The Principle
"Cut latency to 100ms, raise coverage to 90%" isn't strategy — those are goals. Strategy answers three things: what's the core obstacle right now (diagnosis), what overall approach gets around it (guiding policy), and what will you concretely do and give up (coherent actions). Anything without a trade-off is just a wish.
In Their Words
"A good strategy has an essential logical structure I call the kernel. The kernel of a strategy contains three elements: a diagnosis, a guiding policy, and a set of coherent actions."
— Richard Rumelt, Good Strategy Bad Strategy, Ch.5
The Strategy Kernel: Three Parts
① Diagnosis: Simplify the mess into one real bottleneck. E.g., releases are slow not for lack of people, but because a shared test environment makes teams block each other.
↓
② Guiding Policy: The overall direction that answers the diagnosis (not the details). E.g., this year we bet only on "environment isolation + self-serve pipelines," nothing else.
↓
③ Coherent Actions: Mutually reinforcing moves that explicitly drop other things. E.g., cut two nice-to-have projects, move those people to the platform.
Scene
Context: Your director asks "what's your team's strategy for next year?" — they're gauging whether you're ready for the next level.
✗ Fake strategy
"We'll improve stability, speed up delivery, improve dev experience, and drive P0 incidents to zero." Four fine goals — but it's a wish list: no named obstacle, no trade-off, four things at once means no priority. The director just hears "still at the tactical layer."
✓ Real strategy
"Diagnosis: our slow delivery isn't headcount — it's six teams queuing on one shared pre-prod environment. Policy: next year we bet on one thing — environment isolation and self-serve pipelines, so each team ships independently. Trade-off: I'll pause optimizations X and Y and move two engineers to platform. Those two areas stand still this year — I think it's worth it, because they bottleneck everyone."
Checklist: Strategy or Wish?
- Does it name a specific bottleneck, rather than list areas that should all get better?
- Does it say what you'll give up? (No trade-off = not a strategy.)
- Do the actions reinforce each other, or are they parallel to-dos?
- Strip the word "strategy" — does it still read like a slogan? If yes → rewrite.
Common Mistakes
- Treating OKRs as strategy: OKRs are goals and metrics, not "how we win." A table full of "improve X% / hit Y," with no diagnosis or trade-off, is exactly Rumelt's "bad strategy."
- Filling with buzzwords: "build a world-class platform, empower the business" — fancy filler; delete any word and the meaning is unchanged.
- Fear of trade-offs: wanting everything spreads resources so thin that nothing gets done well. The essence of strategy is choosing what not to do.
The Principle
Firefighting feels indispensable but traps you in Quadrant 1 forever. The truly high-leverage work is almost all in Quadrant 2 — important but not urgent: root-cause fixes, building mechanisms, growing people. It never shows up to force your hand, so it's always crowded out.
In Their Words
"I have two kinds of problems, the urgent and the important. The urgent are not important, and the important are never urgent."
— Dwight D. Eisenhower (1954 speech, quoting a former college president)
The Eisenhower Matrix: Which Box Eats Your Time?
Urgent
Not Urgent
Important
Q1 FirefightOutages, deadlines. Must do, but living here = reactive leadership.
Q2 Prevent ★Root-cause fixes, automation, growing people, writing strategy. Where the leverage is — and the first thing crowded out.
Not Important
Q3 Fake-busyMost meetings, pings, others' urgencies. Delegate or decline.
Q4 WasteDoomscrolling, over-reporting. Cut it.
Scene
Context: Your team has fought the same class of incident three weeks running. Morale is dropping; no one has time for root cause.
✗ Firefighter heroics
You lead the charge, fix it by hand overnight, the team is grateful, and you earn a "reliable" reputation. But next week the same incident returns — because no one touched the structure. You've trained yourself into being the most expensive firefighter.
✓ Budget for prevention
"We stop the bleeding this week, but I'm carving out 20% of next sprint's capacity for prevention — not reallocatable." Run a blameless postmortem, turn root causes into actions with an owner and a deadline in the real backlog. To your boss: "Fighting this three more times costs more than one root-cause fix. This is investment, not delay."
Checklist: Preventing or Just Fighting?
- In the last two weeks, how many hours went into Q2 (important, not urgent)? Can't say = probably zero.
- After each fire, does the root cause become an action with an owner and deadline?
- Does next sprint have explicitly protected prevention capacity (not "if we get time")?
