (columns) as a grid. The grades and the four outlined actions come from [Oracle Eloqua's documentation](https://docs.oracle.com/en/cloud/saas/marketing/eloqua-user/Help/LeadScoring/LeadScoring.htm), with the action wording shortened. The shaded zone and the three leads are this page's example.*

There is a base-rate reason to gate. A [Marketing Science report on Ehrenberg-Bass research](https://marketingscience.info/news-and-insights/ehrenberg-bass-95-of-b2b-buyers-are-not-in-the-market-for-your-products) says companies change providers of services such as banking, legal advice, software or telecoms about every five years. So only 20% are in the market in a given year and 5% in a given quarter. Most perfect-fit accounts are not buying this quarter. Fit says who belongs on the list. Engagement, or a signal from our [sales signals guide](https://enrich.chat/blog/sales-signals/), says who to call now.

## What does a worked lead scoring model look like, with weights?

Here is one example model with two scores, each from 0 to 100. The weights are invented to show the mechanics, not benchmarks, so set yours from your own won and lost deals. The scenario is a company that sells route-planning software to US freight carriers.

### Fit points

Fit uses the six rules from the worked example in our [ideal customer profile guide](https://enrich.chat/blog/ideal-customer-profile/): logistics industry 30, 200 to 1,000 employees 25, United States 15, runs HubSpot 10, contact is a VP or director of operations 20, and under 50 employees minus 50. Add the points and floor the total at 0.

### Engagement points

Engagement comes from your own tracking. Strong actions earn more and fade slower. Weak actions are capped, so ten email clicks cannot outweigh one demo request. The model ignores email opens, because [Inbound 281](https://web.inbound281.com/blog/lead-scoring-model-guide) says privacy settings and automated security tools make open data unreliable.

| Action | Points | Counts up to | Fades to zero after |
| --- | --- | --- | --- |
| Demo request or meeting booked | 50 | Once | 180 days |
| Webinar attended | 20 | Once | 90 days |
| Pricing page visit | 10 each | 20 | 30 days |
| Email click | 5 each | 10 | 30 days |
| Unsubscribe | Minus 30 | Once | Never |

A demo request is also a hand raise. Send it to a person the same day, whatever the scores say. [Concept's primer](https://conceptltd.com/blog/lead-scoring-model) makes the same point: when a lead raises a hand, sales should engage.

### Score three example leads

The example rule: a lead is sales-ready when fit is above 50 and engagement is above 25. Those are the A or B and 1, 2 or 3 cut-offs in Figure 1, which are [Oracle Eloqua's default thresholds](https://docs.oracle.com/en/cloud/saas/marketing/eloqua-user/Help/LeadScoring/Tasks/ConfiguringLeadScoringThresholds.htm). The leads and companies are made up.

- **Lead 1, Priya Nair, VP of Operations at Northgate Logistics** (420 employees, logistics, United States, runs HubSpot). Fit: 30 + 25 + 15 + 10 + 20 = 100. Engagement: one webinar 18 days ago, 20 points faded by 20% (18 of 90 days), so 16.
- **Lead 2, Tomas Reyes, Director of Operations at Ridgeline Freight** (180 employees, logistics, United States, no HubSpot). Fit: 30 + 15 + 20 = 65. Engagement: pricing visits today (10) and 6 days ago (8), email clicks today (5) and 6 days ago (4), and the same webinar (16), so 43.
- **Lead 3, Dalton Moss, Operations Manager at Kestrel Haulage** (35 employees, logistics, United States, tech stack not found). Fit: 30 + 15 - 50 = -5, floored at 0. HubSpot is unknown, and even if it applied, fit would be 5. Engagement: a webinar today (20), pricing visits (18) and email clicks (9), so 47.

| Lead | Fit | Engagement | Cell | What happens |
| --- | --- | --- | --- | --- |
| 1. Priya Nair, Northgate Logistics | 100 | 16 | A4 | Right fit, quiet: send why-now messages |
| 2. Tomas Reyes, Ridgeline Freight | 65 | 43 | B3 | Sales-ready: to the sales queue |
| 3. Dalton Moss, Kestrel Haulage | 0 | 47 | D3 | Active, poor fit: nurture, do not send to sales |

Only Lead 2 passes. Now add the scores instead. Lead 1 totals 116 and Lead 2 totals 108, so one line at 100 would send both to sales, although Lead 1 has done almost nothing. That is the case for gating on two scores.

