What is a decision memo?

A short document that leaves a record of a decision: the question it answers, the options considered, the criteria, the evidence and how much each piece counts, the recommendation with its rationale, the assumptions and risks, and how certain you are. It brings together two practices that already existed: the decision analysis Ronald Howard formalized at Stanford in 1964, and the written memo Amazon adopted in 2004 in place of slides.

Method

How to write a decision memo

Q: How long should a decision memo be?

As a sensible rule, one or two pages, with the recommendation up top. Amazon uses a six-page narrative format for big decisions; for a specific call, one or two pages usually do the job.

Q: How is it different from a report or a slide deck?

A report describes a situation, and a slide deck can hide gaps between one bullet and the next. A memo forces you to put the whole reasoning in prose: what you recommend, on what evidence, on what assumptions, and with how much certainty. It can be read, discussed, and revisited later.

By the Verdika team · Jul 10, 2026 · 6 min read

In short

A decision memo is, at heart, writing down in order what any sensible person would want to know before deciding: what's being decided, among which options, on what criteria, what the evidence says and how much it's worth, what you recommend, on what assumptions you're leaning, and how sure you are. Seven pieces. And one rule that changes everything: the recommendation goes first.

We didn't invent this format, and it's worth saying so. It comes from two different places that ended up in the same spot. On one side, decision analysis, the discipline Ronald Howard founded at Stanford in the mid-sixties, which introduced the habit of treating uncertainty with probabilities instead of hiding it¹. On the other, a very practical call from Amazon: in 2004 they banned slide decks from important meetings and replaced them with a written memo, because a list of bullets lets holes in the reasoning slip by unnoticed². A good memo takes the best of both: the rigor of the first and the honesty of the second's continuous prose.

The seven pieces

#	Piece	What it answers
1	The decision	What is being decided, exactly? Better phrased as a question.
2	Options	Between which real paths are you choosing? Ones that can't both happen.
3	Criteria	What do you want to achieve, and what can't be traded away?
4	Weighted evidence	What data is there, and how much do you trust each piece?
5	Recommendation	The preferred option with its score, and the order of the rest.
6	Assumptions and risks	What it rests on, and what would change it.
7	Certainty	How much there is, without dressing it up.

Verdika's own synthesis of decision analysis¹ and the narrative-memo practice².

Start at the end

It's hard to do, because you want to show your work. But whoever reads the memo needs to know where you're headed before they'll follow how you got there. So the first line is the recommendation, plainly: "We recommend keeping the team and consolidating product-market fit. Score 0.82, confidence 78%." The reasons come after, and that's what the rest of the document is for. As a side benefit, a memo that opens with its conclusion gets quoted more accurately (by a hurried person or by an AI search engine) because the sentence that matters sits at the top, on its own.

An example, so you can see it

The decision: do we hire three people now, or stretch the cash six more months?
Recommendation: stretch the cash (0.82). Hiring now comes third (0.31).
Why: with retention below where we want it, adding people speeds up spending without moving the one metric that gates the next round.
Assumptions: that retention won't improve on its own this quarter. Risk: losing a candidate worth keeping. Certainty: 78%, moderate, because the retention evidence is still partial.

A quick test: if the memo can't say how certain it is, it's still just an opinion in nice typography, not a memo.

The three that ruin it

Almost every weak memo fails the same way. The first is confidence with nothing behind it: sounding sure about data that doesn't warrant it. That's the fastest way to lose credibility, which is why the certainty you state has to be tied to the quality of the evidence, not to your mood. The second is options that don't actually compete: if two paths can both be taken, they aren't options, they're tasks. And the third, the most human one, is the assumptions you'd rather not write down. What you leave out is what gets asked in the room, and the improvised answer rarely convinces.

How Verdika handles it

You tell it the problem the way you'd tell a partner, without formatting a thing. Verdika does the heavy lifting: it frames the decision, looks at the evidence you gave it and decides how much to trust each piece, ranks the options with their score and their reason, and puts names to the assumptions, the risks, and the certainty. What you get back is the finished memo, ready to hand over. In minutes, not an afternoon.

Bring the decision you keep putting off.

Tell it in your own words. The structure is on us.

Analyze my decision

Frequently asked

How long should a decision memo be?

One or two pages, recommendation first. Amazon uses six pages for big decisions²; for a specific call, less is plenty.

How is it different from a report or a slide deck?

A report tells you what's happening; a deck can hide gaps between bullets. A memo makes you write the full reasoning, with its evidence, its assumptions, and its degree of certainty. You can read it, argue with it, and come back to it months later.

References

Ronald A. Howard, founder of decision analysis — Stanford, Management Science & Engineering: msande.stanford.edu/people/ronald-howard. Biographical profile at INFORMS: informs.org.
Amazon's narrative memo (2004), told by two of its executives in Working Backwards, Colin Bryar and Bill Carr: book page.

You might also like: Hire now or extend your runway · Is ChatGPT good for decisions?