Lead scoring
Every lead gets a score from 0 to 100 with the reasons shown, so sales calls the right people first. The score orders the list; it never assigns, disqualifies or closes a lead.
Three different things. This page is only the first.
AI lead score This page
How likely a lead is to buy soon: purchase intent. From 0 to 100, built from the six signals below. It orders the call list. It says nothing about whether a customer pays on time.
How likely a lead is to buy soon: purchase intent. From 0 to 100, built from the six signals below. It orders the call list. It says nothing about whether a customer pays on time.
Customer reliability rating Customer rating
How a customer has behaved so far: payment history, cancellations, completed orders. A grade from history under a published rule version. It may lead to a proposal to change terms, which Finance approves. Not an AI output.
How a customer has behaved so far: payment history, cancellations, completed orders. A grade from history under a published rule version. It may lead to a proposal to change terms, which Finance approves. Not an AI output.
Loyalty status Loyalty
Where a customer stands in the loyalty programme: tier and points earned from eligible orders. It gives permitted benefits. It is not a judgement of the customer and is not used to score a lead.
Where a customer stands in the loyalty programme: tier and points earned from eligible orders. It gives permitted benefits. It is not a judgement of the customer and is not used to score a lead.
| Same organisation, three separate values | AI lead score purchase intent of one contact | Reliability rating history of the account | Loyalty status programme |
|---|
All scoring inputs are on this page and nothing hidden is used. The score is a points table: six signals with their weights, shown under “Signals and weights”, and each lead’s points with the evidence behind them (select a score). The rating and the loyalty tier are not inputs. Neither are nationality, gender, age or personal email domain. See also customer analytics.
Open leads scored
203
Model version 6, live since 14 September
Hot leads (70 or above)
41
+9 this week
Hot leads contacted within 1 day
83%
7 hot leads not yet contacted
Sales agreed with the score
78%
Of 142 leads rated by sales, 90 days
Signals and weights
Weights add up to 100Persona fit
Role matches a learning path: piping, stress, layout, integrity, engineering manager. L&D and training managers score highest; students lowest.
25
Company size and team size
Engineers to train, from the form or chat; company headcount when known. A team of 10 or more gets the full weight.
20
Course interest
Named a course or path, asked for dates or a proposal. Extra when a batch for that course starts within 45 days.
20
Webinar attendance
Registered: 5. Attended live: 10. Stayed to the end or asked a question: 15.
15
Pages viewed
Course, schedule, pricing and corporate pages in the last 30 days. Capped so heavy browsing alone cannot make a lead hot.
15
Recency and replies
Replied to an email or WhatsApp message in the last 14 days. Decays to zero after 60 days of silence.
5
Not used on purpose: nationality, gender, age, personal email domain, the customer reliability rating, loyalty tier or points. Country is used only to route the lead to an owner.
Score distribution, open leads
ColdWarmHot
Thresholds: warm from 40, hot from 70. Hot leads create a task for the owner due the same working day.
Cold, below 40
87
Nurture journey only
Warm, 40 to 69
75
Call within 3 days
Hot, 70 or above
41
Call today
Insight. 14 open leads asked about pipe stress in the last 60 days and 5 are hot. Batch B-2614 (, 8 November) needs 3 more seats to break even. Open batch or draft a follow-up.
Top scored leads
Sample data| Score | Lead | Company | Interest | Strongest signals | Change, 7 days | Owner | Status |
|---|
Showing 12 of 203 open leads, highest score firstAll leads in CRM
Did the score predict what happened?
Leads created 1 July to 30 September| Score band when first contacted | Leads | Became a deal | Won | Lead to deal | Average deal |
|---|
A working score shows the conversion rate rising with each band. It does here, but the 40 to 69 band is wide; see the suggestion on the right.
Feedback from sales
| Date | Lead | Score | Sales rating | Comment | By |
|---|---|---|---|---|---|
| 2 Oct | Mostafa Helmy Sample EPC Contractors | 88 | Right | Budget holder, 12 engineers, wants November. | Karim Fathy |
| 1 Oct | Sara Wahba Engineering student | 58 | Too high | Viewed 30 pages and attended two webinars but has no budget. Students should be capped. | Salma Adel |
| 30 Sep | Adel Mansour Sample Gas Pipelines | 84 | Right | Asked for an in-house PD370 proposal. | Karim Fathy |
| 29 Sep | Procurement contact Sample Power Generation | 34 | Too low | Sent a formal tender by email, never visited the site. Tender leads need their own signal. | Karim Fathy |
| 27 Sep | Rana Shawky Sample Fertilizers | 62 | Right | Individual buyer, enrolled in PD386 after one call. | Salma Adel |
| 24 Sep | Training coordinator Sample Refining | 76 | Right | Needs FFS for 8 inspectors; batch is full, moved to waitlist. | Salma Adel |
| 22 Sep | Laila Kamel Fresh graduate | 45 | Right | Interested in ML523, asked about instalments. | Salma Adel |
| 18 Sep | Yassin Gamal Sample Gas Pipelines | 72 | Too high | Already covered by Adel Mansour's deal. Duplicate contact at the same account. | Karim Fathy |
Sales rates a score with one click on the lead page. Ratings are reviewed monthly before any weight changes.
Agreement by month
Share of rated leads where sales chose "Right".
Suggested changes
From feedbackCap students at 45. 9 of 11 "too high" ratings were students or fresh graduates with heavy browsing.
Add a signal: formal request received. Tender or RFQ emails logged by sales would add 20 points regardless of web activity.
Score the account, not only the person. When two contacts belong to one corporate account, show one combined score to avoid duplicate calls.
Suggestions change nothing until the owner of lead scoring approves a new version.
Where ratings came from
Model versions
Audit log| Version | Live from | What changed | Approved by | Agreement | State |
|---|---|---|---|---|---|
| v6 | 14 Sep 2026 | Course interest raised from 15 to 20; bonus when a matching batch starts within 45 days | Karim Fathy | 78% | Live |
| v5 | 10 Aug 2026 | Pages viewed capped at 15; pricing page counts double | Karim Fathy | 72% | Retired |
| v4 | 6 Jul 2026 | Webinar attendance split into registered, attended, stayed | Dina Sabry | 68% | Retired |
| v3 | 1 Jun 2026 | Team size added to company size | Karim Fathy | 63% | Retired |
| v2 | 4 May 2026 | Removed personal email domain as a negative signal | Mahmoud Abdallah | 58% | Retired |
| v1 | 6 Apr 2026 | First rule-based model: persona, company size, pages viewed | Mahmoud Abdallah | 54% | Retired |
In the prototype the scores are sample values: no model runs. The model is a transparent points table, not a black box. A statistical model will be considered once there are at least 500 closed deals to learn from.