Half Strategy

This is where the annual vision becomes a concrete plan for the next six months. It is the Commit step of the framework: you have set the aim, run the diagnosis, and shaped your bets, and the half strategy is where those converge into one written plan.

But a half strategy is not a list of metrics to hit. The best ones are written to be outsized and inspiring: they picture the world once the half has won, they aim past the incremental, and they make people want to build the thing. A strategy that only informs gets executed like a chore. A strategy that moves people gets outsized work out of ordinary weeks. This page is how to write one like that.

Two ground rules carry over. We plan on a six-month horizon, not 13-week quarters: a quarter is too short to be genuinely strategic and too long to truly commit to, so the half sets strategy, the quarterly checkpoint course-corrects it, and the 4-week block underneath is where 100% commitments get made. And the format is discipline-agnostic: the common case is balancing product and tech, but the same doc works for marketing, sales, or operations. You swap the functions you source from and the categories you budget across; the structure below does not change, only the labels do.

Write it end-first#

Borrow Amazon’s discipline of working backwards. Before you plan how, describe the world once you have won, as if it already shipped. Amazon writes a “PR/FAQ” for exactly this: a short future press release of the outcome, plus the hard questions, written before a line of work is scoped. It forces you to be clear about the outcome and who it helps before you fall in love with a solution.

So the strategy opens with the world after: a short, vivid, present-tense picture of the outcome, written for the person it helps. If you cannot write that paragraph, you do not yet have a strategy, you have a task list. And done well, it does something a metric target never can: it gives people a place they actually want to get to.

Picture the before and after#

Impact is a gap: the distance between today’s painful reality and the world once you win. Name both, and put a number on each.

Before: one in five shoppers who reach checkout abandon rather than re-type an address; orders flat at ~1,900 per week for two halves. After: buying is one tap, and checkout completion clears 88% on the way to 95%.

The before earns the urgency; the after earns the belief. A strategy that names both moves people in a way “hit 88%” never will, and it makes the size of the prize impossible to miss.

Aim for outsized, not incremental#

Most strategy defaults to 10% better than last half. Outsized, disproportionate impact comes from a different kind of thinking, and it is a small but non-negotiable part of writing the strategy: before you settle on a Big Rock, pressure-test it against its outsized version.

The 10x question. “What is the 10x version of this, and what would have to be true?” Aiming for 10x forces you to abandon a broken baseline and find a fundamentally better approach; aiming for 10% invites tweaking the baseline. 10x is often easier than 10%, because far fewer people are attempting it. (Peter Thiel; Astro Teller.)
First principles, then delete. Reason from the real constraint, not from “how it is done.” Then delete the step before you optimize it, because the fastest checkout is no checkout page at all. (Elon Musk’s “algorithm”: question every requirement, delete the part or step, and only then simplify, accelerate, automate. Most teams start at “optimize” and polish something that should not exist.)
Name the secret. The contrarian truth the strategy rests on: “most people believe X; we believe Y.” A strategy built on consensus earns consensus, which is to say average, returns. (Thiel: what important truth do very few people agree with you on?)
Concentrate force. Outcomes are power-law, not evenly spread: one rock that could be 10x can outweigh five that are 10%. Overload the one that could go non-linear rather than spreading capacity thin. (Thiel.)

You will not always ship the 10x version this half, and you should not bet the company on it every time. The rule is honest: incrementalism by default is the failure; incrementalism after genuinely considering the 10x version is a legitimate choice. Every rock should have been held against its 10x before you settle for the safe one.

Big Rocks#

A half’s plan is a small number of Big Rocks, roughly six, O(6), never a comprehensive list. Big Rocks are the prioritized bets that survive planning, committed to at the half level: each is tied to an input, and together they represent roughly 30 to 40% of total capacity for the half. The rest is keep-the-lights-on work and each function’s own priorities: real, necessary work, just not named here.

Three litmus tests decide what earns a spot, in order:

Top priorities. If we do nothing else this half, these must happen.
Resolve conflict. When two teams want the same scarce resource, the Big Rock list is what decides who gets it.
Non-goal: be comprehensive. A Big Rock list that tries to cover everything the company is doing has failed at being one. A piece of work can be real, worthwhile, and still not a Big Rock.

Each Big Rock carries two people fields beyond its metric and narrative:

Rock Manager: the person accountable, and a “Pig,” not a “Chicken”: fully committed to this rock, not juggling it among ten other things.
Hoodie Squad: the people, across teams, actually doing the work. Someone wears at most one hoodie at a time; splitting a person across three rocks defeats the point of naming only six.