- Does my team reward "firefighting heroes," or "people who make fires stop happening"?
Common Mistakes + A Note for Women
- Rewarding fighting, ignoring prevention: firefighting is visible, dramatic, easy to praise; prevention is silent and un-awarded. So the team learns to start fires and put them out.
- Busy mistaken for contribution: filling the calendar with Q1+Q3 creates an illusion of productivity, but org progress comes almost entirely from Q2.
A Note for Women
Firefighting and "glue work" (coordinating, covering, cleaning up) often count as non-promotable work (Babcock, The No Club), and it's disproportionately assigned to women — both from a default assumption they're "good at caring" and because it's harder for them to decline. The cost: you keep the system running in Q1/Q3 all day, then get asked at review time "where's your strategic impact?" The fix isn't to stop helping — it's to cap this work and rotate it, and explicitly redirect the saved time into Q2 work that goes into a promotion case.
The Principle
Strategic thinking doesn't happen "once things calm down" — it always loses to something more urgent. The only reliable fix is to make it a fixed cadence: an immovable block, decoupled from execution, just for looking up. Relying on willpower to find a gap is guaranteed to fail.
In Their Words
"Effective executives, in my observation, do not start with their tasks. They start with their time... Time is the scarcest resource, and unless it is managed, nothing else can be managed."
— Peter Drucker, The Effective Executive, Ch.2
Scene
Context: Every week you tell yourself "this week I'll make time to think about team direction," and every week tickets swallow it.
✗ Willpower in the gaps
"Once this rush is over, I'll plan properly." But "this rush" never ends. Tactics are a bottomless pit — they expand precisely to fill every minute you left for strategy. A quarter goes by with not one real zoom-out.
✓ Fix a looking-up cadence
Lock 60 minutes every Friday afternoon as "strategy review" — no tickets, no messages — to answer just three questions: (1) Did we move closer to or further from the North Star this week? (2) Am I spending time firefighting, or did I get into Q2? (3) Is there one important-but-not-urgent thing that, started today, makes next quarter easier? Write the answers in the same rolling doc — it becomes the trace of your strategy.
Checklist: Build a Sustainable Cadence
- Do I have a weekly, calendared, never-cancelled zoom-out block?
- Is it physically isolated from daily execution (no tickets / messages)?
- Do I run layered cadences: weekly for progress, monthly for direction, quarterly for whether the strategy still holds?
- Do the judgments land in a doc I can revisit, rather than evaporating?
Common Mistakes
- Cadence decays into a status meeting: the zoom-out time fills with progress updates. The tell — are you asking "how far have we gotten" (tactics) or "should we still be doing this" (strategy)?
- Cadence without output: you think it through but write nothing, so next week starts from zero. Thinking with no rolling doc equals no thinking.
- First to be sacrificed: when busy, you cut the strategy block first — exactly backwards. It should be the last to go, not the first.
The Principle
The long-term cost you sacrifice for the quarter — tech debt, trust debt, talent loss — doesn't appear on this quarter's invoice, so it's easy to cut again and again. But it accrues interest, and the interest compounds. Bet on the long term deliberately; and admit honestly: not every moment should be about the long term.
In Their Words
"We believe that a fundamental measure of our success will be the shareholder value we create over the long term... We will make bold rather than timid investment decisions where we see a sufficient probability of gaining market leadership advantages."
— Jeff Bezos, 1997 Letter to Shareholders
Scene
Context: To hit a release, the team says for the third time "let's hack it for now, pay the debt next quarter."
✗ Off-the-books accounting
"Fine, ship it, we'll clean up later." Debt piles up silently, no one records it, and "later" never comes. Six months on, the system is too brittle to touch, and your strongest engineer quits because they're "patching a mountain of garbage daily." That's when the invoice arrives — in full.
✓ Make the debt explicit, set a ceiling
"We can hack it, but I'm filing a debt ticket now and writing down the interest (how much it slows us each week), with a rule: no new hacks once repayment capacity drops below X." Translate the cost into your boss's language: "Take three shortcuts in a row and the fourth feature ships at half speed — give me two weeks to repay, and that speeds up next quarter."
Checklist: Is the Long/Short Balance Off?
- Is our tech / trust debt explicitly on the books, or does it only live in people's complaints?
- Can I translate long-term investment into short-term language my boss gets (speed, risk, retention)?
- Each cycle, do I place at least one long-term bet, even just 10% of capacity?
