SANITIZED BUILD RECEIPT · builtgtm.ai/builds
You should never open a list and wonder “is this account any good, and why?” We answer that for you. One number, the reason behind it in plain English, and the contact already found. This page is the short version: what your score means, where we get the data, how it adds up, and how to turn it into a sent sequence in minutes.
We score 5,000+ accounts so you don’t have to guess. We check 8 sources per account, cross-check our own CRM against the truth, and only spend money on the accounts you’re about to work. When a lead lands on your list, it has already cleared the bar — your job is to reach out, not to research. If the data is ever thin, the score says so out loud. No black box.
The score is built in three layers that never compete with each other. Read it left to right:
Why layered, not one big average? Because signals with different fill rates would let the most common one dominate. Gates, fit, and live behavior are kept separate on purpose — a dealbreaker stays a dealbreaker.
| Gate | Effect | Why |
|---|---|---|
| Gmail / Google Workspace | × 1.00 | Prerequisite for our motion. 86% of wins vs 31% of losses (2.67×). |
| Email unknown | × 0.90 | Small haircut for the unconfirmed. |
| Microsoft email | × 0.55 | 0.34× won/lost — disqualifying-grade. The score says so instead of hiding it. |
| Octave disqualifier | cap 40 | A qualitative dealbreaker outranks every number (full list in §06). |
| Zero sales-motion fingerprints | cap 55 | No CTA, no sales hiring, no sales tooling = nothing to sell to. |
The email answer comes from MX records (a DNS lookup) on every domain — free, authoritative, can’t go stale silently. Salesforce’s Email_Provider__c is cross-checked against it, never trusted alone.
Each component is a question about the account, answered by a waterfall of sources. Weights are proportional to won/lost lift. Missing data is renormalized over what we do have — a blank never scores as a zero.
| Weight | Question | Evidence |
|---|---|---|
| .35 | Do they run Salesforce / HubSpot? | 1.64× won 52% vs lost 32% |
| .30 | Are they hiring sales right now? | 1.55× (TheirStack, 100% fill) |
| .13 | Do they sell the way our buyers do? (demo CTA / pricing) | 1.22× + dealbreaker separation |
| .12 | Can we reach a sales decision-maker? | Reason-kept outreach feasibility |
| .10 | Are they the right size? (25–2,000 FTE) | Tested won median 108 vs lost 46 FTE |
| Live signal | Boost | Criteria + evidence |
|---|---|---|
| Already using our product (product-engagement score verdict) — loudest | +12 / +30 | Established/Power adoption. Sequence activation is the top driver — converts at 31%. |
| Company trigger (funding / new sales leader) | +6 | PredictLeads / Crustdata event in last 90d — fresh budget or a new buyer. Timing escalator. |
| Competitor tool in stack (displacement) | +5 | TheirStack — runs Outreach / Salesloft / Apollo. A switching angle. |
| AI adoption (AI Score) | +3 / +5 | 1 / 2+ AI products used. Tested 1.43× SS-paying vs free — the PLG commitment. |
Boosts are additive, never multiplicative — multiplicative escalation jammed 495 accounts into a tie at 100 in testing; additive cut that to 22.
| Band | What you do |
|---|---|
| ≥90 — Strike | The action list. Work first. |
| 80–89.9 — Hot | Real signals, just short of Strike. |
| 65–79.9 — Warm | Personalized touch, not a sprint. |
| 50–64.9 — Watch | Nurture until a signal fires. |
| <50 — DQ / thin | Disqualified or data-thin. Leave it. |
Every account is run through the same 8 sources before it’s scored. Each one answers a specific question. The CRM is the first witness, never the last word — we cross-check it against the outside world on every full pass. Common Room was retired — it fed one intent input we now get from company triggers and product usage.
Ranking every account is free or pennies: DNS, Salesforce, Amplitude, and cheap metered checks. The expensive stuff — verified phone/email and Octave’s deep read — only fires on the few hundred accounts you actually work. Nobody pays to rank an account you’ll never touch, so the budget concentrates on the leads in front of you.
Cheap-and-broad first, expensive last and only where the cheap signals point. Every answer is stored locally with a timestamp (domain · question · answer · source · validated-by). Stamps control freshness — when to re-pull — never trust: on every full pass the filled answer is re-checked against its source.
