Produces the launch plan rather than reviewing one: two dated goals, a channel table that sums, twelve dated weeks, and the five post-launch pillars most plans omit.
We have not measured this one. It is published because the content is checkable on the page rather than because a run beat a control.
What it carries is mostly arithmetic and two fixed offsets. The arithmetic is a channel table in which every line states an expected number of users and a timeline, and the table has to sum to at least the top-line goal before the plan is allowed to exist. That constraint alone kills the most common launch document, the one where the goal is a multiple and the tactics underneath it are a list of good intentions with no quantity attached.
The two offsets are the specific things it knows. The first: relationship-building with writers and community moderators starts in week five and the pitch happens in week eleven, so there are six weeks between first contact and the ask. The second: the affiliate payout window must be at least as long as the refund window, because a thirty-day payout on a sixty-day refund policy means paying commission on sales that later reverse. It also carries a review-request sequence with two real specifics inside it, a request made within roughly 48 hours of signup and an ask phrased as sharing a story rather than leaving a review, and a validation gate that refuses to count email addresses as evidence of demand.
Who it is not for. If you are launching a feature to existing customers rather than a product to strangers, twelve weeks of runway is theatre and you want a release note and a lifecycle email. If your sales cycle involves procurement, the post-launch pillars here are built for self-serve and will not survive contact with a purchasing department. And if you already have a plan on paper, the audit companion is the better first move, because reviewing what exists is cheaper than rewriting it.
Measuring one skill honestly costs about twenty model sessions: five runs with it, five without, on real material, each output graded alone by a session that is not told the other arm exists, against a rubric written by somebody who never saw the skill. We have not spent that on this one yet, so it ships labelled rather than ships silently.
How it would be measured. Four invented product briefs, each stating a price, a list size, a refund window and measured conversion rates, with one brief deliberately supplying no rates at all. Graded mechanically on whether both goals carry a number and a date, whether the channel table's expected contributions sum to at least the goal, whether every week from one to twelve carries at least one dated action, whether first press contact precedes the press pitch by at least five weeks, and whether the affiliate payout window is at least as long as the stated refund window. The spine is computable, so Tier A.
The objective spine here is buildable but it is narrow. You can grade mechanically whether both goals carry a number and a date, whether the channel expectations sum to the goal, whether every week carries a dated action, whether first press contact precedes the pitch by at least five weeks, and whether the affiliate payout window clears the refund window. That is five computable criteria out of a document with perhaps forty judgement calls in it.
The awkward part is the control. A strong model asked for a twelve-week launch plan already produces something with weeks in it, and the question is not whether the output has structure but whether the numbers reconcile. A fair test needs briefs carrying real conversion rates so the summation can be checked, plus one brief with no rates at all, to see whether the plan says so or quietly invents them.
The rule that decides pass or fail was written down before any run was executed and it does not move afterwards. It is in the method note on the hub, along with the full results table including every skill that was tested and cut.
Stated plainly, because a skill that claims everything is useful for nothing.
name and description in the file's frontmatter, so you can also invoke it by name.A competent marketer with a spreadsheet and two hours does the decomposition better than any document, because they know which conversion rates in your category are plausible and this does not. If you have that person, use them, and use this as the checklist they run against at the end.
Reading widely about launches is a genuine alternative and often a better use of the first afternoon, because the biggest launch decisions are about what to launch and to whom, which is a judgement problem rather than a scheduling one.
The model with no skill at all will write you a reasonable-looking twelve-week plan on the first try. What it tends not to do unprompted is force the channel expectations to sum, refuse to treat email signups as validation, or notice that the affiliate payout window is shorter than the refund window. Those are the specific places this earns its keep, and if you do not need them, you do not need this.
