How lead scoring works
What goes into an Embers lead score, what raises and lowers it, how grades and trend are set, and why a lead you like can still rank low.
Every person Embers finds gets one score from 0 to 100. It is a ranking device, not a grade of the human being: it answers “who should I read first”, and it is built from behaviour you can go and verify on LinkedIn.
The four ingredients
The score multiplies four things together:
- An engagement score from 0 to 100, built from how the person engaged.
- An ICP multiplier from how well they fit your ideal customer profile.
- A confidence factor that discounts thin or unclassified data.
- A company boost when several people from one company engage.
The result is capped at 100 for display. Behind the scenes an uncapped version is kept for ordering, so the list stays stable even when several leads reach the top.
What the engagement score measures
Five sub-scores, each normalised against everybody else in your account, then weighted:
| Sub-score | Weight | What lifts it |
|---|---|---|
| Frequency | 25% | Weighted engagement volume relative to how much you published |
| Recency | 25% | Recent engagement, decayed with a 14 day half-life |
| Depth | 20% | Variety of engagement, comments, reposts, comment intent, and escalation over time |
| Consistency | 15% | The share of weeks in the window with any engagement |
| Responsiveness | 15% | A short median gap between your post and their comment |
Not all engagement counts equally. A repost carries the most weight, then a comment, then the more considered reactions, with a plain like weighted lowest. Comments that read as buying intent are weighted up. Escalation matters too: someone who moved from likes to comments to a repost scores higher than someone flat at the same volume.
Engagement on a viral post is discounted, because a post far above your median attracts casual likers who are not signalling anything.
The look-back window depends on your plan: 90 days on Free, 365 days on Solo and above. When your plan changes, everything is rescored silently against the new window, with no notifications for movement nobody caused.
What raises and lowers a score
Raises it
- Comments that show purchase intent or name a pain you solve.
- A recent role change: starting a job within about 90 days boosts the recency component and is called out as a signal.
- Engaging across several weeks rather than in a single burst.
- Two or more people from the same company engaging. The boost grows with the number of colleagues and stops at 1.5 times.
- A confident ICP match, which is the largest single lever in the whole formula.
Lowers it
- A confirmed ICP non-match. Negatives are suppressed hard, far harder than uncertainty is.
- Very little data. Fewer than three engagements applies a discount, and so does never having been classified against your ICP.
- Engagement only on a post that went unusually wide for you.
- Going quiet. Recency decays continuously, so a strong lead cools without any action from you.
Grades and trend
The grade in exports and destinations is the score in bands: 80 and above is A, 60 is B, 40 is C, 20 is D, and anything below that is F. Sixty is the line that matters most in the product, because Hot Leads and the Priority Queue’s new-signal band both start there.
Trend compares the current score with the previous snapshot. A rise of 15 percent or more is rising, a fall of half or more is cooling, and everything between is stable. Cooling leads collect in the Going Cold tab. Trend needs at least two snapshots, so a brand new lead is always stable at first.
Alongside the score, each lead carries up to three plain-English signals, for example “Shows buying intent in comments” or “2 others from their company also engage”. Read those before the number.
Why a lead you like ranks low
The usual causes, in order: they have not been classified yet, they have fewer than three engagements, their only engagement was a like on a post that went wide, or the classifier read their title or industry as a mismatch. The fix for the last one is ICP feedback, which overrides the verdict and rescores the lead straight away.
Scores and destinations
Score is one third of the rule that decides whether a lead leaves the product. A lead is sent to Slack, HubSpot, or a signed webhook when its score is above 0, and it is an ICP match or you approved it by hand, and its company is not on your blocklist. Fail any one of those and nothing is delivered.
Related
Frequently asked questions
- Why is my score capped at 100?
- The internal ranking number is uncapped so ties break cleanly. The number you see is that value clipped to 100, which is why two leads can both show 100 and still sit in a fixed order.
- Does Embers score what a lead posts themselves?
- Only the keyword tracking agent looks at what someone posts. Post and comment scoring measures how they engaged with tracked content, not their own feed.
Was this article helpful?
What went wrong?
Thank you. We read every answer.
Still stuck?
Send us the article you were following and what happened. We answer from the same inbox that writes these pages.