The strategy doc#

A good strategy doc is short, a page or two, because its job is to force decisions, not to document them. It opens end-first (the world after, the before/after, the secret) and only then gets to the plan. Copy this outline:

# [Area] - H1/H2 [Year] Strategy

## The world after   (work backwards)
A short future press release: the outcome as if it already shipped, present tense,
for the person it helps. Vivid enough to picture. The line people remember.

## Before -> After
- Before: today's painful reality, with the number
- After: the world once we win, with the number

## The secret
The contrarian truth this rests on. "Most believe X; we believe Y."

## Diagnosis   (see the Diagnosis page for how to build this)
The governing thought, the value tree, the sized prize, and the options weighed.

## The theme
One sentence: the single most important thing this half is about.

## The Big Rocks   (each written end-first; one numbered rock each)
For each rock:
- After: one line, the world once this rock lands
- Before -> After: the number it moves (now -> target), and the input it serves
- The 10x version: the outsized form (or "none" if genuinely incremental), and the step toward it this half
- Narrative: why we believe this moves it
- Rock Manager / Hoodie Squad

## Capacity budget
Keep-the-lights-on off the top (reliability, security, compliance). Of what is left,
Big Rocks take ~30-40%; the rest is functional-area work. Name the buckets that fit
your function; decide the split before you pick rocks, not after.

## Sequencing
Which Big Rocks front-load into the first blocks, and why. Dependencies.

## Pre-mortem
It is the last day of the half and this plan failed. What most likely killed it?
For each top risk: the early warning sign, and what we will do about it.

## Explicitly NOT doing
The tempting things we are saying no to this half, each with the one-line reason.

The top three sections are what separate a real strategy from a metrics wishlist. A strategy is a diagnosis before it is a plan, and it is a destination before it is a diagnosis. A north star and three inputs listed with no world-after and no secret is a set of numbers to hit that nobody rallies to. Get the top of the doc right and the rest almost writes itself; skip it and you get a task list wearing a strategy’s clothes. The theme is the forcing function: if you cannot name the one thing the half is about, you have a list. The pre-mortem surfaces risks while you can still act on them. Explicitly not doing is where a strategy proves itself, because a plan is as much what you refuse as what you pick.

A filled example#

Here is the template filled in for the illustrative online store, so you can see a whole one at once. A real diagnosis shows its metrics rather than describing them; paste the actual charts so the reader can see the gap at a glance. Illustrative version:

Illustrative. The outcome has not moved in a year, and the three numbers underneath it show why: not one of them is near its target. A flat north star is never the thing to fix directly, it is a report on the inputs.

(In a live doc these are real screenshots from the dashboard, not a mock. The point is the diagnosis is evidenced, at a glance.)

# Storefront - H2 / 2026 Strategy

## The world after
It is the end of H2. A returning shopper taps Buy on the product page and it is
done: no cart, no forms, no waiting. The "checkout" as a separate place to get
lost has quietly disappeared. People buy from us without thinking about it, and
they come back because it is the least work anywhere on the internet.

## Before -> After
- Before: 1 in 5 shoppers who reach checkout abandon rather than re-type an address; orders flat at ~1,900/wk for two halves.
- After: checkout completion 80% -> 88% this half (~+350 orders/wk), on the path to one-tap buy and 95%+ next half.

## The secret
Everyone is racing to add features to win shoppers. We believe shoppers do not
want more features, they want fewer steps. The winner deletes friction; it does
not add capability.

## Diagnosis   (full value tree and options -> the Diagnosis page)
Governing thought: we can add ~350 orders/week (+18%) this half with zero new
traffic, by closing two leaks in a funnel we already own. Sessions are healthy;
the problem is not traffic, it is that we lose value at two steps we control.
- Sized prize: completion 80->88% ~ +190/wk · page load 3.5->1.5s ~ +160/wk · add-on attach 18->30% ~ +$1.9k/wk margin
- Options weighed: fix-conversion (ours, fast, +350) beat buy-traffic, loyalty (pilot moved nothing), new-region (defer)
- Targets from evidence: 88% = saved-address A/B lift; 1.5s = P75 of lazy-loaded pages; 30% = one-tap removes the re-entry step

## The theme
Make buying effortless: turn the carts we already have into orders, and start
deleting the checkout step entirely.

