ingridleiria Strategic Plan and Action Plan
Most strategy documents fail at the join. The strategy half is a set of ambitions nobody in the room disagrees with, the plan half is a list of activities already happening before the offsite, and no line connects the two. The document is circulated, admired, and never opened again, because nothing in it tells anyone to do something different on Monday.
The cost is not the two days of leadership time. It is that the company spends a year executing its previous priorities while believing it changed direction, and that nothing ever stops, because nothing says what stopping would look like. Twelve months later the same team meets, notices the numbers did not move, and writes a longer document. The test applied throughout: could someone who was not in the room execute this on Monday, and would they know what to stop doing.
Start here for next year's plan
If you have arrived with "we need a plan for next year", open this file first. Three skills make up the planning cluster and they run in order.
- This one sets the destination and the bets. Where the company is going, with a number and a date; what has to be true to get there; and what will not be done. Nothing downstream is decidable until those exist, even if for now they exist only in the leadership team's head.
okr-planningcomes next. It takes the bets from this document and converts them into measured objectives with baselines, targets, dates, named owners and named measurement sources, plus the scoring rhythm. This file says what has to be true; that one says how we will know, by when, and who is holding it.annual-planning-and-headcountruns alongside or straight after. It prices the plan and sequences the hiring: whether the bets are affordable, who arrives in which month, and what the whole thing costs to operate.
Run them out of order and the usual failure follows. Objectives written before a direction is agreed turn into a negotiation between functions, and a headcount plan built before the bets exist quietly funds last year's shape for another twelve months.
When to use this, and when not to
Use it when the direction of a company, a function, or a large programme has to be set or reset; when a leadership team has an ambition and no path; when a plan exists but nobody can say which parts are bets and which are certainties; when growth has stalled and the reason is contested; or when a turnaround needs a plan that fits on a page.
Do not use it to set measured targets where the direction is already agreed and stable: that is okr-planning, which starts from this document's bets and converts them into objectives with baselines and scores. Do not use it to build the budget or the hiring sequence; that is annual-planning-and-headcount, which prices this plan and decides who arrives when. Do not use it to design the meetings that review the plan; that is operating-cadence-design. Do not use it for a single contested choice inside an agreed direction, which is decision-memo, or to diagnose a problem whose cause is unknown, which is structured-problem-solving and runs first.
The boundary: this skill decides where the company is going and what has to be true to get there. Everything downstream measures it, funds it, staffs it, or reviews it.
What you need before starting
The current numbers, with their trend over at least four periods. A snapshot hides direction of travel. Missing: build the baseline from what exists, mark every figure as measured or estimated, and list the unavailable ones at the front. A visible gap is a finding; a quiet one is a trap.
What each leader believes is wrong, collected separately and in writing before any group session. The disagreements found here are the real content of the plan. Missing: run the session anyway and treat the first hour as diagnosis.
The constraints already fixed. Cash and runway, a commitment made to a customer or investor, a contract that cannot be exited, a regulatory date, a preference the chief executive has stated that nobody will contradict. Missing: ask directly, because a strategy that ignores a fixed constraint gets found out at the first review.
Who decides when the group does not agree. Missing: name the likely person provisionally and let them correct you. A plan with no decider produces a document of themes.
Last period's plan and what happened to it. It shows which bets were made, which were quietly abandoned, and whether this organisation is capable of stopping anything. Missing: ask what was on the priority list twelve months ago and what became of each item.
Some evidence for each candidate bet. Missing: the bet still goes in, labelled untested, and its first milestone becomes the cheapest test that would move your confidence.
The method
Write the baseline first, and write it so an insider would wince. What is working, what has been quietly failing, and the number nobody says out loud. The rule: if a sceptic inside the business could say "that is not what is happening", it is not finished. A plan built on a flattering baseline collapses at the first review.