- Reverse check: am I using "doing long-term work" as an excuse to dodge what's genuinely urgent and must ship now?
Common Mistakes (with Honest Trade-offs)
- Demonizing all short-termism: sometimes short-term is right — a cash-flow cliff, a survival release, broken trust that needs a quick win first. Lecturing about a three-year vision then is tone-deaf. Survive first — Sun Tzu's "first make yourself invincible."
- Long-term as a fig leaf: some use "I'm doing strategy / platform / long-term" to dodge delivery pressure. Long-termism doesn't exempt you from shipping; it asks you to ship without stealing from the future.
- Equating "expensive" with "long-term": gold-plating and over-engineering get dressed up as long-term investment, but they're just another kind of waste. Long-term ≠ slow and heavy.
Going Deeper · Pushing the Edges
1. In a company that grades by the quarter, will playing the long game get you cut in the short run?
A real tension you can't pretend away. The pragmatic move is to buy long-term room with short-term results: each cycle deliver near-term wins striking enough to bank "this person is reliable" credit, then spend that credit on long-term bets no one is forcing. Pure long-termists who can't show short-term results get cut before they pay off; pure short-termists get crushed in three to five years by the debt they buried. The answer isn't either/or — it's cadence: survive short, compound long.
2. The higher you go, the more you should do only strategy and fully let go of tactics?
Right direction, but "letting go of tactics" isn't "going blind to them." A senior leader fully detached from the front line ends up with strategy as a floating slide deck — they've lost the signal that calibrates strategy. Andy Grove kept "monitoring" depth (sampling a few frontline details) precisely for this. The healthy state: default to strategic altitude, but keep a few probes straight to the ground — sample one incident, sit in on an oncall, read a few raw bits of feedback. A strategist cut off from detail plans from imagination.
3. Strategy needs stability, but the environment changes daily — how often should you tear it up?
Separate the change-frequency of "strategy" and "tactics." The guiding policy (what you bet on, what you drop) should be stable at the quarter scale — flip-flopping it means no strategy and a disoriented team. But tactics (how you execute) should be flexible week to week. The test is Rumelt's diagnosis: only when the diagnosis itself is falsified — your supposed core bottleneck is no longer the bottleneck — should you rewrite the strategy; if it's just execution hitting snags, adjust tactics. Set "diagnosis invalidated" as the trigger line to avoid both over-swinging and clinging to a stale strategy.
4. Does "prevention over firefighting" hold in every culture/industry?
The leverage logic is universal, but the proportion depends on the industry's failure tolerance. Where failure is catastrophic (healthcare, core finance, aerospace), firefighting capability is itself a core competency — Q1 naturally takes a larger share and deserves reward. Even so, Q2 (turning failures into mechanisms that stop them recurring) is still higher leverage — it just shifts form from "add automation" to "add redundancy and drills." In steep-hierarchy East Asian orgs where boss-summons-equals-response is the default, holding Q2 time requires aligning upward first and framing it as "for delivery quality," not appealing to "I need focus."
5. Should strategy be set by a few, or co-created by everyone?
Depends on scale and stage. The "kernel" of diagnosis and trade-off needs a few people with a whole-picture view to converge — committee-style co-creation tends to lose the trade-off (no one wants to offend) and decays into an everyone-gets-a-line wish list. But the inputs to the diagnosis should be gathered broadly (the front line knows the bottleneck), and execution must be co-created (people who didn't design it won't truly run it). Healthy pattern: gather signal openly, make the trade-off in a small room, then transparently explain "why this, what we gave up." The worst is the reverse — decide direction behind closed doors, then make everyone meet to grind your to-do list.
This Issue's Exercise · Your Day 39 Action
Do two concrete things this issue — not reflection, not reading:
Step 1 (strategy audit): Pull up the most recent "strategy / plan" you wrote and run Card 1's three questions over it — does it have a diagnosis? A trade-off? Do the actions reinforce each other? If all three are missing, it's a wish list — rewrite it as one paragraph using the three-part kernel.
Step 2 (cadence): Right now, lock a 60-minute "strategy review" block on next Friday's calendar, titled "do not book · zoom-out." When it fires, answer only Card 3's three questions and write the answers into a fresh rolling doc.
One reflection question: In the past month, how many hours did you truly spend in Q2 (important, not urgent)? If you can't say, or it's near zero — that itself is the diagnosis this issue most needs you to face.