## The Big Rocks
1. Saved address & payment
   - After: a returning shopper never re-types an address again
   - Before -> After: checkout completion 80% -> 88% (~+190/wk) | input: completion
   - The 10x version: one-tap buy from the product page, no checkout page at all; this half ships saved details, the first step toward it
   - Narrative: re-typing is the #1 abandonment reason in session replays
   - Rock Manager: Product PM | Hoodie Squad: PM + 2 eng + 1 design
2. Faster product pages
   - After: pages feel instant; nobody bounces waiting for a load
   - Before -> After: page load 3.5s -> 1.5s (~+160/wk) | input: page load
   - The 10x version: instant pages via a storefront rebuild; this half takes the cheap 80% (compress + lazy-load) and defers the rebuild
   - Narrative: slow pages bounce shoppers before the cart
   - Rock Manager: Platform lead | Hoodie Squad: 2 eng
3. One-tap add-on at checkout
   - After: the relevant add-on is one tap, never a separate step
   - Before -> After: add-on attach 18% -> 30% (~+$1.9k/wk margin) | input: attach
   - The 10x version: none. A margin play, not an order driver; kept small, first to cut if a squad is needed elsewhere
   - Narrative: shoppers skip add-ons because they are a separate step
   - Rock Manager: Growth PM | Hoodie Squad: PM + 1 eng + 1 design

## Capacity budget
Compliance 10% · Security 10% · Reliability 15% · Expansion 25% · Product 40%
(These three rocks are ~32% of total capacity; the rest is functional-area work.)

## Sequencing
Faster pages first: most uncertain, and upstream of the other two. Saved-address
and one-tap add-on run in parallel in later blocks; they are independent surfaces.

## Pre-mortem: it failed because...
- one-tap add-on read as a pushy upsell and returns rose -> watch refund rate
  weekly; only surface add-ons relevant to the cart.
- image compression hurt perceived quality -> A/B before full rollout.
- the payment-vault change stalled in security review -> start that review in
  the first block, not the last.

## Explicitly NOT doing
- Loyalty points: the pilot moved nothing; revisit only with a new hypothesis.
- Full storefront redesign: a rock with no clear first pebble this half.
- New-region launch: deferred to H1 next year; it would eat the expansion budget.

Notice what makes it good: it opens with a world people want to build, not a metric; the before and after make the prize impossible to miss; the secret stakes out a real point of view; every rock carries its after-state, the 10x version it was tested against (even when the honest answer is “none”), and a number that traces back to the sized diagnosis (+190/wk, +160/wk), not an asserted one; the keep-the-lights-on reserve is set before the rocks; the pre-mortem names concrete failure modes each with an early signal; and the “not doing” list is specific enough to hold the line. That is the bar: outsized, evidenced, decisive, and short enough to be read.

Producing a high-quality strategy#

The doc is the artifact; here is the work that fills it, on the six-month cadence.

Start from Learn, diagnosis first. The retrospective from the current half is the direct input to the next, and so is every team’s two-pager. You plan forward from what actually happened, not from what you hoped last time or a shiny idea nobody has tested against the data.

Two-pager intake, bottom-up. Every team’s leads write a two-pager ahead of planning. Leadership pre-reads all of them, then runs a fast review per team, 10 to 15 minutes, built on questions rather than a walkthrough of slides.

Prioritization, top-down: $100 voting. Planning opens with a kickoff brainstorm: candidate Big Rocks go up, sourced from the two-pagers and from leadership’s own read of the half ahead. Everyone in the room gets $100 to dot-vote with, allocated across candidates however they see fit. The vote reads the room’s conviction; it is input the leader weighs, not a rule the leader is bound by. It surfaces where energy is, not necessarily where the leverage is, and that is still a judgment call.

Challenge every rock in the room. The questions that matter:

Is this the right problem, or a symptom of a bigger one?
Is the scope right: too wide, too narrow?
What is the 10x version, and what would have to be true for it?
What is our conviction: confident this is real, or exploring?
What would we have to believe for this to be the highest-ROI thing we could do this half?
Who else has to be involved when it ships: marketing, sales, operations?

What survives the challenge is worth committing to.

Run a mini pre-mortem. Before you commit, imagine it is the last day of the half and the plan failed, and ask the room what most likely killed it. Imagining the failure loosens tongues that “any concerns?” never will. For each risk, either name the early warning sign and a mitigation, or re-scope the rock now.

Ending offsite: pick the rocks, staff them. Pick the roughly six Big Rocks, assign a Rock Manager to each, and allocate the Hoodie Squads. There are usually more good candidates than people to staff them; that reconciliation, cut rather than overcommit, is the point of the room.

Writeup and all-hands. The decisions get published as the live strategy doc, and each Rock Manager presents a short poster for their rock at the all-hands: the world-after, the metric, and who is on the squad. That is the moment the plan stops being leadership’s decision and becomes the company’s shared picture of the half.

