Section 21 · Final Revenue Multipliers & Control Plane
21.9Attribution confidence & incrementality control
Built in the POC- No.
- 21.9
- DMA's reference
- Decision Book p.46
- Phase
- Phase 1
- Autonomy
- L1 · Recommend
- DMA's decision
- Yes
◆ Amended in v3 of the book.
What DMA asked for · Decision Book p.46
Separate revenue the AI created from revenue it only touched, and keep control groups for the major reactivation experiments.
Why it matters to DMA: It stops the system claiming credit for revenue that would have happened anyway, and keeps management decisions honest.
In our words, from DMA's Product Decision Book v3, page 46.
How we do it
- Every reactivation wave keeps a matched holdout that isn't contacted. Replies, meetings, new pipeline and won deals are counted for both groups inside the wave's window, from the moment each account was first contacted.
- The lift is the difference between the two, in percentage points, with a 95% range and a plain verdict ("meaningful", "no difference yet"), so a small or noisy result isn't mistaken for success.
- How each outcome counts is a versioned rule (what counts as a reply, a meeting, pipeline, a win), shown next to every number with "How this was counted".
- Simulated history is labelled everywhere; the backtest shows how well the score ranked past outcomes.
The flow
It plays on its own while it's on screen; hover or use the controls to pause or step through.
Two groups
Emailed and held out, matched by tier and segment.
- Screen
- Person
- Rule in code
- Routing
Where the data goes
The same six stages on every page. Nothing reaches Nutshell except through the write gate.
- 1Waves and outcomes
Source
Sends, replies, meetings, deals.
- 2Sync
Copies Nutshell changes into the bridge database
Brings meetings and won deals from Nutshell.
- 3Bridge database
A copy of the CRM data, plus what the AI works out
Experiments, holdouts, outcomes, attribution rules.
- 4AI
Claude models, only through DMA's own gateway
Not involved: counting is code.
- 5Write gate
The only way back into Nutshell: checked, approved when needed, sent once
Not involved.
- 6Nutshell
The system of record
The source of won revenue.
The tables behind it
Drawn from the POC's database catalogue: structure only, no data.
PK primary keyFK reference the database enforces (solid line)ID reference kept by id (dashed line)
| Table | What it holds | Columns |
|---|---|---|
| ctrl.attribution_rules | How each outcome counts, versioned. | 10 |
| ctrl.decision_outcomes | Each outcome, its group and the wave. | 13 |
How the tables connect (2)
| Column | Points to | Kept by |
|---|---|---|
| ctrl.decision_outcomes.decision_id | ctrl.decision_log | The application (by id) |
| ctrl.decision_outcomes.experiment_id | ctrl.experiments | The database (foreign key) |
Worked example
Synthetic demo data: every name, business and number is made up.
Measurement
What you see
1.Each wave against its holdout
Replies, meetings, pipeline, won, lift and verdict.
What happens in the system
- Rule engine
attribution v1The versioned rules count both groups the same way.
How the lift is worked out (the prototype's rules, version 1).
Attribution rules
- Unit
- The account
- Lift
- Emailed rate minus holdout rate, in percentage points, with a 95% range
- Won revenue
- The value of deals won inside the window
- Counting starts
- When the account is first contacted in the wave
In the running POC
Measurement: waves against their holdouts.
- 1Simulated history, labelled.
- 2A wave: emailed vs held out, lift and verdict.
Dashboard · Measurement. Captured from the running POC on September 25, 2026; synthetic data.
See it live
Opens the exact screen in the running POC, in a new tab.
| Open | Where | What to try |
|---|---|---|
| Measurement | DMA AI dashboard | Open any number's "How this was counted". |
Status
Built in the POC. This capability runs in the POC today, on synthetic data, end to end.