| Question | First answer → cross-validate | What it proves | Refresh |
|---|---|---|---|
| Lifecycle + commercial state | SFDC RecordType + DWH ARR (never cached) | Play gate + model | every run |
| Which email stack? | MX records → audits SFDC Email_Provider__c | Can we serve them (2.67×) | every full pass |
| Run Salesforce/HubSpot? | SFDC CRM__c → Crustdata / BuiltWith | Win-correlate (1.64×) | monthly |
| Hiring sales? | TheirStack via Deepline | Motion growing (1.55×) | weekly |
| Sell like our buyers? | PredictLeads → Firecrawl on miss | CTA + pricing (1.22×) | monthly |
| Can we reach the buyer? | SFDC contacts → FullEnrich search (free) | DM findability + SS→DS gate | full pass; waterfall on play entry |
| Right size? | Crustdata → SFDC CR_Number_of_Employees__c | 25–2,000 band | monthly |
| Already use us? | Amplitude (gp:domain) → product-engagement score | +6/+12 escalator (4.7×) | every run |
| Any company trigger? | PredictLeads / Crustdata events | Funding / new sales leader (+6 escalator) | weekly |
| ICP at all? | Octave qualify_company (shortlist) | Fit + hard dealbreakers | shortlist entry |
No score is ever handed to you naked. Every account carries a plain-English explanation, a matched play, and a contact. Read top to bottom and you have your opener:
| You get | What it gives you |
|---|---|
| Live signals first | What they’re doing in the product + any fresh company trigger. This is your opening line. |
| The fit case | CRM in stack, sales hiring, a competitor you displace, demo CTA. |
| The qualifier + caveats | “They’re on Gmail, they qualify.” And anything we’re unsure about, said plainly. |
| Product story | What they use and ignore — the expansion wedge and the AI gap line. |
| Octave play + the record | The exact messaging motion to fire, one click to the Salesforce account. |
Your Salesforce lifecycle — not the score — decides the motion. The score ranks; the play is chosen by where the account sits in its journey:
| If the account is… | Your play |
|---|---|
| An open opportunity | In-Flight — AE closes, no new sequence |
| A current customer (DS ARR > 0) | Expansion — seats + products (Save first if usage is weak) |
| A self-serve payer (SS ARR, no DS) | SS→DS Conversion — pitch on real usage |
| A churned account (Fall Off) | Win-Back — sequence built on the loss story |
| A prospect with a PQA date | PQA Re-Engage — they already know us |
| A cold prospect | Net-New Outbound — top-down by score |
| What’s the score? | 100 / 100 — top of the Strike band. |
| Where’s the data? | Amplitude (they’re active in-product), SFDC + Crustdata (CRM, size, loss history), TheirStack (hiring + SalesLoft in stack), PredictLeads (company triggers), MX (Gmail confirmed). |
| How did it add up? | Gmail gate ×1.00 · strong fit base (Salesforce in stack, hiring 16 sales roles, right size) · +12 already using the product · +6 a fresh trigger and a SalesLoft displacement angle → ceiling. |
| What do you do? | Open with what their team is doing in the company. They run Salesforce (our strongest win signal), are hiring 16 sales roles, and pay for SalesLoft today — a tool we displace. Loss context: lost March 2023 on a $15,300 renewal after 4 prior renewals. Lead with the product re-entry, not a cold pitch. |
That entire paragraph is generated for you, per account, in the brief. You don’t assemble it — you read it and reach out.
Your brief gives you the context. Three connected tools turn that context into a launched, personalized sequence — without leaving the chat. Ask in plain English; the MCPs do the work.
Score, Product Engagement verdict, the AI gap line, the best verified contact, PLAN notes, and which play it’s in — in one ask.
Octave grounds the copy in the account’s real situation — the product story, the fit case, the matched play — and writes a multi-step sequence in your voice. Not a generic template.
the company builds the sequence, enrolls your contact, and tracks opens and replies — so the follow-up is automatic and you see who’s engaging.
Score → brief → draft → launch, in one sitting. You bring the judgment and the relationship. The system brings the research, the writing, and the plumbing. That’s the deal: we did the homework so you can spend your day in conversations, not spreadsheets.
Everything below is the full methodology. You don’t need it to work your list — it’s here so the model is never a black box. Open what you’re curious about.
An Octave disqualifier caps the score at 40 and escalators at +4 — no quantitative signal rescues a disqualified account. These are hard rule-based overrides. In the June 2026 run, 39 of 600 shortlisted accounts (6.5%) were disqualified.
| Override | Fired | What it means |
|---|---|---|
| Sub-50 / founder-led | ~22 | No sales team to buy a sales platform; also catches defunct/acquired shells and B2C apps. |
| Non–North America HQ | ~14 | Outside our serving geography. Size never overrides geography. |
| Government / RFP motion | 2 | B2G procurement cycles don’t fit our motion. |
| Competitor | 1 | Operates in the sales-execution space we sell into. |
| SMB ecommerce | 1 | DTC/Shopify SMB — high churn, low ACV. |
Dealbreakers are graded, not blindly trusted: carVertical fired the Non-NA override and then closed won at $10,000. Counterexamples get logged and reviewed at the quarterly re-fit.
Customers are scored on a different question: where’s the room to grow? Adoption (.30, Product Engagement), seat whitespace (.25), sales-hiring timing (.20), ARR headroom (.15), utilization (.10). A SAVE-FIRST override fires when utilization <40% and product engagement is weak — retention before any expansion pitch.
The biggest expansion question isn’t seat count — it’s which teams we’ve sold into and which we haven’t. Validated on the paying book (2026-06-07): 185 addressable whitespace targets on $3.25M ARR — e.g. Northwind (120 unsold-team users), Globex ($125.7K ARR, 93 users).
Signals tested against the won/lost truth cohort and excluded. Not reintroduced without a new backtest.
| Signal | Lift | Status |
|---|---|---|
| Sales-role presence (binary) | 0.96× | flat — excluded |
| Hiring volume ≥3 postings | 0.91× | falsified — presence used instead |
| Bombora / third-party topic intent | 0.57× | inverse — excluded, stays excluded |
| SFDC history gradient | 1.02× | flat + circular — context-grade only |
| Octave continuous score (as ranker) | 1.03× | gate + context only in v6 |
| the fit score scores | — | weight zero, graded — computed, never weighted; feeds the the fit score accuracy loop |
Nothing changes the score except a pre-registered lift test. New signals enter at weight zero (shadow-scored, no effect) and only promote with ≥~1.5× won/lost lift. Weights re-fit quarterly on the grown outcome cohort; every re-fit ships as a versioned, QA-verified page. The current forward test: 173 strike accounts vs 173 matched controls, outcome = new opp or external meeting, reviewed 2026-09-04, pass = ≥2.0×.