## How should lead score decay work?

Decay lowers the points of older actions so that a score shows recent behavior. Without it, a lead that clicked an email six months ago ranks beside one that asked for a demo yesterday. [Count's guide](https://count.co/metric/lead-scoring-analysis) lists ignoring decay as a common mistake, and [Kissmetrics](https://kissmetrics.io/glossary/lead-scoring) says last week's engagement should count more than six-month-old engagement.

Tools differ. In HubSpot, [score decay](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) applies to engagement and combined scores, event by event. You pick an interval of 1, 3, 6 or 12 months, each counted as 30 days, and the points fall in a straight line, past events included. [Salesforce's Trailhead module](https://trailhead.salesforce.com/content/learn/modules/lead-scoring-and-grading-in-account-engagement/get-started-with-lead-scoring) describes decay in Account Engagement as reducing a score for inactivity, done with automation rules that adjust a score or reset it to zero. Figure 2 draws the fade rules from the example model. Four actions on the same day add up to 85 points. Forty-five days later, 47.5 are left.

**Figure 2: How four example actions fade over 180 days.** A line chart with days since the action from 0 to 180 across and points left from 0 to 50 up. A demo request starts at 50 points and falls in a straight line to zero at 180 days. A webinar starts at 20 and reaches zero at 90 days. A pricing page visit starts at 10 and an email click at 5, and both reach zero at 30 days. At day 45 the demo request has 37.5 points left, the webinar 10, and the other two zero, so 47.5 of the original 85 points remain.

*Illustrative maths for this page's example model: each action fades in a straight line to zero. The weights and fade times are invented, not benchmarks. HubSpot's [linear decay](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) works the same way, with its own interval choices.*

Four rules of thumb follow, based on the sources.

1. **Fade engagement, not fit.** A practitioner in [Adobe's Marketo community](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/lead-score-limit-for-individual-and-account-178836) decays activity only, and rescores the demographic score when a job title or another data point changes. Fit goes stale through real-world change, so look it up again. [Lusha](https://www.lusha.com/blog/b2b-data-decay-rate-measured/) measured about 1% of US sales leaders changing roles each month, and our [B2B data decay guide](https://enrich.chat/blog/b2b-data-decay/) covers the rates.
2. **Match the fade to your sales cycle.** Inbound 281 says a 30-day inactivity penalty can suit a short cycle but be too aggressive for capital equipment or enterprise services. An answer in a [Marketo community thread on expiration](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/lead-scoring-expiration-164600) names sales cycle length as a good cut-off.
3. **Fade weak actions fast, strong ones slowly, and cap repeats.** An email click says little after a month. A demo request still matters after a quarter. A user in the Marketo community reports that content viewed again and again keeps feeding a score, so decay never catches up.
4. **Decide whether to keep a lifetime count.** Opinions differ. One poster in a [community thread](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/never-decay-your-behavior-score-164638) keeps a never-decaying behavior score beside a decaying recent-activity score.

## Where should you set the MQL threshold?

Set it from your conversion history and your sales capacity, not from a published number. Inbound 281 says there is no universal sales-qualified score: 50 may suit one company and mean nothing at another. [Concept](https://conceptltd.com/blog/lead-scoring-model) says to keep the threshold attainable and to raise or lower it after launch. Published bands are only examples: [HubSpot's](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) sample labels are High at 70 to 100, Medium at 40 to 69 and Low below that.

1. **Count your capacity.** How many leads can reps follow up well each week? Inbound 281's example service level is contact within one business day.
2. **Rank last year's leads by score.** Find the cut where the leads above it match your capacity and conversion visibly steps up.
3. **Require a fit floor and an engagement floor,** as in Figure 1, and add the hand-raise override.
4. **Review quarterly.** Inbound 281 says to watch for point inflation, where too many contacts cross the line.

