I'm writing next quarter's plan and last quarter's didn't land.

What to do when last quarter's plan didn't land

October twenty fourteen. Lisa Su has just become chief executive of Advanced Micro Devices. The stock drops fifteen per cent on the announcement.

The market's read is straightforward. Amd is in trouble. NVIDIA holds eighty-something per cent of the discrete graphics-processor market. The CUDA software ecosystem has made the moat structural — developers write for CUDA because their customers run CUDA, and customers run CUDA because developers write for it.

Lisa Su, AMD, and the plan that kept being the wrong plan

For most of the previous decade, amd's quarterly plan has been the same plan, slightly modified. Close the performance gap. Win back the consumer.

Each year, the plan has not landed. Each year, a slightly different version of the same plan has gone up for board approval the following quarter.

Su's response is not a better consumer-GPU plan.

It is a multi-year reallocation of the firm's resource commitment toward enterprise and AI infrastructure — where NVIDIA's software moat is weaker, the customers fewer and larger, and amd's silicon-design talent can compete on architecture. The EPYC server chips, the Instinct accelerators, the open ROCm stack — all consequences of that decision. The consumer roadmap becomes notably sparse. The gaming press hits amd for years.

By twenty twenty-five, the enterprise bet has produced record revenues and a sixty-one per cent year-on-year increase in net income.

The lesson isn't that consumer GPUs were a mistake. It is that the previous quarter's plan, and several before it, had not been marginal failures. They had been the wrong plan. And the plan-revision process had been reliably producing a slightly different wrong plan.

So let's go to the office and work through it.

"I'm writing next quarter's plan and last quarter's didn't land."

The feeling is exposed.

You're behind a desk people can see across, with a document that has to be defensible, and the question hanging over it is whether anyone in the room still thinks you can write a plan that lands.

Two reasons the plan failed, and each needs a different move

Two choices. They look identical when you open the document. They need different moves.

If it was credibility: name the miss before anyone else does

Choice one: the plan was fine on paper and you wore the failure. The numbers were sensible, the targets reachable, the execution slipped just enough that the result didn't land the way the plan said it would. Your credibility took the hit even though the plan, in the abstract, was fine.

If that's the trap, open the new plan by naming what you got wrong last quarter. In plain language. Before anyone else gets to name it for you. Not a performance of humility. A specific admission. People trust people who can describe their own mistakes accurately, and they stop trusting people who can't.

If it was the numbers: present three stories from the ground

Choice two: the plan wasn't fine. The metrics you were optimising for turned out to be measuring the wrong thing. You hit your targets and the business still moved against you, because the targets weren't the right targets.

If that's the trap, put the spreadsheets away. Present the new plan as three short stories from the ground. Not forecasts. Observations. The predictive models that failed you last quarter will fail you again if you build the next plan with them, and a story that's true beats a model that's wrong.

Credibility, or numbers. The same document, two different first sentences.

Three tools for understanding last quarter before you write the next one

Three tools. The discipline is to understand last quarter before you write the next one.

What, So What, Now What: an honest review in three columns

The first is

What, So What, Now What.

Terry Borton wrote down What, So What, Now What in Reach, Touch and Teach in nineteen seventy, drawn from his work in education and refined since through clinical training and reflective practice.

Three columns on a board, three questions in sequence. The reason the tool exists is that teams reviewing a quarter that didn't land tend to jump from what happened straight to what we'll do differently next time — skipping the question in the middle that does the actual work.

The unique insight is that the middle column is where the defensiveness lives. Managers go from facts to actions because actions feel like progress. So what asks whether the facts mean what we'd like them to mean — and that's the question every team starts the conversation wanting to avoid.

What you get is a retrospective that surfaces the implications honestly, and a now what that follows from those implications rather than from each person's pre-existing view.

So. How to run it.

What. The facts of last quarter, on the board, agreed on. No interpretation yet — just what happened, in order.

So what. The implications of those facts. What they mean. The defensiveness shows up here; let the room sit with it.

Now what. The moves that follow. Each one tied to a specific so what on the board, not floating as a wish-list item.

Don't let the team get to now what until so what is on the board.

How to run a retrospective that changes the system

The second is

Retrospective.

Norman Kerth wrote about the Retrospective in Project Retrospectives in two thousand and one, codifying the current shape during the Agile Manifesto movement of the early two-thousands. The mechanics are well-known: three columns. What went well, what didn't, what we'd change.

The reason the tool exists is that teams running back-to-back delivery cycles need a structured pause — a place to look at how the work got done, not what got delivered.

The unique insight is what separates a retrospective from a complaint session. The difference sits in the what we'd change column, and in what happens to it after the meeting. The check is whether last quarter's what we'd change list shows up implemented in this quarter's what went well. If it does, the system is changing. If the same five items reappear three quarters running, you've been holding a complaint session and calling it a retrospective.

What you get is a delivery cycle that improves itself — the team's own work shaping the conditions of the next round.

