# Decision Memo

Appendix C Kit 6, expanded for a working memo. Options first, recommendation last. The criteria table has to be fair to the option you did not pick; that is the line a skeptic checks first. The flip rule is the one you defend in the room.

## The Memo

```
Context: [The decision and why it is due now. Three sentences max.]

Options (three or more):
Option          | Cost this year | Cost to maintain | Risk          | Notes
[Option one]    | [Money, time, people] | [Per month or quarter] | [What breaks] | [One honest line]
[Option two]    | [...]          | [...]            | [...]         | [...]
[Option three]  | [...]          | [...]            | [...]         | [...]

Board findings:
Strongest objection: [What it was.]
Changed after it: [What you changed.]
Kept anyway: [What you kept and why.]

My call: [One sentence.] I would change it if: [The one condition that flips the call.]

Revisit date: [Date.]
```

## The Weighted Score Table

Optional, for when the options need a defensible score. Weights are yours; say them out loud to a skeptic before you fill the scores.

```
Criterion                | Weight | Option one | Option two | Option three
[Cost this year]         | [1-5]  | [1-5]      | [1-5]      | [1-5]
[Cost to maintain]       | [1-5]  | [...]      | [...]      | [...]
[Risk]                   | [1-5]  | [...]      | [...]      | [...]
[Speed to value]         | [1-5]  | [...]      | [...]      | [...]
[What the team can absorb] | [1-5] | [...]     | [...]      | [...]
Weighted total           |        | [Sum of weight times score] | [...] | [...]
```

Then the honesty pass, from prompt card P18: read the table as someone who wants the option you did not pick. Where does the weighting favor a conclusion you already hold?

## A Filled Example (all names invented)

Context: our quarterly report is hand-built from three exports and takes two evenings. It lands the same shape every month and the numbers are right, but the process does not survive someone else running it. I need a decision on how we produce it next quarter.

| Option | Cost this year | Cost to maintain | Risk | Notes |
| --- | --- | --- | --- | --- |
| Keep the manual build | Two evenings per month, mine | Same, forever | Errors when I am out | Honest option, no new exposure |
| Agent builds the draft, I review | One afternoon of setup | Ten minutes per month | Draft needs checking every run | The factory from Chapter 9 |
| Buy the reporting add-on | License, one line item | Vendor renewal yearly | Data lives in the vendor | Fastest, but the tool holds our numbers |

Board findings:

- Strongest objection: a draft that needs checking is still two jobs, not one.
- Changed after it: the factory draft ships with the two numbers I check by hand flagged, not buried.
- Kept anyway: review before ship. The checking is the point, and it stays mine.

My call: agent-built draft with my review, piloting for one quarter. I would change it if: the draft needs more than ten minutes of edits twice in a row.

Revisit date: end of next quarter, same week as the QBR.