1. Courses
  2. Product Management
  3. Communication & Leadership
  4. Stop 1

35 minAdvanced

Writing for decisions

Most of a PM's influence travels in documents. This lesson shows you how to write the one-pagers and decision memos that get a clear yes or no from a busy leader in one meeting, how to match the effort to how reversible the decision is, and how to find the ways a plan will fail before you commit to it.

By the end you can

  • Pick the right document for a decision: a one-pager, a decision memo or a PR/FAQ
  • Write a decision memo with the ask, the recommendation and fair options at the top
  • Tell one-way doors from two-way doors, and turn a one-way door into a two-way one
  • Run a 30-minute pre-mortem and put what it finds into the plan
  • Record a decision so nobody has to argue later about what was agreed

The everyday picture

A family is deciding whether to keep renting their flat or buy one. The eldest daughter has done the homework, so before Sunday dinner she sends everyone a one-page note. The first line says what they are deciding and by when: "Do we buy the flat on Lake Road before the offer closes on the 30th?" Then her recommendation, the three options (keep renting, buy Lake Road, buy something cheaper further out), the monthly cost of each, and what could go wrong. She adds one honest line: "Renting we can undo with a month's notice. Buying we can't undo for years."

At dinner nobody asks her to explain the numbers again. They argue about the thing that matters: will they still want to live in this city in ten years? The decision takes 40 minutes instead of three weekends, and when someone asks in June "why did we pick this one?", the note answers.

That note is a decision document. It states the question, gives a recommendation, treats the options fairly, says how hard the choice is to reverse, and leaves a record.

Why write, and which document to write

When you need a decision from someone senior, the easiest thing is to book a meeting and talk through some slides. Writing takes longer and works better, for four reasons. Writing forces you to finish your thinking, because gaps that hide between bullet points show up in sentences. Readers go at their own speed and can reread the hard part. People who miss the meeting can still follow. And the document becomes the record of what was decided and why.

Amazon is the best-known example. In his 2017 letter to shareholders, Jeff Bezos wrote that the company does not use slide presentations in senior meetings: "Instead, we write narratively structured six-page memos." Each meeting starts with everyone reading the memo in silence, "in a kind of study hall". He added that a great memo is "written and re-written", shared with colleagues who improve it, set aside for a couple of days and edited again, and that a great one "probably should take a week or more".

You will not need six pages for most decisions. A comes in three common sizes:

DocumentUse it whenLengthExample at Mangrove
One-pagerThe decision is small or easy to undo, and one or two people decideOne page, 300 to 500 wordsPause store counts during the two festival peak weeks?
Decision memoMoney, many people or a hard-to-undo choice is involved, or people disagreeTwo to six pages, plus an appendixCount the 300 fastest-moving items every day in all 320 stores?
PR/FAQThe bet creates something new that customers or staff will feel, and the benefit itself is in doubtA one-page press release and up to five pages of FAQLaunch a 15-minute delivery promise in one city?

All three share one rule: the ask comes first. Who must decide, what exactly, and by when. A reader who stops after the first paragraph should still know what you want from them. If you cannot write the ask in two sentences, you are not ready to send the document.

Quick check

Ken, the engineering manager, and Imran need to choose whether the Store app's count list refreshes every night or every four hours. Either is easy to change later. Which document fits?

The decision memo, part by part

A decision memo has a fixed order. Each part answers a question the decider will ask, in the order they will ask it.

  1. Title as a question. "Should we count our 300 fastest-moving items every day in all 320 stores?" A question title tells the reader this is a decision, and the answer is yes or no.
  2. The ask. Who decides, what you need, and the date. Name the door type (see the next section) in the same block.
  3. The recommendation in one paragraph. The problem in numbers, what you propose, what it costs, what it is worth, and what you are not asking for yet.
  4. What we know. Only the facts the decision depends on, each with its source. Background the reader already has goes in the appendix or nowhere.
  5. Options, judged on the same criteria. Two to four real options, always including "keep doing what we do now". Put them in a table with the same rows for each: effect on the goal, cost, risk, time to first result, and how reversible it is.
  6. Why we recommend this option. The trade-off you are choosing, in plain words, including what you give up.
  7. Risks and the pre-mortem. What could make this fail and what the plan does about each. More on this below.
  8. How we will know, and when we look again. The metric, the target, the guardrails and the date of the review.
  9. Appendix. Method, data, detailed costs, the full option analysis. The decider may never open it, but contributors will check it.

