I need better leads.
Call the twenty who are ready.

The Whale Finder — demo run

Input: sample-crm-export.csv, 40 accounts, no ICP explicitly provided.

Inferred ICP (per the rubric's instruction to infer from existing "Customer"-stage accounts when none is given): four accounts are marked Customer — Northbeam-adjacent Software companies at 115-220 employees (Peregrine Data Systems, Cinderpath Analytics), plus two larger outliers in Retail (900 employees) and Insurance (1,200 employees). Core band taken as 100-250 employees, Software-weighted; Retail and Insurance accepted as fit-qualifying verticals on the strength of those two accounts, but scored strictly — no vertical without a Customer record on file got fit credit for industry match, even where the signal looked good (Fintech, Healthcare, Agriculture, Logistics all got scored on size/relationship/completeness only, not vertical).

15 of 40 accounts dropped at the fit gate (fit < 4/10) — mostly enterprise manufacturing/logistics far outside the size band, and a handful of list-sourced records with no real data to score honestly (Briarcliff Manufacturing, Nettlebed Media, Quillfeather Design Co had near-blank rows and were dropped rather than guessed at).

The top 20

Honesty check

• 15 of 40 dropped at the fit gate, 3 of those because the record was essentially blank (list-sourced, no real data to score) rather than because they were disqualified on merit.
• Signal data was genuinely usable for most accounts — the free-text notes field carried real trigger events (funding, hires, inbound activity), which is what made specific "why this week" reasons possible instead of generic scores.
• What to start capturing for sharper output next time: an explicit ICP definition (this run had to infer one from four Customer records, which is thin evidence — Retail and Insurance both got included on a single account each), and a structured "last activity type" field instead of only free-text notes, so signal strength doesn't depend on whoever happened to write a good note.

One thing worth flagging back to the founder

The rubric scored vertical fit strictly — only Software, Retail, and Insurance got fit credit, because those are the only verticals with a Customer record. That correctly kept enterprise Manufacturing out, but it also fit-penalized some of the best signals in the file: Ashgrove Fintech, Bellwether Insurance Tech's near-miss on Financial Services scoring, and Ravenscroft Biotech all had strong triggers in verticals DEVENTA hasn't sold into yet. Worth deciding deliberately whether The Whale Finder should treat "no evidence yet" the same as "not a fit" — right now it does, which is the conservative and defensible default, but it's a real product decision, not an accident of the code.

Cold Outbound Diagnostic — demo run

Input: sample-questionnaire.md, Thistledown Scheduling, all 20 questions answered.

Score

22 / 60 — Mixed Signal. Real practice exists in this team — a written segmentation, an enforced call-task trigger — but it's sitting underneath five active theatre patterns that are almost certainly costing meetings, not just looking inefficient on paper.

Patterns triggered

The Template Loop

Q3=C (2), Q4=A (0), Q5=B (1), Q17=B (1) — 3 of 4 at theatre level.

Someone on this team is doing real homework: Q3 says reps check the clinic's site and reviews before the first touch. But Q4 confirms that homework never reaches the message — the first-touch email is one merge-field template, and Q5 confirms only two pre-approved openers exist team-wide. The research is being thrown away between the browser tab and the send button. That's the most expensive version of this pattern, because the team is already paying the time cost of research and getting none of the return.

Next step: stop treating "research" and "writing the email" as separate steps done by different amounts of attention. Require one specific, checkable fact from the research step to appear in the first line of the email — not industry, the clinic's own detail — before a message is allowed to send.

The Follow-Up Cliff

Q6=B (1), Q7=B (1), Q18=B (1) — 3 of 3 at theatre level.

Four touches, all saying the same thing reworded, and a mid-sequence reply that goes quiet just gets noticed "eventually" with no process. This isn't a volume problem — it's that the fourth touch carries no more information than the first one did, so a prospect who didn't respond to touch 1 has no real reason to respond to touch 4.

Next step: touch 3 or 4 has to carry something touch 1 didn't — a different proof point, a different channel, a reference to what the prospect actually did (opened twice, visited pricing) — not a rewrite of the same sentence.

The Slow Reply

Q10=B (1), Q11=B (1) — 2 of 2 at theatre level.

This is the pattern most worth fixing first. Every other pattern on this list describes wasted effort generating interest that was never going to convert. This one describes real interest — a prospect who replied — being handled by "whoever's logged into the shared inbox" with no SLA and no specific answer to what was actually asked. The team is spending budget and rep-hours to generate the one signal that matters, then handling it worse than the cold outreach that produced it.

Next step: route replies to a named owner, not a shared inbox, with a same-business-day answer target that gets tracked, not assumed.

The Vanity Metric Trap

Q12=A (0), Q13=B (1), Q19=A (0) — 3 of 3 at theatre level.

Dials and emails sent lead the Monday standup. A rep with high activity and low bookings is told to push more volume, not audited on message or targeting. And if outbound stopped entirely, the first thing anyone would notice is the dashboard going quiet — not a drop in booked meetings. This is very likely the root cause underneath the other three patterns: if activity is the number that gets rewarded, a team will rationally keep sending the same template at the same list rather than slow down to fix either one.

Next step: move "meetings booked" above "activity sent" in the standup deck itself — literally reorder the slide. What leadership looks at first is what the team optimizes for by Friday.

The Static Loop

Q15=B (1), Q16=B (1), Q20=B (1) — 3 of 3 at theatre level.

The message was last properly reviewed against reply data roughly eight months ago. Low-reply stretches get a subject-line swap, not a rewrite. This means whatever is underperforming right now has likely been underperforming for the better part of a year without a structured re-check.

Next step: put a recurring calendar review on the actual message against actual reply/booking data — monthly is enough — before touching cadence or list quality again. Fixing list or cadence against a message that hasn't been checked in eight months is fixing the wrong layer first.

Patterns not triggered — genuine strengths

The Spray List — not triggered (Q1=B/1, Q2=C/2, Q14=C/2; only 1 of 3 at theatre level). Lists come from HubSpot with a manual glance, not a cold purchase, and segmentation by clinic type with a distinct pain-point line per segment is real, defensible practice. This is worth protecting as the team fixes the rest — don't let a fix to messaging or cadence erode the segmentation discipline that's already working.

The Channel Silo — not triggered (Q8=C/2, Q9=C/2; 0 of 2 at theatre level). The sequence tool enforces an automatic call task on day 5 and reps mostly follow it. Multi-channel logic with a real trigger rule is a genuine strength most teams this size don't have.

Honesty check

All 20 questions were answered, so every pattern's gate is computed on complete data. If the team fixes exactly one thing, fix the Slow Reply pattern first — it's the only one where the team is failing to capture demand it already paid to generate, and it's a two-question, low-effort process change (an SLA and a named owner) rather than a rewrite of anything. The Vanity Metric Trap is the more structural fix, but it's slower to land and its benefit shows up mainly by making the other three patterns easier to justify fixing, not by itself producing a meeting next week.