Agree one destination, with a number and a date, and test it for reachability. One sentence, not three themes. The number must reflect the health of the business rather than its activity, and it must be movable inside the horizon: take the lag between action and effect, sales cycle plus implementation plus renewal cycle, and ask whether anything started in month one can show in the metric by the date. Where it cannot, lengthen the horizon or pick a causal leading number. If the team cannot agree on a destination at all, stop; that disagreement is the real work.
Generate candidates by finishing "we will reach the destination only if ..." Eight to twelve, from the whole team, in writing and separately, without filtering. What survives grouping is usually four or five real claims and a lot of activity described as strategy.
Cut to three to five bets, each a claim that could be false. For each: the claim, the evidence for it, the evidence that would overturn it, and confidence as high, medium, or low. The cutting rule: if nobody could argue against a candidate, it is a platitude taking a slot from a real bet.
Write what will not be done, naming things currently being done. At least three items, at least one with a current owner and budget. Where the group will not name anything, ask the resourcing question instead: which bet gets the next available engineer, and which gives one up.
Build one workstream per bet, and nothing else. Anything unattached goes on the run-the-business list or gets cut, and both are visible decisions rather than omissions.
Give each workstream an owner, outcome, milestones, first step, dependencies, resources, and kill criteria. One named person, never a team and never two. Milestones are verifiable events with dates, never percentages, because a percentage cannot be disputed and therefore cannot be acted on. The first step starts within two weeks; a first step eight weeks out means the workstream starts never.
Write the kill criteria while everyone is still optimistic. Per bet: a number or an event, a date, and a named checker who does not own the workstream. "If time to second use case is not below fourteen days by 30 April, we revert" qualifies; "if it is not working we will reconsider" does not. A bet with no stopping condition never stops, it only gets quieter, and it keeps consuming people who could be elsewhere.
Name the two or three risks that would break the whole plan, each with its earliest observable signal. A risk with no signal cannot be managed, only regretted.
Set the review before the plan circulates, then circulate with the narrative first. Monthly on milestones and dependencies, quarterly on whether the bets hold, dates in calendars in the same session. The narrative is two to four pages a new joiner could read and repeat, tables in the appendix.
Writing a bet so it can actually be disproved
Four tests before a bet enters the document. Could a competent colleague argue the opposite in one sentence? If not, it is a platitude. Does it name a mechanism? "Enterprise demand will grow" is a forecast; "enterprise buyers switch when their incumbent raises prices at renewal, so a renewal-window campaign converts them" says how the money moves. Is there a cheap test? Name the cheapest experiment that would move confidence usefully and its cost in weeks; where one exists it becomes the first milestone, and where none does, say so, because that bet is more expensive than it looks. Would you notice it failing? Name the number that moves first if the bet is wrong, and where none exists, instrument it in month one.
Confidence is stated openly. One high-confidence bet and three low-confidence ones is a normal plan honestly described. Four high-confidence bets is a plan nobody stress-tested.
Worked example
Situation. Kestrel Systems, 210 people, field-service scheduling software, revenue 31.4 million dollars growing at 14 percent, all figures in this example being US dollars, down from 39 percent two years earlier. Net revenue retention 96 percent. The previous year's offsite had produced nine strategic pillars, none with an owner. The chief executive's brief: the board meets in five weeks and asks every quarter why growth stopped, and the answer has been a different theory each time.
Task. One plan, circulated in four weeks and presented to the board in five, that named a destination and the bets and could be executed by people who were not at the offsite. Good meant that at the next quarterly review each owner could say whether their bet was holding.
Action. Written input from all seven leaders beforehand, on one question: what is actually wrong. The answers split into two camps that had never been stated side by side. Three believed the product had fallen behind on integrations and enterprise buyers were leaving. Four believed mid-market accounts, 44 percent of new logos, churned at twice the enterprise rate. That was the first hour, settled by cutting cohort data by segment for the first time: mid-market gross retention 79 percent, enterprise 93, enterprise win rates flat for six quarters. The integration theory was wrong and had driven the roadmap for three quarters.