A few writing rules make the difference between a memo that decides and one that starts a debate about the memo:

  • Make every option a real one. If option A is "do nothing" and option C is "spend $5 million", the reader will feel steered, and they will be right. Describe the option you reject the way its strongest supporter would.
  • Give every number a source and a range. "About 5%, range 4 to 6%, from a 5-day trial in one store" invites the right question. "5%" invites the question "says who?".
  • Say what you don't know. An unknown that you name is a risk you are managing. An unknown the reader finds is a hole in your credibility.
  • Write "we recommend". Phrases like "it could be considered" hide who is accountable. The memo's author owns the recommendation.
  • Cut the story of your work. The reader needs the conclusion and the evidence, not the order in which you did the analysis.

Quick check

Imran's draft options table shows A "keep monthly counts" and B "daily counts", and leaves out the shelf-camera proposal a vendor pitched to the managing director last month. What should he do?

Draft the ask question, options Driver Contributors comment in doc Experts rewrite Pre-mortem risks into plan 6 to 10 people Pre-read sent 24 h before Approver Meeting read 10 min first Approver missing evidence decided Decision logged review date set Driver
The life of a decision memo. The driver drafts, experts comment and the draft loops until it is clear. A pre-mortem adds the risks. The approver gets it a day ahead, the meeting starts with silent reading, and it ends either in a logged decision with a review date or with the named evidence that is missing.

One-way doors and two-way doors

Not every decision deserves a memo and a pre-mortem. The most useful test for how much process a decision needs is how hard it is to undo. Bezos described it in his 2015 letter to shareholders: "Some decisions are consequential and irreversible or nearly irreversible – one-way doors – and these decisions must be made methodically, carefully, slowly, with great deliberation and consultation." But, he went on, "most decisions aren't like that – they are changeable, reversible – they're two-way doors", and those should be made quickly by a small group or one person with good judgement. In his 2016 letter he added that "most decisions should probably be made with somewhere around 70% of the information you wish you had."

Teams make two opposite mistakes. They treat two-way doors as one-way doors, so a button label goes through three review meetings and the team slows to a crawl. Or they treat one-way doors as two-way doors: they sign a three-year contract, or change how 1,900 store staff work, on the strength of one meeting.

A gets a light process: one owner decides, a one-pager or a log entry records it, and it takes days. A gets the full process: a memo, a pre-mortem, a named approver, and weeks.

The best move is often to turn a one-way door into a two-way door before you walk through it:

  • Pilot first. Twenty stores for four weeks before 320 stores for good.
  • Stage the purchase. Buy 20 devices for the pilot, and buy the other 300 only if the pilot shows you need them.
  • Negotiate an exit. A break clause after six months, or a month-to-month contract at a slightly higher price.
  • Hide it behind a switch. A lets you turn a change off in minutes (Launching).
  • Keep the old way running for a while. Keep the paper count sheet available for one month after the app count goes live.

Reversibility is rarely all or nothing. A change to a store routine can be undone, but every reversal costs training time and trust. Say which parts of your decision are one-way and which are two-way. That usually lets the decider approve the two-way part today and keep the one-way part for later, with better evidence.

Quick check

A vendor offers shelf cameras for Mangrove's 40 busiest stores on a three-year contract with no exit. What is the best first move?

A decision Easy to undo, cheaply, soon? yes Two-way door owner decides one-pager, days no Can we make it two-way? yes: pilot, stage it, break clause no One-way door memo, pre-mortem approver, weeks
Match the process to the door. Easy-to-undo decisions are made fast by their owner. Hard-to-undo ones first go through the question "can we make it two-way?", and most can be, with a pilot, a staged purchase or an exit clause. Only what remains gets the full memo, pre-mortem and approver.

