Confidential · © 2026 Space-O Technologies Inc. · Prepared exclusively for DMA. Please do not copy, save or share.

Section 2 · Qualification & Opportunity Scoring

2.2Learn scoring from DMA's real history

Designed for Phase 2
No.
2.2
DMA's reference
Decision Book p.11
Phase
Phase 2
Autonomy
L1 · Recommend
DMA's decision
Yes

What DMA asked for · Decision Book p.11

Use DMA's real won and lost deals to learn what actually predicts revenue, instead of guessing the weights.

Why it matters to DMA: It turns the score from an opinion into something backed by DMA's own outcomes.

In our words, from DMA's Product Decision Book v3, page 11.

How we do it

  • DMA's book schedules this for Phase 2, after the review of the 30,000 records and clean lost-reason data. Today's scores use reasoned starting weights, as the book allows.
  • The design: the backtest already built in the prototype re-scores past accounts and deals as they stood at the time, and measures how well the score ranked the ones that were won or revived. Learning adds a proposed set of weights that ranks DMA's real history better.
  • A learned version is never switched on automatically (L1: it recommends). It becomes a new scoring version with its backtest result next to it, and a person at DMA approves it, exactly like any other rule change.

Already in place

  • Measurement → Backtest: scores past reactivation waves and closed deals as they stood at the time, using only what the CRM held then, with deciles, top-10 precision and the winners that ranked low, and why.
  • Won and lost outcomes are already linked back to the accounts and waves that produced them.
  • Scoring weights are a versioned setting that only changes with a person's approval.

What the later phase adds

  • Fitting the weights to DMA's real won and lost history once the record review is done.
  • A proposed scoring version with its backtest comparison, for approval.
  • A periodic re-check as new outcomes arrive.

The flow

It plays on its own while it's on screen; hover or use the controls to pause or step through.

Won and lost dealsDMA's own historyBacktestas things stood thenLearn weightswhat predicted revenueProposed versionwith its comparisonA person approvesor keeps the old one
Step 1 of 4

DMA's history

Every closed deal and every past outreach with its outcome.

  • Nutshell
  • Rule in code
  • AI
  • Stored data
  • Person

Where the data goes

The same six stages on every page. Nothing reaches Nutshell except through the write gate.

  1. 1Past deals

    Source

    Won and lost deals, and past outreach outcomes.

  2. 2Sync

    Copies Nutshell changes into the bridge database

    Mirrors closed deals from Nutshell.

  3. 3Bridge database

    A copy of the CRM data, plus what the AI works out

    Backtest runs and proposed scoring versions.

  4. 4AI

    Claude models, only through DMA's own gateway

    No model call: the backtest re-uses the stored AI reading of the notes.

  5. 5Write gate

    The only way back into Nutshell: checked, approved when needed, sent once

    Not involved.

  6. 6Nutshell

    The system of record

    Unchanged until a new version is approved.

The tables behind it

Drawn from the POC's database catalogue: structure only, no data.

ai.backtest_runsPKiduuidcreated_attimestampsyntheticbooleanprimary_itemsintegerprimary_aucnumericsecondary_itemsintegersecondary_aucnumeric+ 11 more columnsai.scoring_configsPKiduuidrubrictextversionintegerweightsjsonbstatustextIDapproval_iduuid+ 9 more columns

PK primary keyFK reference the database enforces (solid line)ID reference kept by id (dashed line)

TableWhat it holdsColumns
ai.backtest_runsSaved backtest runs with their key numbers and full report.18
ai.scoring_configsScoring versions; a learned one would be a new version waiting for approval.15
How the tables connect (1)
ColumnPoints toKept by
ai.scoring_configs.approval_idctrl.approvalsThe application (by id)

Worked example

Synthetic demo data: every name, business and number is made up.

The prototype's backtest on its synthetic history, which the learning step would build on.

Backtest today (synthetic data)

Reactivation
24 of 176 accounts revived; the 10 highest-scored revived at 20%, against 14% overall
Closed deals
Not what this score is for: no better than chance on synthetic dealsReal outcomes are needed to judge or learn the weights.
Learning
Phase 2, on DMA's real history

In the running POC

There is no screen for this one yet: it is designed for Phase 2, as DMA's book decides. The parts listed under “Already in place” run in the POC today.

See it live

Opens the exact screen in the running POC, in a new tab.

OpenWhereWhat to try
Measurement → BacktestDMA AI dashboardHow the current score ranked past accounts and deals.

Status

Designed for Phase 2

Designed for Phase 2. DMA's book schedules this for Phase 2. The design is ready and builds on parts that already run in the POC.

2.2 Learn scoring from DMA's real history · DMA AI proof