Guide
Decision memo template
A decision memo records what was decided, on what evidence, against which objection, and by whom. It's the document someone will read in six months when the decision is questioned: by a board, an auditor, a new CEO, or a court. A good one takes an hour to write and saves weeks of reconstruction later.
Copy the template below. A filled-in example follows, then the rules that make a memo hold up.
The template
DECISION MEMO
──────────────────────────────────────────────
Title: [short name of the decision]
Date: [date] Decision owner: [name, role]
Decision needed by: [date] Reversible? [yes / partly / no]
1. DECISION
One sentence. What is being decided, by whom, by when.
2. RECOMMENDATION
The answer first. Then its conditions:
"Proceed, provided that ___ and ___."
3. OPTIONS CONSIDERED
A. [option]: one line on why / why not
B. [option]
C. Do nothing: what happens if we don't decide
4. WHAT WE KNOW (fact, with source)
- ___ (source: ___)
5. WHAT WE DON'T KNOW (missing inputs)
- ___ (when it would arrive / cost to find out)
6. ASSUMPTIONS (stated as assumptions)
- We assume ___ because ___
7. DERIVED FIGURES (calculated, and from what)
- ___ = ___ × ___
8. STRONGEST COUNTERARGUMENT
The best case against the recommendation, in its best form.
Our answer to it.
9. RISKS AND REVERSIBILITY
If we're wrong: what happens, how we'd know, can we undo it?
10. WHAT WOULD CHANGE THIS DECISION
Triggers to revisit: ___
11. DECIDED BY / DATE / AUTHORITY
Who approved it, and under what delegated authority.
──────────────────────────────────────────────
A filled-in example (fictional)
Title: Move mid-market plan from $49 to $59 per seat. Decision owner: CEO. Needed by: 15 August. Reversible? Partly (customers notice price changes).
1. Decision: whether to raise the mid-market seat price by $10 from 1 October.
2. Recommendation: proceed for new customers from 1 October. Hold existing customers until renewal, provided the August churn reading stays under 2.5%.
3. Options: A. raise for all (fastest revenue, highest churn risk). B. new customers only (recommended). C. do nothing (margin keeps eroding as support costs rise).
4. What we know: 1,240 mid-market seats (billing system, 31 July). Support cost per seat up 18% year on year (finance pack, Q2).
5. What we don't know: the price sensitivity of this segment. We have never raised prices. Competitor reaction.
6. Assumptions: we assume new-customer conversion falls by no more than 10%. This is a guess, based on one sales leader's view.
7. Derived: if conversion holds, new-customer ARR rises by about 20% per seat ($10 ÷ $49).
8. Counterargument: "We have no elasticity data, and assumption 6 is carrying the case." Answer: limiting the change to new customers caps the downside, and we will measure conversion weekly.
9. Risks: a conversion drop larger than 10% would show within six weeks. It can be reversed for new customers. Grandfathering limits the damage to existing ones.
10. Revisit if: conversion falls more than 10% for four consecutive weeks.
11. Decided by: CEO, under board-delegated pricing authority, 12 August.
Look at section 6. The memo admits the key assumption is a guess. That's not weakness. It's what makes the memo trustworthy, and it tells the team exactly what to measure.
The four-line rule
Most memos fail in the same way: they blend four different kinds of statement in the prose.
| Line | Example | Test |
|---|---|---|
| Known | "1,240 seats" | Can you point to the source? |
| Unknown | "Price sensitivity" | Is it named, not ignored? |
| Assumption | "Conversion falls ≤10%" | Is it labelled as an assumption? |
| Derived | "+20% ARR per seat" | Can you show the calculation? |
When "churn will stay at 3%" appears in a paragraph, a reader can't tell whether it's a measurement or a hope. Separating the four lines is most of the work.
Other rules that help
- Answer first. The reader's first question is what you want them to agree to. Don't make them hunt for it.
- Conditions stay attached. "Proceed, provided that…" is one recommendation. Don't drop the condition in the summary.
- One to two pages. Push detail to an appendix.
- Write the counterargument well. If the strongest objection reads like a straw man, the board will write the real one for you, in the meeting.
- Record authority. Who decided, and under what delegation. This is the line that matters most when a decision is reviewed later.
Writing the memo with ExecMind
ExecMind is an institutional executive desk built around this structure.
- Talk it through. Open a conversation and talk the options through the way you would with a colleague who has read the file. The desk matches your register.
- Convert to Review. When it has to stand, convert. The evidence ledger classifies known, unknown, assumption and derived on the file rather than in the prose, and an adversary comes on, so the counterargument is on the page before you walk in.
- Quant, if numbers matter. Put ARR, price, margin or runway on Quant as named expressions. If a binding input such as churn is missing, the desk proceeds symbolically and names the gap. It won't invent the number.
- Ask for the memo. Ask for a single, answer-first pass. Conditions on the recommendation stay on the recommendation. For a board, Pack writes the paper from the active file: conclusion, constraints, options, recommendation and counterargument.
Questions
What is a decision memo?
A short written record of a decision: the recommendation, options, evidence, assumptions, counterargument and who decided under what authority.
How long should a decision memo be?
One to two pages, answer first. Put supporting detail in an appendix.
What is the difference between a decision memo and a board paper?
A decision memo records one decision. A board paper presents a decision to a board for approval, usually with more context. ExecMind's Pack writes the board paper from the same file.
Can AI write a decision memo?
It can draft one from your file. ExecMind keeps known, unknown, assumed and derived on separate lines, and won't invent a binding number that is missing.
How do I start?
Sign in and open a conversation. A new account gets 150 free credits, no card, so you can see the desk work before you take a seat. A run that fails, or that you stop, isn't charged.