The first destination drafted was "growth back to 30 percent by year end", and it was abandoned on the reachability test: the sales cycle averaged 4.7 months and the mid-market base renewed across the full year, so nothing started in month one could move an annual growth rate by December. It became net revenue retention of 108 percent by 31 December, from 96, which the board accepted as the leading indicator of the growth question.
Four bets survived from eleven candidates: that mid-market churn is caused by never reaching a second use case, evidenced by 31 of 38 lost accounts having used scheduling only and never dispatch; that the same product serves mid-market if onboarding differs; that a usage-based tier would raise expansion revenue; and that the enterprise motion needs no change, stated as a bet so it could be disproved rather than assumed. The exclusion list took longest and was where the plan became real: the integrations marketplace with its two engineers, a reseller programme that had produced 140,000 dollars in eighteen months, and a planned move into an adjacent vertical. The marketplace was hardest because it was the previous strategy, and it was recorded as paused with a date to revisit rather than cancelled.
The wrong turn: the second draft added two cross-cutting workstreams, one for data and one for enablement, because several bets needed both. Within a week neither owner could say what would be different when their workstream was done. Both were dissolved and their work written into the bets that needed it, as dependencies with a named person who owed delivery.
Result. Four pages of narrative and a two-page table, circulated in week four and presented in week five. The board said it was the first plan from the company that stated what it would not do. At the April review the onboarding bet cleared its criterion at 17 days against a threshold of 21 and continued. The pricing bet did not: nine accounts adopted by 31 May against a threshold of twelve, so the tier was withdrawn rather than promoted harder, releasing a product manager and part of an engineering team six weeks earlier than an argument would have. Retention reached 104 by 31 December, short of the destination. The plan was not achieved, and the review recorded why: the mid-market fix worked and expansion did not, which is a more useful year-end position than a met target with no attribution.
A second scenario, where it goes differently
A 60-person hardware services business with 7.5 months of cash, whose board had declined to extend runway without a plan. The destination is not a growth number but a cash date: monthly operating breakeven by 31 August. The exclusion list moves to the front and is written before the bets, because under a cash constraint what you stop funds what you continue. Bets shrink from five to two, because a business with seven months of cash cannot run five experiments and staff none of them properly. Kill criteria become cash thresholds rather than performance thresholds: not "if this does not work we stop" but "if monthly burn exceeds 620,000 dollars in any month, the second bet stops regardless of progress". The review goes weekly on cash.
What changed: horizon, number of bets, position of the exclusion list, currency of the kill criteria, review frequency. What did not: every bet still falsifiable, every workstream still one owner and a first step inside two weeks.
Output
Two artefacts, circulated together.
WHERE WE ARE
[Baseline. Numbers with trend. What is working. What has been quietly
failing. The figure nobody says out loud.]
WHERE WE ARE GOING
[One sentence. One number. One date. Then why this number and not
another, and what it is a proxy for.]
WHAT HAS TO BE TRUE
[Three to five bets, a short paragraph each: the claim, why we believe
it, what would overturn it, confidence.]
WHAT WE WILL NOT DO
[Named items, including things currently resourced, with one line each
on what happens to the people and money released.]
WHAT WOULD BREAK THIS
[Two or three risks, each with the signal we would see first.]
HOW WE WILL KNOW
[Review cadence, attendees, dates.]
The workstream table, one row per bet:
| Workstream | Bet it serves | Owner | Outcome (measurable) | First step (date within 2 weeks) | Milestones (event, date) | Dependencies (what, from whom, by when) | Resources (people, money, source) | Kill criteria (condition, date, checker) |
|---|
The run-the-business list, separately, so the cost of keeping the lights on is visible rather than assumed, with columns for activity, owner, effort in people, and why it continues.
Failure modes
The current activity list, reordered. Recognise it because no team would have to change anything to comply, and the exclusion list is empty. Ask each leader what they would stop if the plan were real, and put the answers in.