The routing table covers all 16 cells of the grid. It is an example, built from Eloqua's documented follow-ups.

| Situation | Example rule | Action |
| --- | --- | --- |
| Hand raise | Demo request, any scores | A person replies the same day |
| Sales-ready | Fit A or B, engagement 1 to 3 | Sales queue, contact within one business day |
| Right fit, quiet | Fit A or B, engagement 4 | Why-now messages; watch for engagement |
| Active, weak fit | Fit C or D, engagement 1 to 3 | Nurture, and look up the fit data again |
| Weak fit, quiet | Fit C or D, engagement 4 | No outreach; keep in the database |

## How do you test a lead scoring model?

Backtest it on leads whose outcome you know, then watch it live. Count's guide lists insufficient validation as a common mistake: scoring never tested against actual conversion rates leads to misaligned sales effort.

1. **Pull a mixed sample.** Inbound 281 suggests closed-won deals, closed-lost opportunities and leads that never progressed.
2. **Score only what was known then.** Use actions from before the outcome date. A lookup made today shows the company as it is today, so treat fit for old leads as approximate.
3. **Fill in a band table**, as below. The rate should fall from the top band to the bottom. Kissmetrics says that if mid-score leads convert at the same rate as high-score leads, recalibrate.
4. **Respect small numbers.** In a band of 20 leads, one lead moves the rate by 5 points.
5. **Beat a one-rule baseline,** such as title level alone. If the model does not, simplify. Kissmetrics suggests starting with 5 to 7 factors.
6. **Run a blind round, then track it live.** A commenter in an older [Marketo community thread](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/never-decay-your-behavior-score-164638) describes hiding the score from sales, then holding a quarterly review that compares scores with the leads that became customers. Count suggests comparing lead-to-opportunity rates across score ranges each month.

| Score band (fit and engagement) | Leads | Reached SQL | Won | Reached SQL rate |
| --- | --- | --- | --- | --- |
| Fit A or B, engagement 1 to 3 | fill in | fill in | fill in | fill in |
| Fit A or B, engagement 4 | fill in | fill in | fill in | fill in |
| Fit C or D, engagement 1 to 3 | fill in | fill in | fill in | fill in |
| Fit C or D, engagement 4 | fill in | fill in | fill in | fill in |

## What mistakes break a lead scoring model?

Most failures come from how the points are used.

- **Summing fit and engagement into one number.** A busy poor-fit lead can outrank a well-matched one.
- **Scoring email opens.** Inbound 281 calls a single open a weak signal that privacy settings and security tools distort. Clicks are a better input.
- **Reading a title as authority.** Inbound 281 says titles are inconsistent across companies, so combine title with company facts. See [decision makers at a company](https://enrich.chat/blog/decision-makers-in-a-company/).
- **Using too many factors.** Kissmetrics warns against dozens of weighted factors nobody can maintain or explain.
- **Using negative points as a wall.** Inbound 281 says to lower the priority of a smaller company and nurture it rather than disqualify it.
- **Treating missing data as a no.** A company whose tech stack was not found is unknown, not a company without HubSpot. Score it 0 and show the best-case score next to it.
- **Treating the score as a verdict or as permission.** Inbound 281 says scoring should inform judgment, not replace it. A high score also does not change the contact rules: CAN-SPAM, GDPR and the TCPA still apply. See [cold email](https://enrich.chat/blog/cold-email/) and [TCPA and B2B cold calling](https://enrich.chat/blog/tcpa-b2b-cold-calling/), and anyone can opt out at [enrich.chat/opt-out/](https://enrich.chat/opt-out/). This is general information, not legal advice.

## How does Enrich.chat fill the fit half of a lead scoring model?

Enrich.chat supplies fit data for a lead's company and contact: industry, headcount, country, tech stack and job title. It does not track engagement, and it does not compute or store scores. Your CRM or a spreadsheet holds the rules.

Enrich.chat is contact and company data for AI agents. The agent reads [enrich.chat/agents.md](https://enrich.chat/agents.md) once, or you add the MCP server at `https://enrich.chat/mcp`. The [install page](https://enrich.chat/install/) and the [agents page](https://enrich.chat/agents/) have steps for ChatGPT, Claude, Cursor and more. For a one-off stack check, try the free [tech stack lookup](https://enrich.chat/tools/tech-stack-lookup/). The [company enrichment API](https://enrich.chat/api/company-enrichment/) and [person enrichment API](https://enrich.chat/api/person-enrichment/) pages list every field.