Reserve capacity before you spend it#

Here is the first-principles reason this section exists: capacity is finite, and a strategy that commits 100% of it to growth dies the first time an incident or a regulator shows up. So the hardest part of a half’s strategy is not picking exciting Big Rocks, it is not spending everything on them. You never allocate 100% of capacity to growth. Before any Big Rock, the half reserves what keeps the lights on and keeps you legal: reliability, security, and compliance come off the top. Of what is left, Big Rocks claim roughly 30 to 40% of total capacity; the rest funds each function’s ordinary work.

The mechanism is the capacity budget: split the half’s total person-sprints by category before any specific item is named, so the tradeoff is an explicit decision instead of something discovered after you have over-committed. An illustrative split for an online store, a few years old and expanding into new regions:

Category	Allocation	What it covers (online store)
Compliance & regulatory	10%	Payment regulation, data-privacy law, consumer-protection rules
Security	10%	Fraud prevention, payment and account security
Reliability & bugs	15%	Checkout errors, search downtime, crashes users report
Expansion	25%	New categories, new payment methods, new regions
Product	40%	Checkout, the add-on flow, search, the shopping experience

The split shifts every half with where the pressure is: launching in three new regions, expansion grows and product shrinks; a spike in payment fraud, security takes more. What is fixed is the discipline of deciding it up front, and the buckets themselves adapt to the function (a marketing team might split across brand, experiments, campaigns, and compliance).

The reframe that makes this easy: compliance and security are not the enemy of strategy, they protect it. A data breach or a regulatory shutdown is the fastest way to lose the entire strategy at once, not just a half of it. See Security & Data Handling for the posture behind that budget line.

The quarterly checkpoint#

Six months is a long time to go without checking the plan against reality. Halfway through the half, leadership runs a quarterly checkpoint, a quick review of the Big Rocks, not a re-plan. The questions are simple: are these still the right rocks, has the diagnosis changed, is a rock’s narrative holding up or has Learn already shown it is wrong?

The checkpoint can reallocate, kill, or add a Big Rock and adjust the resources behind it. A Rock Manager whose rock is clearly won can free their squad; a rock that is not landing can be killed rather than protected out of sunk cost; a problem nobody saw in the two-pagers can earn a spot if it has become a top priority since. It is deliberately lightweight: no two-pager cycle, no full offsite, no re-running $100 voting. The point is a mid-course correction so a six-month horizon never drifts a full half unchecked, without paying the full cost of re-planning to get one.

The output is a one-page checkpoint memo appended to the live strategy doc: a single verdict per rock, the numbers behind it, and any reallocation.

# Storefront - H2 Checkpoint (end of Q3)

## The read in one line
Two rocks on track, one killed, freeing a squad to double down on the winner.

## Rock-by-rock verdict
| Big Rock            | Target            | Now    | Verdict   | Move                          |
|---------------------|-------------------|--------|-----------|-------------------------------|
| Faster product pages| 3.5s -> 1.5s      | 1.6s   | ON TRACK  | Finish rollout, then done     |
| Saved address & pay | completion -> 88% | 85%    | ON TRACK  | Hold squad, ship in Block 3   |
| One-tap add-on      | attach -> 30%     | 19%    | KILL      | Read as pushy; refunds +2pt.  |
|                     |                   |        |           | Free squad -> Saved address   |

## What changed since planning
The add-on narrative was wrong: the intent we assumed is not there. Diagnosis
otherwise holds; checkout friction is still the leverage.

## Reallocation
Growth PM + 1 eng move off add-on onto Saved address to pull its ship forward.
No new rock added; the freed capacity compounds an existing bet.

## Still explicitly NOT doing
Loyalty, redesign, new-region: unchanged from the H2 strategy.

A checkpoint that changes nothing is a valid outcome, but write the memo anyway, so “we looked and held” is on the record rather than assumed.

From Big Rocks to blocks, and keeping it live#

Big Rocks are not worked in one six-month push. They are delivered through the 4-week Blocks & Sprints cadence underneath: the half sets the rocks, the Rock Manager, and each Hoodie Squad; every block, that squad turns a slice of its rock into a true 100% commitment for the block. Sequencing across rocks and reconciling dependencies happens at that level, not here.

The strategy doc itself is live for the whole half. It is visible to clients and leadership at any time and should always reflect what is actually happening, not what was planned in week one. If a rock gets re-scoped, a conflict surfaces, or a block reveals the narrative was wrong, update it immediately. The roadmap is not a contract; it is the team’s best current judgment, in the open.

Connection to Discover#

The half strategy decides what is worth discovering; Discover confirms whether the belief behind a Big Rock is real. As a block turns a slice of a rock into a block commitment, PM takes its metric and narrative into discovery to validate against data and rule out non-tech solutions before Define. If Discover finds the problem is not what you thought, the strategy doc updates. That is not a failure of planning; it is the system working.