Bets nobody could argue with. Recognise it when every bet is high confidence and none has an overturning condition. Ask someone outside the leadership team to argue the opposite of each in one sentence; whichever they cannot argue against is a platitude.
A destination the company cannot move inside the horizon. Recognise it by tracing the lag against the date. Choose a causal leading number and state the trailing number as the reason for it.
Milestones as percentages. Recognise it because status never goes backwards. Rewrite each as an event a sceptic could confirm happened.
Dependencies noted but not owed. Recognise it when a dependency has a description and no name. Name the person and the date, confirmed with them before circulation.
Cross-cutting workstreams with no outcome of their own. Recognise it when the owner cannot say what is different when it is done. Dissolve them into the bets that need them.
Kill criteria written as sentiment. Recognise it when the condition has no number, date, or named checker. Fix it in the room; it cannot be written honestly once a bet is struggling.
Edge cases
The team cannot agree on the destination. Do not average the candidates into a compromise. Write the competing destinations as one sentence each with what each implies for resourcing, and take that to the decider as a decision-memo.
The strategy is already set by an owner or a parent company. State the given destination as a constraint at the top and spend the effort on the bets and exclusions, where the remaining judgement lives.
No usable baseline data. Write the baseline from qualitative evidence, mark it as such, and make instrumentation the first milestone of the most affected workstream.
Mid-year, the world changes. Do not patch. Re-run the baseline, the destination and the bets only, keep the workstreams that survive, and stop the rest explicitly with a note on what changed. A silently amended plan teaches people the document is not real.
The plan is for a function rather than a company. Same method, with the destination expressed as that function's contribution to the company destination, and the exclusion list negotiated with whoever loses the capacity.
Quality bar
- One destination, with a number and a date, reachable inside the horizon.
- The baseline is one a sceptical insider would accept, including the number nobody says out loud.
- Three to five bets, each a claim that could be false, each with the evidence that would overturn it and a stated confidence.
- An explicit list of what will not be done, naming at least one thing currently resourced.
- Every workstream has one named owner, dated event milestones, and a first step inside two weeks.
- Every bet has kill criteria with a number, a date, and a named checker.
- Every dependency names the person who owes it and the date.
- The review cadence exists, with dates in calendars, before the plan circulates.
Adapting this to your context
The twelve-month horizon, the three to five bets and the monthly and quarterly review split come from venture-backed commercial companies of fifty to five hundred people planning on an annual cycle.
- Three to five bets over twelve months. Cut to two where cash is short. Genuinely multi-year businesses, infrastructure, pharmaceuticals, capital projects, hold a three-year horizon with the bets re-tested annually.
- A destination as a growth or retention number. For a public body, a charity, a university department or an internal function it becomes a service level, a cost per unit or an outcome measure. The reachability test does not change.
- A first step inside two weeks. Where work can begin only at a budget round, a procurement cycle or a committee date, the first step is the paper that enters that cycle, dated to its deadline.
- Kill criteria as performance thresholds. Under a cash constraint they become cash thresholds. In a regulated or contracted setting they become dates: the last moment the option can still be exited without cost.
- What not to change. Every bet is a claim that could be false, written with the evidence that would overturn it, and the plan names at least one thing currently resourced that will stop.
Related skills
structured-problem-solving runs first when the cause of the problem is contested, and produces the diagnosis the baseline rests on. okr-planning takes the bets from this document and converts them into measured objectives with baselines, targets and a scoring rhythm. annual-planning-and-headcount prices the plan and sequences the hiring that staffs it. operating-cadence-design builds the rhythm that reviews it, and weekly-status-update and meeting-to-decisions are the artefacts that rhythm produces. decision-memo handles a single contested choice inside the plan rather than the plan itself. board-and-investor-management carries the plan to the board, and ceo-communications carries the narrative to the company.