Pre-mortems: borrow hindsight before you need it

After a project fails, everyone can explain why. The moves that explanation to the start, when you can still act on it. Gary Klein described it in the Harvard Business Review in 2007. The team imagines that the project has already failed, then each person writes down the reasons. Klein cites 1989 research by Deborah Mitchell, Jay Russo and Nancy Pennington, which found that imagining an event has already happened (they called it "prospective hindsight") increased people's ability to identify reasons for future outcomes by 30%.

A pre-mortem works for a social reason as well. In a normal planning meeting, raising doubts can look disloyal, especially to a senior sponsor's idea. In a pre-mortem, finding reasons for failure is the task, so the quiet operations lead who has seen three similar projects die can say what she knows.

How to run one in 30 minutes:

  1. Invite 6 to 10 people who know different parts of the work: the team, a frontline user such as a store manager, someone from finance or operations, and at least one sceptic.
  2. Set the scene in one sentence. "It is the end of March. The daily counts were a failure. Store managers want them gone and cut orders did not move. Why?"
  3. Write silently for three minutes. Each person lists every reason they can think of. Silence stops the first speaker from setting the agenda.
  4. Go round the room, one reason per person per turn, until the lists are empty. Write them where everyone can see.
  5. Vote on the reasons that are most likely and most damaging. Each person gets three votes.
  6. Change the plan for the top three to five. Each gets a mitigation, an owner, or an early warning sign you will watch. Put the result in the memo's risk section.

The output is different from a normal risk list. It tends to be specific ("stores skip counts in festival weeks") and human ("store managers see it as head-office extra work"), where risk lists tend to be generic ("adoption risk"). Look again at the list at each review, because some reasons turn into early warning signs you can measure.

Quick check

Why does a pre-mortem start with three minutes of silent writing instead of open discussion?

The PR/FAQ as a decision tool, and the one-pager

You met the PR/FAQ in Writing specs people read, as the step before a spec. Use it for a decision when the open question is whether customers or staff will value the result at all. The press release forces a one-line benefit ("Fewer 'Now' orders arrive with something missing"), and the internal FAQ carries the decision: what it costs, what must be true, which parts are one-way doors, what you are not doing, and what you need from the reader today. When the benefit is already accepted and the debate is about cost, risk or timing, a decision memo is the better fit, because its options table compares choices side by side.

For two-way doors, write a one-pager. It has six short parts: the decision needed (with a date), the recommendation, why, the options you considered in one line each, the risk and the door type, and the next step. Send it, give people a deadline to object ("if I hear nothing by Thursday, we go ahead"), and log the result. Most of a PM's decisions should take this path, so that the full memo stays rare and gets read.

Do it

Write a decision memo and run a pre-mortem for one real decision you need from a senior person, using Imran's memo as a model.

  1. Write the question and the ask first

    Before any analysis, write the title as a yes-or-no question, then the ask in two sentences: who decides, what exactly, by when. If the ask has three parts, check whether it is really three decisions.

  2. Name the door for each part of the ask

    List what you are asking for and mark each part two-way or one-way. For each one-way part, try the four moves: pilot, stage, exit clause, switch. Move whatever stays one-way into a later, separate decision if you can.

  3. Build the options table on shared criteria

    Pick four or five rows that the decider cares about (effect on the goal, cost, risk, time to first result, reversibility) and fill them for every option, including the current way of working. Ask a supporter of each option you reject to check its column.

  4. Write the top half of the memo

    Title, ask, recommendation, what you know, options. Every number has a source; every estimate has a range.

    Markdown
    # Should we count our 300 fastest-moving items every day in all 320 stores?
Decision memo | Driver: Imran (PM, Store app) | Approver: managing director
Decide by: 24 Oct | Contributors: head of store ops, head of finance, Ken, Tomás

## The ask
1. Approve a 4-week pilot of daily fast-mover counts in 20 stores from 3 Nov.
   Door: two-way (stop any day; cost under $1,000 of staff time).
2. Pre-approve rollout to all 320 stores if the pilot passes the gates below.
   Door: two-way, but costly to reverse (training, store routines).
Not asked today: 320 extra scanners ($192k, one-way). Only if the pilot shows queues.

