Account Scoring Model: How to Rank In-Market Accounts [2026]
Building an in-market account list answers who might buy. Scoring answers the question your reps actually ask every morning: who first?
Most scoring models fail not because the math is wrong but because nobody trusts them — the weights were guessed in a workshop, the score never matches what reps see, and within a quarter everyone's back to gut feel. Here's a model built the other way: start from your closed-won data, keep the dimensions legible, and wire the score to routing so it changes behavior instead of decorating a dashboard.
The Three Dimensions
Every workable account score combines three things:
| Dimension | Question it answers | Example signals |
|---|---|---|
| Fit | Could they buy? | Industry, employee band, tech stack, funding stage, geography |
| Intent | Are they researching? | Topic surge, competitor comparisons, review-site activity, category searches |
| Engagement | Are they researching you? | Pricing-page visits, return sessions, demo views, email/LinkedIn replies |
Keep them separate. A blended single number hides the difference between "perfect fit, zero activity" (nurture) and "mediocre fit, on your pricing page daily" (call now) — and those demand opposite plays.
Step 1: Calibrate Weights on Closed-Won, Not Opinion
Pull your last 12 months of closed-won and closed-lost accounts. For each, reconstruct what was observable before the first meeting: firmographics, signals, site behavior. Then ask which attributes actually separated winners from losers.
Every team that does this honestly finds surprises. Common ones:
- First-party engagement outpredicts third-party surge by a wide margin — a pricing-page visit is worth more than a month of topic surge (our buying-signal hierarchy analysis ranks the common signals by closed-won correlation)
- One or two firmographic traits you thought mattered don't
- Negative signals (wrong stack, recent competitor contract, hiring freeze) predict losses better than positive signals predict wins
Recalibrate quarterly. Around three-quarters of B2B teams will run some form of AI-assisted scoring by end of 2026 — but AI calibration on top of unexamined assumptions just automates the guessing.
Step 2: Score Each Dimension 0-100
A simple, legible structure beats a clever one:
Fit (gate + score). Hard disqualifiers first — wrong size, wrong region, incompatible stack → score 0, exit. Survivors get scored on weighted firmographic/technographic match.
Intent (recency-weighted). Score signals, then decay them: a signal this week at full value, halved next week, gone by week four. Intent has a half-life; a score that ignores decay ranks last month's shoppers above this week's.
Engagement (behavior-tiered). Not all touches are equal. A blog visit is a point; a pricing page is ten; a second pricing visit within a week is twenty; an identified decision-maker doing it is fifty. Person-level resolution matters here — knowing who is on the page is the difference between account warmth and an actual buyer.
Step 3: Convert Scores to Tiers
Reps don't act on a 73. They act on tiers:
| Tier | Threshold (typical) | Play | SLA |
|---|---|---|---|
| A — Act now | High fit + high engagement | Direct outreach to identified people, personalized | 24 hours |
| B — Working | High fit + intent, low engagement | Warm outbound referencing category research | 72 hours |
| C — Watch | Fit, weak/no signal | Automated nurture, monitor for signal | — |
| D — Disqualified | Failed fit gate | None. Genuinely none. | — |
Set thresholds so Tier A matches your team's actual capacity. A tier system that flags 400 "act now" accounts for six reps is a random number generator with extra steps.
Step 4: Wire the Score to Routing
This is where scoring lives or dies. A score that only sorts a dashboard changes nothing. The score should do things:
- Tier A entry → account routed to a named rep with the triggering signal attached, SLA timer running (routing rules here)
- Tier transitions → notifications, not reports ("Acme moved B→A: 3rd pricing visit this week")
- SLA breach → escalation or re-route
The signal-to-meeting playbook covers the outreach side of the handoff.
The Failure Modes
- Opinion-weighted models. If the weights came from a meeting instead of closed-won data, the model encodes the loudest person's intuition.
- No decay. Scores that only go up produce a leaderboard of accounts that were hot in Q1.
- Score without routing. If a tier change doesn't move work to a person with a deadline, the model is decorative.
- Precision theater. Seventeen weighted sub-factors nobody can explain lose to four factors every rep understands. Trust drives adoption; adoption drives revenue.
How MarketBetter Handles Scoring
MarketBetter scores accounts and people continuously — fit gate, blended intent with built-in decay, person-level engagement — and routes tier changes straight into SDR queues with the evidence attached and outreach drafted for human review. Calibration runs against your actual pipeline outcomes, not a static rubric.
Book a demo to see your accounts scored on live signals.
FAQ
Account scoring vs lead scoring — what's the difference? Lead scoring ranks individual form-fills; account scoring ranks companies using signals from everyone in the buying committee, including people who never converted. In committee-driven B2B sales, account scoring is the one that matches how deals actually happen.
How many factors should the model use? As few as survive the closed-won analysis — usually 4-7. Every factor you add that reps can't verify against reality costs trust.
How often should scores update? Continuously, or daily at minimum. Weekly batch scoring means your fastest-moving buyers spend their hottest days invisible.