HubSpot Scoring — lead-triage

Field names, scoring weights, and ICP defaults.

Fields to pull

Company fit fields live on the COMPANY record, not the contact

industry and numemployees exist on the contact object but are almost never filled in. The populated copies live on the company associated with the contact. Pull those instead:

Fetch the associated company for each lead, then read fit from there. Only fall back to the contact-level industry / numemployees if there is no associated company at all.

The enrichment leg: when there is no associated company

When every lead in a portal has no company association, company fit scores 5 across the board and the dimension contributes nothing. With a lead-data connector connected — Apollo or Clay, peers per ../../../shared/connector-neutrality.md — fill the gap from the lead's email domain instead of scoring it flat:

1. Collect the leads with no associated company and a business email domain (skip gmail, outlook, yahoo, icloud and the like — nothing to enrich).
2. Say the count and ask once. Enrichment is metered: each row spends credits. "32 leads have no company on file. Enriching them from their email domains costs roughly 32 credits in Clay. Go ahead?" Run only on a yes for this run; a no means the flat ladder below applies to those rows.
3. Enrich by domain:
4. Clay: find-and-enrich-company per domain, reading industry and employee count from the returned company data points. Batch through find-and-enrich-list-of-contacts when the connector offers it for the whole set.
5. Apollo: apollo_organizations_enrich per domain, reading industry and estimated_num_employees.
6. Score company fit from the enriched values using the same table as an associated company. Mark those rows "fit from enrichment" in the caveats.
7. Never write the enriched fields back to the CRM from this skill. Offer it as a crm-autopilot hygiene run afterward, behind its write gate, if the owner wants the company records created.

Without a lead-data connector, or on a no, the flat-score ladder stands.

Fetch note bodies for the urgency scan

The urgency keywords live in note bodies, and note bodies are engagements — they never arrive with the contact property pull. num_notes only says a note exists; the +15 keyword signal cannot fire unless the bodies are actually read.

How to do it without flooding the API:

1. Score every lead first without the note-keyword signal (provisional score).
2. Fetch associated notes only for the top ~25 leads by provisional score — or for every lead with num_notes > 0 when fewer than 25 have notes.
3. Pull via the contact → notes association and read the note body (hs_note_body). One page of notes per lead, most recent first, is enough.
4. Scan the bodies for the urgency keywords and re-score with the +15 where a keyword hits. Say in the caveats when notes existed but were not fetched (beyond the cap) rather than claiming they contained nothing.

Scoring model (0–100 composite)

Four dimensions, each 0–25. Sum for composite.

Engagement (0–25)

Only count signals from the last 30 days. Older signals score 0.

Company fit (0–25)

Default ICP if owner hasn't stated one: any industry, 1–50 employees.

Urgency (0–25)

Recency penalty (subtracted from composite)

Custom ICP (runtime override)

If the owner states an ICP at runtime ("focus on SaaS, 10–100 employees"), apply it for that session only.

When the scores come out flat

A CRM with no logged engagement scores every lead the same: ten leads, ten identical composites of 0. Sorting identical scores produces arbitrary order, and presenting that as a priority list is worse than saying nothing.

Check for this before showing a ranked list. If the top and bottom composite differ by less than 10 points, the model has not separated anything. Say so plainly:

"These leads score almost identically — your CRM has no logged emails, calls, or site visits for them, so the engagement and fit signals are all empty. Here they are ordered by what the CRM does know."

Then order by this ladder instead of by composite, and label it as such:

1. hs_lead_status — NEW before IN_PROGRESS before empty
2. hs_is_unworked = true — nobody has touched it yet
3. hs_buying_role — contains DECISION_MAKER, then CHAMPION, then the rest
4. jobtitle — seniority words (VP, Director, Head, Chief, Owner, Manager)
5. num_associated_deals ≥ 1 — pipeline already attached
6. createdate — newest first

Never present ladder order as a score. Show it as "ordered by CRM signal" and drop the numeric score column entirely, so the owner is not reading precision that does not exist.