## Recommendation
About 7% of app orders (about 4,200 a day) lose an item at picking. 55% of the
missing items are among 300 fast movers, which we count once a month. Counting
those 300 each afternoon (about 25 min per store) and the full range quarterly
instead of monthly should cut missing items to about 5%. Base case: about $330k
a year of margin kept, for about $69k a year of extra staff time and a $20k build.

## What we know (sources in the appendix)
- Stock accuracy 88% (weekly 2,000-line sample); cut-order rate 7% (8-week average)
- Fast movers drift within days; a monthly count finds an error 2 weeks late on average
- 300 lines take about 25 min, 2-4 pm (5-day trial in Ama's store)

## Options
| Criterion          | A. Monthly counts | B. Daily fast movers (rec.) | C. Shelf cameras, 40 stores |
| Cut-order rate     | stays about 7%    | about 5% (range 4-6%)       | unknown, no trial data      |
| Cost per year      | $0                | $69k staff + $20k once      | about $400k (3-yr, no exit) |
| Time to result     | -                 | 4 weeks                     | about 6 months (install)    |
| Reversibility      | -                 | two-way                     | one-way                     |

  5. Run the pre-mortem and write the bottom half

    Run the 30-minute pre-mortem from this lesson, then write the risks, the gates, the review date and the appendix list. Show which reasons changed the plan.

    Markdown
    ## Pre-mortem: "It is March and the daily counts failed. Why?" (8 people, 30 min)
| Reason (votes)                                  | What the plan now does                        |
| Stores skipped counts in festival weeks (6)      | Counts pause in the 2 peak weeks; completion  |
|                                                 | tracked per store, shown to area managers     |
| Staff typed a guess instead of counting (5)      | Random recount of 10 lines per store a week   |
| Counting fought receiving for scanners (4)       | Counts 2-4 pm; scanner purchase is a separate |
|                                                 | decision, made only on queue data             |
| Cuts came from picking errors, not stock (3)     | Pilot tags every cut by cause                 |
| Managers saw it as head-office extra work (3)    | Full count becomes quarterly; managers see    |
|                                                 | their own store's cut-order rate daily        |

## Gates for rollout (pilot stores vs 20 matched stores, 4 weeks)
1. Cut-order rate falls at least 1 point more than in the matched stores
2. Counts completed on at least 85% of store-days
3. Recount accuracy at least 95%
Guardrail: median receiving time per delivery stays at or under 20 min (18 today)

## Value and cost (detail in appendix B)
Each point of cut-order rate = 600 orders a day x $3 lost sale = about $657k sales,
about $164k margin a year (25% gross margin). Low 1 pt / base 2 pts / high 3 pts.
Net of extra staff time: about $95k / $260k / $424k a year.

## Review: pilot readout 4 Dec; 90-day review of rollout 15 Mar
## Appendix: A. method and data  B. cost model  C. pilot store list  D. option C detail

  6. Test it before you send it

    Give the first paragraph and the options table to someone outside the project for two minutes, then ask them what you want and why. Ask the person most likely to oppose you to read the whole draft and mark the weakest sentence. Then cut about a third of the words.

  7. Record the decision the same day

    Add a row to the product's decision log, with the conditions and the review date, and send a three-line note to everyone who contributed. Use a one-pager for the small decisions that follow.

    Markdown
    ## Decision log: Store app
| Date   | Decision                                   | Decider | Door    | Conditions / review          |
| 24 Oct | Pilot daily fast-mover counts, 20 stores,  | MD      | two-way | Rollout pre-approved if 3    |
|        | 3 Nov to 30 Nov                            |         |         | gates pass; review 4 Dec     |
| 27 Oct | Counts pause in the 2 festival peak weeks  | Head of | two-way | One-pager; look again after  |
|        |                                            | st. ops |         | the first festival           |

# One-pager: Pause daily counts during festival peak weeks?
Decision needed: by 30 Oct, from the head of store operations
Recommendation: pause counts for the 2 peak weeks; restart the day after.
Why: store staff are needed on the floor; drift in 2 weeks is small (pilot data).
Options considered: keep counting (staff stretched); count 100 items only (complex).
Risk / door: two-way. Accuracy may dip; we watch the cut rate daily.
Next step: if no objection by 30 Oct, the count list is switched off for 14 days.

