Buying signal scoring Claude skill

Use this to turn a noisy stream of events into a short list of accounts to contact today, with a reason and a play for each.

Inputs

• The signals available: job listings, leadership changes, funding, web visits, content engagement, third-party intent, technology changes, product usage for existing users.
• The ICP and account tiers.
• The offer, so each signal can be tied to a reason to talk.

The play

1. Build the signal catalogue

List every signal you can detect. For each, write the "so what": why this event makes your offer more relevant now. Signals without a clear "so what" are noise, however easy they are to collect.

2. Weight by strength and freshness

Strength reflects how often the signal precedes a real opportunity. Freshness reflects how fast it decays. Read references/signal-weights.md for the starter weights, decay curves and the stacking rules.

3. Stack signals per account

One signal is a hint; two or three on the same account in the same month is a pattern. Score = sum of decayed signal weights × fit multiplier (Tier 1: 1.0, Tier 2: 0.7, Tier 3: 0.4). Accounts outside the ICP get no multiplier and do not make the list, however loud their signals.

4. Pick the play

Each signal type maps to a play: who to contact, the angle, the channel and the timing. A new VP gets a "first 90 days" angle; a pricing-page visit gets a fast, low-pressure helpful touch; a hiring wave gets a "before your new team starts" angle.

5. Produce the daily list

Rank by score, cap at the team’s capacity (typically 15-30 signal-led touches per rep per day), and include the evidence so reps can write in seconds.

Output format

rank | account | score | signals (type - date - source) | contact | play | suggested opener

Plus: signals that fired on accounts outside the ICP (for review), and signals that decayed without action this week.

What good looks like

• Every row has evidence a rep can check in one click.
• The list is short enough to finish. A 300-row "hot list" is a cold list.
• Plays reference the situation, never the surveillance: "teams that just raised often…" not "I saw you were on our pricing page".
• Stacked signals beat single loud ones. One webinar attendance is weaker than a new VP plus two relevant job listings.
• The common mistake is treating every signal as equal and blasting the same template the moment anything fires.

Rules

• MUST apply the ICP filter before ranking.
• MUST decay signals over time.
• NEVER reveal in outreach that you tracked someone’s website visit or profile view.
• NEVER contact the same person on more than one signal in the same week.

Example prompts

• "Here are this week’s signals across our 800 target accounts. Who should we contact today?"
• "Design a signal catalogue and weights for our HR software."
• "Turn these funding and hiring alerts into a ranked list with openers."

Starter weights

Calibrate after one quarter by comparing which signals preceded real opportunities.

Decay: weight × 0.5^(days since signal ÷ half-life).

Stacking rules

• Same signal type twice in a week counts once.
• Two different signal types in 30 days: add a 1.2x bonus.
• Three or more types in 30 days: 1.5x bonus and a Tier 1 treatment regardless of the account’s static tier.

Play map

Calibration table

Raise the weight of any signal converting at 2x the average, cut any converting below the average.