### Which fit rules can a lookup fill?

| Fit rule | Field to fill | Call |
| --- | --- | --- |
| Headcount in your band | employees | GET /v1/company?domain= |
| Industry matches | industry | The same company call |
| Country or state | country, state | The same company call |
| A tool in the stack | technologies, each with a category | GET /v1/company/tech?domain= |
| Title at your target level | title | GET /v1/person with fields=profile |
| Seniority level | Not a returned field | Map the title to a level with your own rule |
| Revenue, funding, growth, intent | Not available | Bring them from another source |

Each found profile, company or tech stack uses 5 credits, so fit data for one lead costs 15 credits when all three lookups find a match. A lookup that finds nothing uses 0. Figure 3 shows where each column of a scoring row comes from.

**Figure 3: Where each part of a lead score comes from.** Two panels feed two scores. On the left, three Enrich.chat lookups fill the fit score: a company call for employees, industry, state and country, a tech call for the tools in the stack, and a person call with fields set to profile for the title. Together they use 15 credits per lead when all are found. On the right, engagement comes from your own systems, such as pricing page visits, email clicks, demo requests and webinar attendance, and is not looked up by Enrich.chat. Both scores, each from 0 to 100, feed a gate that sends fit A or B with engagement 1 to 3 to the sales queue.

*The fit columns come from three lookups. The engagement columns come from your own tracking. Credits are used only when a lookup finds a match. Endpoints and credits are from [Enrich.chat's agents.md](https://enrich.chat/agents.md).*

### What do the lookups return for Lead 1?

A lead from a form gives you a work email. The part after the @ is the company website, and the same email finds the person. The responses below are illustrative, for the made-up Northgate Logistics. Run them one at a time, about one request a second.

```
GET https://enrich.chat/v1/company?domain=example.com

{
  "found": true,
  "name": "Northgate Logistics",
  "domain": "example.com",
  "industry": "logistics & supply chain",
  "employees": 420,
  "state": "Texas",
  "country": "United States",
  "credits_used": 5,
  "free_lookups_remaining": 9
}
```

```
GET https://enrich.chat/v1/company/tech?domain=example.com

{
  "found": true,
  "domain": "example.com",
  "company": "Northgate Logistics",
  "technologies": [
    {"name": "HubSpot", "category": "CRM"},
    {"name": "Stripe", "category": "Payments"}
  ],
  "credits_used": 5,
  "free_lookups_remaining": 8
}
```

```
GET https://enrich.chat/v1/person
  ?email=priya.nair@example.com
  &fields=profile

{
  "found": true,
  "name": "Priya Nair",
  "title": "VP Operations",
  "company": {"name": "Northgate Logistics"},
  "credits_used": 5,
  "free_lookups_remaining": 7
}
```

For Lead 3, the tech lookup finds nothing. The response is `{"found": false}` with `credits_used` at 0, and no free lookup is used. Your agent writes unknown in the HubSpot column, not no. Here is the scoring-ready table the agent builds:

| Lead website | Employees | Industry | Country | HubSpot in stack | Title |
| --- | --- | --- | --- | --- | --- |
| example.com | 420 | logistics and supply chain | United States | yes | VP Operations |
| example.net | 180 | logistics and supply chain | United States | no | Director of Operations |
| example.org | 35 | logistics and supply chain | United States | unknown | Operations Manager |

Your rules turn the title into a level. The person lookup returns a title, not a seniority field. To find VP-level people at a company, `/v1/company/people` and `/v1/search` accept a seniority filter. For the leads that pass the gate, add `fields=email`. Each found work email comes back with a status label (verified, catch_all or unknown), which is a label and not a delivery guarantee. See [what a catch-all email is](https://enrich.chat/blog/what-is-a-catch-all-email/).

```text
Use www.enrich.chat to build the fit side of a lead score. For each lead in my list, take the work email, use the part after the @ as the company website, and look up the company for its industry, employee count and country, look up its tech stack, and look up the person for their title. Skip personal mailboxes such as Gmail and mark them not looked up. Then score fit out of 100: logistics industry 30, 200 to 1,000 employees 25, United States 15, runs HubSpot 10, title is VP or director of operations 20, under 50 employees minus 50, with a floor of 0. Mark any field not found as unknown, not no, show the best-case score, and report the total credits used. Do not score engagement.
```