Bring it to your team

Ask these before any decision document leaves your hands.

  • If the decider reads only the first paragraph, does she know what we want, from whom and by when?
  • Is "keep doing what we do now" in the options table, and would its supporters say we described it fairly?
  • Which part of this ask is a one-way door, and can we make it two-way or move it to a later decision?
  • Does every number have a source and a range? Is our value in sales or in margin, and do we say which?
  • Which pre-mortem reasons changed the plan, and who owns each one?
  • Where will this decision be recorded, and when do we look at it again?

Red flags

  • The ask appears after page one, or the memo ends with "next steps to be discussed".
  • Only two options, one of them obviously bad.
  • Costs described as "minimal" or "within existing budget" with no number.
  • A three-year contract, a large purchase or a 320-store change approved in the same meeting it was first proposed.
  • Three months later, two people remember the decision differently and there is no log entry.

Practice

Turn a buried email into a one-pager

The head of partner sales received this email from a PM and replied, "What do you need from me?"

"Hi, following our discussions last month we looked at how partner shops pay. As you know about 4,000 partner orders a week go out and many shops pay late. We did some interviews and the main reason seems to be that shops wait for their sales rep to collect cash on the next visit. We also looked at wallets. Lagoon Digital says adding wallet payment to the Partner app is about 6 weeks of work and the wallet provider charges a fee per payment, maybe around 1%. Finance is a bit worried about the fee. We could try it with some shops first. Let me know your thoughts."

Rewrite it as a one-pager with all six parts. Invent nothing that changes the facts; where something is missing, say so in the document.

  • The first two lines state who decides, what exactly, and by when
  • The recommendation is one clear sentence with a scope, such as a pilot size and length
  • At least two options are named, including keeping cash collection by reps
  • The fee concern is stated as a number to confirm, with an owner, instead of being hidden
  • The door type is named, and the pilot is what makes it two-way
  • Missing facts (how late payments are today, what late payment costs) are listed as open questions
Show a model answer

Decision needed: by 14 Nov, from the head of partner sales: may we pilot wallet payments in the Partner app with 100 shops for six weeks?

Recommendation: yes. Ask Lagoon Digital to add wallet payment (about 6 weeks of work), then pilot it with 100 shops that often pay late, alongside rep collection.

Why: interviews suggest shops pay late mainly because they wait for the rep's next visit to hand over cash. Paying in the app removes the wait.

Options considered: (A) keep rep cash collection only; (B) wallet payment for all shops at once; (C) a 100-shop pilot (recommended).

Risk and door: the wallet fee is "maybe around 1%"; finance to confirm the real rate and who pays it by 10 Nov. Building the feature is a two-way door for shops (they can keep paying cash) and a modest sunk cost of about 6 weeks of agency work.

Open questions: how many days late do shops pay today, and what does that cost us? What share of shops already use a wallet? Next step: if you agree, we brief Lagoon Digital on 15 Nov.

Sort five decisions by door, and make one two-way

For each decision, say whether it is a one-way or a two-way door, who should decide, and what document it needs. Then pick the most one-way decision and describe how to make most of it two-way.

(1) Change the label on the Store app's "Close GRN" button. (2) Sign a five-year lease for a new distribution centre. (3) Move all 320 stores from paper count sheets to the app in one weekend. (4) Raise the Partner app's minimum order from $50 to $80. (5) Replace the customer app's payment provider.

  • The button label is two-way, decided by the team, with a log entry at most
  • The lease is one-way and goes to a senior approver with a full memo
  • The weekend cutover is recognised as mostly one-way as stated, with a staged alternative proposed
  • The minimum order change is treated as reversible but with a real cost to shops' trust, so it gets a one-pager and notice to shops
  • The payment provider change is one-way-ish, with parallel running or a staged migration named as the way to make it safer
Show a model answer

(1) Two-way. The designer and PM decide; log it. (2) One-way. The managing director or board decides on a full memo with a pre-mortem; ask for a break clause anyway. (3) As written, close to one-way: if the app fails on the first Monday, every store's counts are lost or redone. Make it two-way by running the app in 10 stores while keeping paper as a backup for a month, then moving in waves. (4) Two-way in software but costly in trust: shops plan their cash around it. A one-pager to the head of partner sales, with four weeks' notice to shops and a review after a month. (5) Mostly one-way: contracts, card data, reconciliation. A decision memo; run both providers in parallel for a share of orders and move in stages, so you can switch back.

