score-accounts

name	score-accounts
description	Rank a list of accounts by signal strength to surface the highest-priority targets for outreach. Reads native fit + urgency scores from Saber when available, and falls back to client-side weighted scoring for Apollo, HubSpot, or pasted data.

Score Accounts

Use this skill to rank accounts so outreach focuses on the highest-intent targets first.

Two paths:

• Path A — Saber CLI: read pre-computed fit and urgency scores from native scoring. No client-side math.
• Path B — any other source (Apollo / HubSpot exports, pasted data): client-side weighted scoring against the model from generate-signals.

For scoring concepts (dimensions, profiles, contributions, auto-trigger), see _shared/scoring.md.

Step 1 — Determine the path

Run saber --help to confirm the CLI is installed.

If installed:

saber list company list
saber scoring profile list

If a relevant company list and at least one company-scoped scoring profile exist, take Path A.

If a list exists but no scoring profile is assigned to its accounts, route to configure-scoring first, then return here.

If not installed: take Path B with whatever signal data the user has.

---

Path A — Read native scores

A1. Pull the list

saber list company companies <listId>

Capture the domains.

A2. Read scores

Read scores for all domains in a single call (the --object flag is repeatable):

saber scoring scores --type company \
  --object acme.com --object stripe.com --object ramp.com [...] \
  --detailed

--detailed returns per-rule contributions and the previous score (for Δ display). Reading scores does not consume credits.

You'll get one row per (profile, domain, dimension). Group by domain to get fit and urgency side-by-side.

A3. Decide the sort

Ask the user how to rank when both dimensions exist:

• Combined (default) — (fit + urgency) / 2. Balanced.
• Fit-first — sort by fit desc, break ties by urgency. Best for cold outbound where ICP-match comes first.
• Urgency-first — sort by urgency desc, break ties by fit. Best for time-sensitive plays (event-triggered campaigns).

A4. Present

## Account Scores — [List Name]

### High priority — combined ≥ 70
| Rank | Company | Domain | Fit | Urgency | Δ | Top contributing rules |
|---|---|---|---|---|---|---|
| 1 | Acme Corp | acme.com | 84 | 72 | +6 | New VP Sales hired (+20), Hiring SDRs (+15) |
| 2 | Beta Inc | beta.io | 76 | 80 | +4 | Series B (+25), HubSpot migration (+15) |

### Medium priority — combined 50–69
| Rank | Company | Domain | Fit | Urgency | Δ | Top contributing rules |
|---|---|---|---|---|---|---|
| ... | | | | | | |

### Watch list — combined 30–49
| Rank | Company | Domain | Fit | Urgency | Δ |
|---|---|---|---|---|---|
| ... | | | | | |

### Deprioritise — combined < 30
| Company | Domain | Notes |
|---|---|---|
| ... | | |

Include the Δ (score - previousScore) — it's the single best signal for "what changed since last week".

A5. Watch for low coverage

If signalCoverage is much lower than totalRules for many objects, the underlying signals haven't fully run yet. Note this in the output:

"[N] accounts have partial signal coverage — scores will firm up as remaining subscriptions complete."

Auto-trigger keeps scores fresh; you generally don't need to call saber scoring compute.

---

Path B — Client-side weighted scoring

Used when there's no Saber CLI or no scoring profile is set up. Works with Apollo enrichment, HubSpot exports, manual research — any signal data the user can describe.

B1. Get the signal data

Ask the user for their data. Accepted formats:

• Paste a table or CSV with columns: Company, Domain, [Signal 1], [Signal 2], ...
• Describe what they know per account ("Acme raised a round, Beta is hiring SDRs")
• HubSpot export, Apollo enrichment output

Confirm which signals are present and what a positive result looks like.

B2. Apply the weighted algorithm

If signal metadata from generate-signals is in conversation context, use the full model:

1. For each signal, determine if the answer is positive per interpretation rules
2. earnedWeight = sum of weights for positive-answer signals
3. totalWeight  = sum of all signal weights
4. rawScore     = (earnedWeight / totalWeight) × 100

5. Apply penalties:
   — Any disqualifier fires wrong  → score = 0 immediately
   — Zero buying_signal hits       → score = 0 (no pain evidence)
   — Zero urgency hits             → score − 15 (right fit, wrong time)

6. Final score: 0–100

Thresholds:

• 70+ — High fit, prioritise outreach
• 50–69 — Moderate fit, worth pursuing
• 30–49 — Low fit, monitor for triggers
• < 30 — Poor fit, deprioritise

If no metadata is available, ask which signals matter most, then:

• Base: +1 per positive signal
• Priority signals: 2× weight for the user's flagged signals
• Recency bonus: +0.5 if from the last 7 days
• Normalise to 0–10 relative to the highest scorer.

B3. Present

Same output structure as Path A's table, but with one combined score column instead of separate fit + urgency.

---

Step 4 — Suggest next steps

• High priority accounts: use write-outreach to draft personalised messages referencing the top contributing signals
• Watch list: re-read scores in 2–4 weeks; auto-trigger will have updated them as signals run
• Low scorers: pause or remove from active list; revisit if a trigger fires
• Use deal-coaching on any high-priority account already in pipeline
• Use manage-scoring to tune rules if the rankings don't match your gut

Related

• _shared/scoring.md — concepts and CLI reference
• configure-scoring — first-time scoring setup
• manage-scoring — tune rules and assignments
• qualify-inbound — single-lead version of this flow