Why a score without reasons gets ignored
Most CRMs that offer lead scoring show a number and nothing else. A counsellor looks at "72", cannot tell whether it means "replied twice yesterday" or "filled every field but never answered the phone", and goes back to calling from the top of the list by date. The score has to explain itself or it is decoration.
Crosvia's scoring was built the other way round: the reasons came first, the number is the sum. Each of the four factors writes a short sentence about why it gave the points it gave, and the lead record shows the four sentences beside the total. A manager can audit a ranking in ten seconds; an agent knows what to open the call with.
The four factors
The score is out of 100 and is the sum of four capped factors. The weights below are the defaults in the rules driver.
Recency (up to 30 points)
How recently anything happened on the lead: an inbound message, a call, a note, a stage change. A lead gets full marks for the first two days, then the points slide linearly to zero at 30 days. Touched today or yesterday, 30; touched 15 days ago, 16; untouched for a month, nothing.
The reason line reads "Last activity 2 days ago" or "Created 12 days ago, no activity since". Linear rather than exponential decay was a deliberate choice: counsellors reason in days, and a curve that drops 80 percent in the first week feels wrong to the people using it.
Source (up to 20 points)
Where the lead came from, multiplied by a fixed quality factor. Referrals score 1.0 and walk-ins 0.95; WhatsApp 0.8; paid campaigns 0.7; website forms 0.6; anything else 0.4; bulk imports 0.25. The reason line is simply "Source: referral" or "Source: paid campaign".
The table is the same for every workspace, so a referral is worth the same 20 points in your account as in anyone else's. Per-workspace tuning is on the roadmap; until it lands, the factor is something to rely on rather than something to configure.
Engagement (up to 30 points)
What the lead has done: replies to your messages, connected calls, appointments booked or attended, documents uploaded. Inbound counts more than outbound: a reply or a connected call is worth 4 raw points and an appointment 5, while everything you send adds at most 5 raw points in total. Ten unanswered messages therefore cannot inflate a lead.
Around 20 raw engagement points — two replies, two connected calls and an appointment, for instance — saturate the factor. A lead that engaged is already maximally hot on this axis, and the other three factors decide the rest.
The reason line lists what was found, such as "2 replies, 1 connected call, appointment booked", or "No engagement yet".
Completeness (up to 20 points)
How much of the lead profile is filled: phone and email carry half the weight between them, then destination country, programme interest, deal value and at least one pack section. The reason names only what is missing from the contactable set, so it reads "Missing email, interest" or "Profile complete".
Completeness is capped at 20 on purpose. It is a weak signal, because a counsellor can fill every field on a lead who will never answer, and the score should not reward data entry as if it were interest.
A worked example: three leads on a Tuesday morning
Lead A, Riya. Paid campaign three days ago, replied once to the first template, no call yet, phone, email, country and programme on file.
- Recency: last activity 3 days ago, 29 of 30.
- Source: paid campaign, 14 of 20.
- Engagement: 1 reply plus the one template we sent, 8 of 30.
- Completeness: profile complete on the contactable fields, 15 of 20 — deal value and the pack sections are still empty.
- Score 66. Reason summary: active recently, replied once, never called.
Lead B, Arjun. Referral from an enrolled student two weeks ago, one connected nine-minute call, appointment booked, every field filled, but nothing since the call.
- Recency: last activity 14 days ago, 17 of 30.
- Source: referral, 20 of 20.
- Engagement: 1 connected call, 1 appointment, 18 of 30.
- Completeness: profile complete, 20 of 20.
- Score 75. Reason summary: strong source, engaged, but going quiet.
Lead C, Fathima. Website form yesterday, nothing sent yet, only name and phone.
- Recency: created yesterday, no activity since, 30 of 30.
- Source: website, 12 of 20.
- Engagement: none yet, 0 of 30.
- Completeness: missing email, interest, 5 of 20.
- Score 47. Reason summary: brand new, nothing known yet.
The call queue that morning puts Arjun first among the hot uncalled leads, because the reasons say a good source is cooling and one call will likely either book the enrolment conversation or mark the lead lost. Riya is next; the reason says "never called", which is a clear action. Fathima comes third, and her reason tells the agent the first job is to learn what she wants.
Notice that the order is not the same as the order by score alone within a task list, because the queue's first three bands (awaiting disposition, overdue, due today) sit above all of this. Scoring decides what happens when the explicit commitments are done.
What the reasons change in practice
- Agents stop asking "why is this one on top?". The answer is on the screen.
- Managers can spot drift. If half the top-20 reasons read "Source: bulk import", the problem is the list you bought, not the score.
- Coaching becomes specific. "You called six leads whose reason was no engagement yet before the two that had replied" is a conversation a team lead can have in two minutes.
- The AI reply drafts on the Growth plan read the same reasons. A draft for a lead whose reason says "appointment booked, no activity since" opens differently from one whose reason says "replied once, never called".
What the score is not
It is not a conversion probability. Nothing in the four factors knows whether the person has the budget or the visa eligibility.
It is not fixed: it is recomputed whenever the lead changes, and at least daily for recency decay. And it is not hidden: every computation writes the breakdown to the lead, so the history of a score is visible in the timeline.
What to do next
- See the queue the score feeds: Call queue.
- See which pack fields feed the completeness factor: Verticals explains the fields per pipeline.
- Let reasons shape the first message: AI reply drafts.
- Related terms: lead score, lead source, call disposition.
- Read why a queue beats a task list for counsellors: Call queue vs task list.