The most one-way is the lease. Even there, ask for a break clause at year two, or a shorter lease on a smaller site first, so the expansion becomes a sequence of smaller decisions.

Run a pre-mortem on paper

Mangrove plans to let customers choose a substitute item in the app when the item they ordered is out of stock, before the picker packs the order. Pickers will send a one-tap choice to the customer and wait up to 2 minutes for an answer. The goal is to halve refunds for cut items. Imagine it is six months later and the feature failed. Write six reasons, pick the top three, and give each a change to the plan or an early warning sign.

  • Reasons cover different people: customers, pickers, riders, store managers
  • At least one reason is about timing or workload, not only the software
  • Each of the top three has a concrete change or a measurable warning sign
  • At least one reason questions whether the goal metric can move at all
Show a model answer

Reasons: (1) customers don't see the message in time, so pickers wait 2 minutes and orders run late; (2) pickers skip the step at peak because it slows them down; (3) suggested substitutes are poor (wrong size or brand), so customers refuse; (4) riders wait at the store for late orders and on-time delivery falls; (5) refunds don't fall because most cuts happen after packing; (6) store managers are measured on picking speed, so they tell staff to skip it.

Top three: (1) a push notification plus a default the customer sets in advance ("always substitute similar" or "never"), with a warning sign of reply rate under 50%; (2) and (6) together, by counting substitutions offered in the store's picking measures, and capping the wait at 60 seconds at peak; (5) by checking first where cuts happen, using a week of tagged data, before building anything.

Check yourself

Your memo asks the managing director to approve a 20-store pilot, a rollout to 320 stores and a $192,000 hardware purchase, all in one meeting. The pilot has not started. What is the best restructure?

A pre-mortem for a new Partner app feature produced 14 reasons for failure. What should happen to them?

The managing director reads your memo and says, "I can't tell whether $1.3 million is a lot or a little." What did the memo most likely get wrong?

Flashcards

Open my review

Key takeaways

  • Put the ask first: who decides, what, by when. If you can't write it in two sentences, you aren't ready.
  • Match the process to the door. Two-way doors get a one-pager and a quick owner decision; one-way doors get a memo, a pre-mortem and an approver.
  • Before walking through a one-way door, try to make it two-way with a pilot, a staged purchase, an exit clause or a switch.
  • Judge every option, including today's way, on the same criteria, and give every number a source, a range and its kind (sales or margin).
  • A pre-mortem is useful only when it changes the plan. Record every decision with its conditions and review date.