So. How to run it.

Wins. Each person names one or two things, on the board.

Misses. Same shape — concrete, attached to specific moments rather than generalised complaints.

Change. Systemic items. What we'd do differently in the system, not personal commitments.

Consolidation. The what we'd change list goes onto next quarter's tracking. A retrospective that doesn't change the system is a retrospective that hasn't happened.

How to run a pre-mortem before the plan locks

The third is

Pre-mortem.

Gary Klein wrote about the Pre-mortem in a two thousand and seven Harvard Business Review piece, refining a technique from his decades of naturalistic-decision-making research.

The optimism bias blinding you this quarter is the same one that blinded you last quarter, and the standard tools — risk registers, sensitivity analysis — don't reliably get underneath it.

The unique insight is that the mechanism inverts the question. Imagine the plan has already failed. Six months from now. Disastrously. Why did it fail? The shift is the entire move. Why might this fail puts the team into a defensive posture against a future they're trying to prevent. Why did this fail puts the team into a diagnostic posture about a past they can describe. The brain does different work on a future it has already accepted.

What you get is the failure causes the team will spot for real — surfaced now, while you can still change the plan that locks them in. The diagnostic value is in the room generating the causes, because the room generating the causes is the room that will spot them when they start happening for real.

So. How to run it.

Setup. Think about who you need in the room rather than how many. The people who own the plan. The people whose work the plan depends on. No outsiders — observers change the conversation, even when they're senior and well-meaning. Cap it around ten, because past ten the room stops being a room and starts being an audience. Ninety minutes, a whiteboard. State the proposition twice, slowly: the plan we are about to commit to has, six months from today, failed disastrously. Don't soften it. The room will resist. Let it.

Silent. Five minutes of silent individual writing. Each person, on their own, lists the reasons the plan failed. Silent matters. The first cause of pre-mortem drift is one confident voice saying well, obviously the main risk is X before anyone else has had time to think. Once the room hears it, that's the answer everyone writes down.

Cluster. Round the room, one cause per person at a time, until everything's on the wall. Then cluster. Don't rank yet. Let related causes find each other — the supply-chain cluster, the team-capacity cluster, the customer-assumption cluster. The clusters are the diagnosis. If supply-chain has eight notes and customer-assumption has two, that's information about where the team's uncertainty is concentrated.

Finish. Each person gets three dots. Place them on the clusters they think are most dangerous. The top three by vote commit to mitigations before the meeting ends. Before it ends — not "we'll think about this." Actual mitigations, with names and dates.

Consolidation. The mitigations get written into the plan before it locks. Not a separate document. Not a risk-register appendix. In the plan. The Pre-mortem failed if the mitigations don't change what the plan says.

So. For next quarter's plan.

You've done the what, so what, now what on last quarter. You've held the retrospective and committed the systemic changes. You're about to lock the new document. The Pre-mortem is the last move before lock.

Six months from now, this plan has failed.

Why?

That's the toolkit. One more story before we close.

The Lisa Su story we opened with is about recognising that the plan-revision process had been producing the wrong plan. The story we close with is one step further — the recognition that the planning cadence itself was producing the wrong question.

Indra Nooyi, PepsiCo, and the wrong unit of planning

October two thousand and six. Indra Nooyi has just become chief executive of PepsiCo. The prior year's plan has landed by the numbers. Revenue up. Margins holding. The quarterly narrative defensible.

What the prior year's plan has not done is resolve the question growing underneath the portfolio for the decade prior. PepsiCo's core categories — carbonated soft drinks, salty snacks — are on trajectories that well-run quarterly execution will not fix. Obesity rates are rising. Regulatory pressure is gathering. The analyst questions are asking, beneath the language, whether PepsiCo is structurally the wrong shape for the decade ahead.

Nooyi's response, formalised as Performance with Purpose, is not a better quarterly plan. It is an acknowledgement that the quarterly plan is the wrong unit of answer to the question being asked. The portfolio shifts toward healthier products. The environmental footprint shrinks. The firm invests in its people. None of which shows as legible gains in any single quarter's numbers.

When you're writing next quarter's plan and last quarter's didn't land, the honest question is sometimes whether the quarter is the right unit to be planning in. Nooyi's move wasn't to write a better three-month plan. It was to notice that the three-month plan was the form the wrong question was taking.

The quarterly plan still existed. It stopped being the top of the stack.

When the quarter is the wrong unit to plan in

So. Su recognised the plan-revision process was producing the wrong plan. Nooyi recognised the quarterly cadence was producing the wrong question. Both moves are stepping outside the planning frame to see what the frame was excluding.

The pre-mortem is that move at meeting scale. The plan, looked at from outside the plan, before it locks.

Your next move: the sentence you'd say at the post-mortem

So. The question from this Playbook.

Imagine standing at the post-mortem six months from now, explaining why this failed.

What's the one sentence you already know you'd have to say?