Methodology · public
How deal health is scored. No magic numbers.
Most AI CRMs hand you a deal score and ask you to trust it. We don't. Every Outcome Engine score is a weighted sum of six evidence categories you can audit. When the score moves, the signal that moved it is named.
Worked example
Acme · Q3 Renewal · $80k
72
/ 100
Reads at the “moderate” tier — the deal is real but missing a confirmed economic buyer and the next-meeting date.
Signal breakdown
The six signal categories
Where each point comes from.
Weights are starting defaults — your workspace admin can re-tune them as you learn which signals predict your wins.
Engagement quality
- Reply rate from the prospect side in the last 7/14/30 days
- Whether the champion attended the most recent meeting
- Whether email-open / link-click signal exists for shared assets
Read from · Activity timeline, email tracking, meeting attendance
Multi-thread coverage
- Number of distinct contacts engaged at the prospect company
- Whether economic buyer (decision-maker) is on a thread
- Whether technical / legal / finance roles are introduced when relevant
Read from · Contact + activity records
Stage velocity
- Days in current stage vs. your team's historical median
- Stage progression cadence over the deal lifetime
- Whether the next scheduled milestone has a confirmed date
Read from · Stage history + meeting calendar
Captured objections + asks
- Explicit pain captured in the latest rep update
- Stated objections (price, timing, fit, authority)
- Whether requested artifacts (proposal, case study, demo) were delivered
Read from · AI-parsed rep updates
Asset engagement
- Proposal opens + time on page
- Pricing or contract page repeat views
- Resources downloaded from share links
Read from · Proposal view tracking + share-link analytics
Forward-motion indicators
- Next-step scheduled within the expected interval for this stage
- Owner-created follow-ups completed vs. overdue
- Forecast category set by the rep (commit / upside / pipeline)
Read from · Tasks + stage-specific cadence rules
Backtesting commitment
We score predictions against actual outcomes — and ship the results.
Every closed deal in your workspace becomes a data point. We compare the deal-health score at each stage transition against the final close outcome. Your admin sees which signal categories actually predict wins for your team — and can adjust the weights accordingly.
If a signal stops predicting outcomes, we deprecate it. If a new signal emerges from your data, we surface it. The methodology is meant to get better over time, not stay frozen.
Calibration shipped
Monthly
Workspace admin override
Yes, per signal
Methodology version
v1.0 · May 2026
What we deliberately don't do
- We don't score from black-box LLM intuition. The model parses rep updates into structured fields; the scoring math is rule-based and inspectable.
- We don't hide which inputs moved a score. Every deal-detail view shows the signal breakdown. Same view a manager sees in 1:1 coaching.
- We don't auto-write to your CRM unattended. The rep confirms every structured field before save. The audit log records what was changed and why.
- We don't train models on your deal data. Anthropic's API terms prohibit it by default and we enforce that contractually.
Defend the forecast with the receipts.
See the live methodology in your own workspace. Free plan available, no credit card.