### What does it cost?

You get 10 free API lookups a day with no signup. That covers three leads a day at three lookups each. Plans start at $29 a month. Say you score 1,000 new leads a month, look up 300 open leads again for their titles, and fetch a work email for the 100 that pass the gate. That uses 17,500 credits, which fits the Growth plan at $29 a month. The 3,000 fit lookups for the new leads take about 50 minutes at one request a second. A lookup that finds nothing uses 0 credits. Plans bill monthly with unlimited seats and no annual contract. See [Enrich.chat pricing](https://enrich.chat/pricing/).

### What are the limits?

- **No engagement and no scoring.** Enrich.chat has no page, email or product events and no scoring engine. The points live in your CRM or a spreadsheet.
- **No revenue, funding, growth or intent data.** A rule that needs them has to use another source. See [intent data](https://enrich.chat/blog/intent-data/).
- **Websites, not company names, and work emails only.** A person lookup takes a LinkedIn URL, a work email or a name plus a company website. A lead who gave a Gmail address cannot be looked up by email.
- **A title, not a seniority level.** You map titles to levels yourself.
- **No saved lists or bulk export on self-serve plans.** Enterprise adds bulk files and data licenses.

### When is another tool the better choice?

If you want scores computed and updated inside your CRM as leads act, use the scoring built into [HubSpot](https://enrich.chat/integrations/hubspot/), Salesforce, Marketo or Eloqua, and feed it the fit fields above. If your model needs funding, hiring or buying signals, read [what sales intelligence is](https://enrich.chat/blog/what-is-sales-intelligence/) and compare the [best sales intelligence tools](https://enrich.chat/blog/best-sales-intelligence-tools/). To keep a CRM's fit fields current, see [CRM data enrichment](https://enrich.chat/blog/crm-data-enrichment/).

## Frequently asked questions

### What is a lead scoring model?

A lead scoring model is a set of rules that gives each lead points, so sales can work the likeliest buyers first. [HubSpot](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) builds scores from record properties and actions. Most models score fit (title, company size, industry) and engagement (visits, clicks, meetings) separately, then route leads by both. Our [lead qualification guide](https://enrich.chat/blog/lead-qualification/) covers the questions asked on the call.

### What is the difference between fit and engagement scoring?

Fit scoring rates who a lead is: title, company size, industry and tools. Engagement scoring rates what the lead does: page visits, email clicks and meetings. Salesforce Account Engagement calls them a grade and a score. [Oracle Eloqua](https://docs.oracle.com/en/cloud/saas/marketing/eloqua-user/Help/LeadScoring/LeadScoring.htm) uses a profile score from A to D and an engagement score from 1 to 4. Keeping them apart shows a quiet ideal lead and an active poor-fit lead as different.

### What lead score makes a lead sales-ready?

No published number fits every company. [Inbound 281](https://web.inbound281.com/blog/lead-scoring-model-guide) says a threshold of 50 may suit one company and mean nothing at another, so set it from historical conversion and the number of leads sales can follow up well. Require a minimum fit score and a minimum engagement score, review them each quarter, and send demo requests to a person at once.

### What is a predictive lead scoring model?

A predictive model learns from past outcomes instead of points you write. It needs history. [HubSpot](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) says an AI-built score needs a sample of at least 50 contacts, 25 converted and 25 not. With less history than that, a points model like the one on this page is the practical start, and you can swap a trained model in later and keep the same routing.

### How does lead score decay work?

Decay lowers the points of older actions so the score reflects recent behavior. In [HubSpot](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool) you set it per event, with intervals of 1, 3, 6 or 12 months, and points fall in a straight line. Fade weak actions such as email clicks fast and strong ones such as demo requests slowly, and match the window to your sales cycle. Re-check fit data instead of decaying it. See [B2B data decay](https://enrich.chat/blog/b2b-data-decay/).