Sources and further reading

Episode transcriptPut the ask in the first line · about 11 minutes · tap any line to play from there
  1. NadiaImran, when did you last send a document to the managing director?
  2. ImranTwo weeks ago. A proposal for daily stock counts in every store. She replied with one line.
  3. NadiaWhat did the line say?
  4. Imran"What are you asking me for?" The ask was on page three.
  5. NadiaThat's today's episode. Writing for decisions: one-pagers, decision memos, one-way and two-way doors, and the pre-mortem, which finds out how a plan fails before you commit to it.
  6. ImranStart with why I should write at all. A meeting with slides is faster.
  7. NadiaFaster to prepare, slower to decide. Picture a family deciding whether to buy a flat. The eldest daughter sends everyone a one-page note before Sunday dinner. First line: do we buy the flat on Lake Road before the offer closes on the thirtieth?
  8. NadiaThen her recommendation, three options with the monthly cost of each, what could go wrong, and one honest line. Renting we can undo with a month's notice. Buying we can't undo for years.
  9. ImranSo at dinner nobody asks her to explain the numbers again.
  10. NadiaRight. They argue about what actually matters, whether they'll still want to live in that city in ten years. Forty minutes instead of three weekends. And in June, when someone asks why they chose it, the note answers.
  11. ImranAmazon is famous for this, isn't it?
  12. NadiaYes. In his twenty seventeen letter to shareholders, Jeff Bezos wrote that they don't use slides in senior meetings. They write narratively structured six-page memos, and every meeting starts with silent reading, in a kind of study hall. He also said a great memo probably takes a week or more to write.
  13. ImranA week. I don't have a week for every decision.
  14. NadiaYou don't need one. There are three sizes. A one-pager, for small or easy-to-undo decisions that one or two people make. A decision memo, two to six pages, when there's money, many people, a hard-to-undo choice, or disagreement.
  15. ImranAnd the third?
  16. NadiaThe press release and FAQ from the spec course, when the open question is whether anyone will value the result at all.
  17. ImranSo the press release isn't only for specs. It works for a bet too.
  18. NadiaWhen the benefit itself is in doubt, yes. When everyone accepts the benefit and the argument is about cost or timing, the memo is better, because its options table puts the choices side by side.
  19. ImranIs there anything all three have in common?
  20. NadiaOne rule. The ask comes first. Who decides, exactly what, and by when. If you can't write the ask in two sentences, you're not ready to send it.
  21. ImranLet's do the memo itself. What's the order?
  22. NadiaThe order follows the questions the decider asks. Title as a yes or no question. The ask. The recommendation in one paragraph. What we know, with sources. Options in a table, judged on the same criteria. Why this option. Risks and the pre-mortem. How we'll know, and when we look again. Then an appendix.
  23. ImranIf you can see the screen, the first diagram is the life of a memo. A column of boxes from draft to decision log, with two loops going back up. If you can't see it, picture a ladder where you sometimes climb back down a rung.
  24. NadiaThe first loop is the rewrite after contributors comment. The second starts in the meeting itself. If the approver finds evidence missing, you name the gap and go back, instead of pretending you've decided.
  25. ImranMy draft had two options. Do nothing, and my idea.
  26. NadiaAnd how did that look to the managing director?
  27. ImranLike I was steering her. Especially because a vendor had pitched her shelf cameras the month before, and I'd left them out.
  28. NadiaThat's the most common way memos lose trust. Every option has to be a real one, described the way its strongest supporter would describe it. Put the shelf cameras in, on the same rows: effect on cut orders, cost, time to a result, how reversible it is. Let her see the trade-off herself.
  29. ImranAnd the numbers. Ken asked whether my value was sales or profit, and I didn't know.
  30. NadiaEvery number needs a source, a range, and its kind. Sales or margin. In your case, each point of cut-order rate is about six hundred orders a day at roughly three dollars each, so about six hundred and fifty thousand dollars of sales a year. At a twenty-five percent margin that's roughly a hundred and sixty thousand.
  31. ImranSo my base case, two points, is about three hundred and thirty thousand dollars of margin, not one point three million.
  32. NadiaBoth are true. Say which is which, and give the low case. A managing director who spots that difference herself starts checking everything else you wrote.
  33. ImranOK. One-way and two-way doors.
  34. NadiaAlso from Bezos, his twenty fifteen letter. Some decisions are consequential and irreversible or nearly irreversible. One-way doors. They need to be made carefully, slowly, with consultation.
  35. ImranAnd the rest are two-way doors.
  36. NadiaMost decisions, he says. They're reversible, so they should be made quickly, by one person or a small group with good judgement.
  37. ImranAnd in the twenty sixteen letter he said most decisions should be made with about seventy percent of the information you wish you had.
  38. NadiaRight. So teams go wrong in two directions. They treat a button label like a one-way door, with three review meetings. Or they treat a three-year contract like a two-way door and sign it after one meeting.
  39. NadiaThe second diagram is a small decision tree. Easy to undo? If yes, the owner decides with a one-pager, in days. If no, there's a middle box: can we make it two-way? Only if the answer is still no do you get the full memo, the pre-mortem and the approver.
  40. ImranHow do you make it two-way?
  41. NadiaFour moves. Pilot first, twenty stores before three hundred and twenty. Stage the purchase, buy twenty devices before three hundred. Negotiate an exit clause. Or put the change behind a feature flag, or keep the old way running for a month.
  42. ImranThat's what happened with my scanners. I'd put a hundred and ninety-two thousand dollars of extra scanners in the ask.
  43. NadiaAnd what did you do?
  44. ImranTook it out. The pilot runs on the scanners stores already have, with counts in the quiet hours between two and four in the afternoon, after most deliveries. The scanner purchase comes back as a separate decision, only if pilot data shows staff queueing for them.
  45. NadiaSo the riskiest part of your memo became a later, better-informed decision. And the managing director can say yes today to the part that's easy to undo.
  46. ImranWhich brings us to the pre-mortem. That's where the scanner problem came from.
  47. NadiaGary Klein described the pre-mortem in the Harvard Business Review in two thousand seven. You gather the team and say: imagine it's six months from now and the project failed. Why?
  48. ImranDoes imagining it really help?
  49. NadiaKlein points to research from nineteen eighty-nine showing that imagining an outcome has already happened improved people's ability to identify the reasons for it by about thirty percent.
  50. ImranAnd it gives people permission to doubt.
  51. NadiaExactly. In a normal planning meeting, doubting the sponsor's idea can look disloyal. In a pre-mortem it's the job.
  52. ImranHow do I run one?
  53. NadiaSix to ten people, including a frontline user and a sceptic. Set the scene in one sentence. Three minutes of silent writing. Then go round the room one reason at a time, vote, and change the plan for the top three to five.
  54. ImranWhy the silent writing?
  55. NadiaBecause once the most senior person names a reason, everyone builds on theirs and the unusual ones disappear.
  56. ImranOurs came out with stores skipping counts in festival weeks, staff typing a guess instead of counting, and counting fighting receiving for the scanners. Each one changed the plan. Counts pause in the two peak weeks, there's a random recount of ten lines per store each week, and the scanners came out of the ask.
  57. NadiaThat's the test of a good pre-mortem. If nothing in the plan changed, you held a meeting, not a pre-mortem.
  58. ImranSo the Do it section, in words. Write the question and the ask before any analysis. Mark each part of the ask as one-way or two-way, and try the four moves on the one-way parts.
  59. NadiaThen the options table on shared criteria, the top half of the memo, the pre-mortem, and the bottom half with gates and a review date.
  60. ImranThen a two-minute test with someone outside the project, cut about a third of the words, and log the decision the same day.
  61. NadiaAnd for the small decisions that follow, a one-pager. Pausing counts during the festival is a good example. Decision needed by the thirtieth, from the head of store operations. If nobody objects by Thursday, it goes ahead.
  62. ImranFor my team, the questions I'll ask before any document goes out. If she reads only the first paragraph, does she know what we want? Is today's way of working in the options, described fairly? Which part is a one-way door, and can we move it to later?
  63. NadiaAnd one more. Is the value in sales or in margin, and does the memo say which?
  64. NadiaHere's a question for you. Take a second. A pre-mortem produced fourteen reasons the project could fail. What do you do with them?
  65. ImranFix all fourteen before we start?
  66. NadiaThat's the tempting answer, and it would delay you for risks that will probably never happen. Vote, take the top few, give each a change to the plan, an owner or an early warning sign, and look at the whole list again at each review.
  67. ImranLet me sum up. One: the ask goes first, who decides, what and by when. Two: match the process to the door, and try to make a one-way door two-way before you walk through it. Three: a pre-mortem is only worth it if it changes the plan.
  68. NadiaOne nuance on the first. The ask first doesn't mean the argument is short. It means the busiest reader can stop after one paragraph, and the reader who wants to check still finds every number and its source further down.
  69. ImranAnd next time?
  70. NadiaNext time you take that memo into the room. Twenty minutes with the managing director, answer first, and the tough questions she's going to ask.
Finished reading?Mark it complete to fill in its stop on your course line.

Use it at work

PRD for a new feature

Problem with numbers, users, goals and non-goals, measurable metrics with a guardrail, user stories with acceptance criteria, scope, risks and launch.

Open in the Product & BA Lab
Next stopExecutive updates and presentations
Was this lesson helpful?
Questions and discussion

Loading the discussion…