### Can Enrich.chat score leads?

No. Enrich.chat does not score leads or track engagement. It fills the fit half: industry, headcount, country and tech stack for a company website, and a job title for a person. Your CRM or spreadsheet applies the points. Fit data for one lead takes three lookups, 15 credits when all are found, and a lookup that finds nothing uses 0. See [pricing](https://enrich.chat/pricing/).

## Sources

1. [Overview of the lead scoring tool](https://knowledge.hubspot.com/scoring/understand-the-lead-scoring-tool)
2. [Build lead scores to qualify contacts, companies, and deals](https://knowledge.hubspot.com/scoring/build-lead-scores)
3. [Lead scoring](https://docs.oracle.com/en/cloud/saas/marketing/eloqua-user/Help/LeadScoring/LeadScoring.htm)
4. [Configuring the thresholds for profile and engagement scores](https://docs.oracle.com/en/cloud/saas/marketing/eloqua-user/Help/LeadScoring/Tasks/ConfiguringLeadScoringThresholds.htm)
5. [Lead Scoring and Grading in Account Engagement: get started with lead scoring](https://trailhead.salesforce.com/content/learn/modules/lead-scoring-and-grading-in-account-engagement/get-started-with-lead-scoring)
6. [Lead Scoring and Grading in Account Engagement: get started with lead grading](https://trailhead.salesforce.com/content/learn/modules/lead-scoring-and-grading-in-account-engagement/get-started-with-lead-grading)
7. [Lead Scoring Model Guide for Better Sales Focus](https://web.inbound281.com/blog/lead-scoring-model-guide)
8. [Lead Scoring Model Essentials: Three-Part Primer](https://conceptltd.com/blog/lead-scoring-model)
9. [Lead Scoring: Definition and How to Track It](https://kissmetrics.io/glossary/lead-scoring)
10. [Lead Scoring Analysis: Methods & Optimization](https://count.co/metric/lead-scoring-analysis)
11. [Lead Scoring: Simplify the Process With Leadfeeder](https://www.leadfeeder.com/blog/sales-prospecting/sales-intent-score/)
12. [Foundational Marketo User Group: Score Smarter, Not Harder](https://experienceleague.adobe.com/en/on-demand-events/scoring)
13. [Lead Score vs. Behavior/Demographic Score (Why Both?)](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/lead-score-vs-behavior-demographic-score-why-both-196894)
14. [Lead score limit for individual and account](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/lead-score-limit-for-individual-and-account-178836)
15. [Lead Scoring Expiration](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/lead-scoring-expiration-164600)
16. [Never Decay Your Behavior Score](https://experienceleaguecommunities.adobe.com/adobe-marketo-engage-27/never-decay-your-behavior-score-164638)
17. [How fast does B2B contact data decay?](https://www.lusha.com/blog/b2b-data-decay-rate-measured/)
18. [Ehrenberg-Bass: 95% of B2B buyers are not in the market for your products](https://marketingscience.info/news-and-insights/ehrenberg-bass-95-of-b2b-buyers-are-not-in-the-market-for-your-products)
19. [Enrich.chat agents.md](https://enrich.chat/agents.md)

## Try this prompt

Score the fit side of a lead list

```text
Use www.enrich.chat to build the fit side of a lead score. For each lead in my list, take the work email, use the part after the @ as the company website, and look up the company for its industry, employee count and country, look up its tech stack, and look up the person for their title. Skip personal mailboxes such as Gmail and mark them not looked up. Then score fit out of 100: logistics industry 30, 200 to 1,000 employees 25, United States 15, runs HubSpot 10, title is VP or director of operations 20, under 50 employees minus 50, with a floor of 0. Mark any field not found as unknown, not no, show the best-case score, and report the total credits used. Do not score engagement.
```

## Related

- [Lead qualification frameworks compared](https://enrich.chat/blog/lead-qualification/)
- [Firmographic vs technographic data](https://enrich.chat/blog/firmographic-data/)
- [15 sales signals and how to act on each](https://enrich.chat/blog/sales-signals/)