--- name: launch-runway-plan description: Writes a dated twelve-week launch plan from nothing. Sets two quantified goals with dates, decomposes the top line into a channel table whose expected contributions must sum to the goal, assigns dated deliverables to every week including the eight required elements of a pre-launch page and the six-week offset between first press contact and the press pitch, and specifies the five post-launch pillars: affiliate programme with a payout window at least as long as the refund window, community, multi-platform reviews, recurring revenue and upsells, and a second campaign. This skill should be used when a launch date has been agreed and no plan exists yet, when a launch goal is stated as a multiple rather than a number, or when a plan ends at launch day with nothing written for the weeks after it. --- # Launch runway plan ## The claim this skill is built on Most launch plans fail as documents before they fail as launches, and they fail in one specific way: the goal at the top is not connected by arithmetic to anything below it. The usual shape is a headline ambition, then a list of activities. Build a landing page. Post on social. Get on a launch aggregator. Do some PR. Every item is sensible. Not one of them carries a number. So when the launch delivers a fifth of the goal, nobody can say which line underdelivered, because no line had an expectation to miss. This is hope marketing, and its defining feature is that it cannot be wrong, only disappointing. The correction is structural rather than motivational. Every channel in the plan is named, is given an expected contribution in users and a timeline, and is therefore independently magnifiable if it beats the expectation or killable if it misses. If you cannot say what a channel is expected to contribute, it does not go in the plan. It goes in a list of experiments, which is a different document with a different budget. The second structural claim is that the sequence has two hard offsets in it that are almost always compressed, and compressing either one destroys the value of the activity rather than merely delaying it. They are named at week five and in the affiliate section, and they are the parts of this file most worth stealing. ## What you produce Five artefacts, in this order. Nothing here is a recommendation to consider. Each is a thing that exists at the end. 1. **Two goals**, each with a quantity and a date. 2. **A channel table**: channel, expected users, timeline, owner. The expected column sums. 3. **A twelve-week runway**, each week carrying at least one dated deliverable. 4. **A validation checkpoint** with a customer count on it, scheduled before launch week. 5. **A post-launch document**, covering five pillars, written before launch rather than after. ## Step 0. Two goals, not one Set exactly two, and make them different in kind. **One adoption goal.** A quantity of the thing that means the business works: paying customers, active teams, completed jobs. Not signups, unless signups are genuinely what you sell. It carries a date. **One earned credibility goal.** Something you cannot buy directly, stated as a countable outcome: a stated number of reviews on independent platforms, a written mention in a named publication that your segment actually reads, a specified number of customer stories you are permitted to publish. It carries a date too. Two goals rather than one, because a plan optimised only for the adoption number will spend the entire runway on the cheapest converting channel and arrive at launch with no evidence that anyone outside your own list believes you. Credibility is the asset that makes the second and third campaigns cheaper, and it has a lead time measured in weeks, so it has to be a goal or it will not be scheduled. ## Step 1. Decompose until you reach something a person does Work backwards from the adoption number through each conversion step, stopping only when you reach a quantity of raw input that a named person has to produce. For a goal of 600 paying customers in 90 days, at an assumed trial-to-paid rate of 20 percent and a visitor-to-trial rate of 4 percent, the chain reads: 600 customers, 3,000 trials, 75,000 visits, which is roughly 833 visits a day for every one of the ninety days, including the eighty-nine that are not launch day. Write every assumed rate into the document, next to where it came from. A rate that lives only in someone's head cannot be found to be wrong afterwards, which guarantees the post-launch review blames execution instead of arithmetic. If a rate has never been measured, mark it as an assumption in the plan itself, and mark it a second time in the channel table, so the reader knows which lines are evidence and which are hope. ## Step 2. The channel table, and the summation rule Now decompose the visits by channel. Each row: channel, expected users, timeline, owner, and the basis for the number. The basis column is what makes this different from a list. "Email: 6,000 addresses, 35 percent cumulative unique click across three sends, so 2,100 visits" is a basis. "Email: significant" is not. **The summation rule.** Add the expected column. If the total is below the goal, you do not have a plan, you have a goal and some activities. Three legitimate responses, and exactly one illegitimate one. - Lower the goal to what the table produces, and state the higher number as a twelve-month target instead. - Fund the one line that can be bought, and state the cost. - Add a channel with a real basis, which usually means somebody has to go and get a commitment from a partner this week. The illegitimate response is raising the expected numbers until the column sums. Everybody does this. It is why launch retrospectives are so uniformly confusing. **The concentration check.** If any single row carries more than about half the total, the plan has one channel and some decoration. Ask what happens if that row simply does not occur, because the rows most often overweighted are exactly the ones you do not control: one aggregator, one publication, one partner's audience, one platform's ranking decision. ## The twelve-week runway Weeks are counted forwards from the start of work, so week 12 is launch week. Adjust the absolute length to your runway, but do not reorder, because several of these block each other. ### Week 1. Segments, written from inside their heads Define 5 to 8 niche customer segments. The constraint that makes this useful: **each segment must have a different problem and a different use case**, not a different job title with the same problem. If two segments would read the same landing page and nod at the same sentence, they are one segment. For each, write down their fears, their motivations, and how they currently spend money on this problem, in their own vocabulary rather than yours. The output is a paragraph per segment that you could read aloud to a member of that segment without them wincing. This is week one because everything downstream, the page copy, the channel choice, the outreach list, the ad targeting, is a function of it. ### Week 2. The pre-launch page, with eight required elements Not a coming-soon page. A page that can convert a stranger into an interested, identified human. Eight elements, all of them: 1. **A founder or origin story with something verifiable behind it.** Not a mission statement. A specific reason this exists, attached to a checkable fact: a number of years spent doing the thing, a previous product, a named problem you had. 2. **A visible human element.** A face, a name, a way to reach a person. Anonymous software from an unknown domain converts badly for a reason that has nothing to do with copy. 3. **Exactly three highlighted features.** Three, not seven. The discipline of choosing three is the point, and it forces the segment work from week one to have actually happened. 4. **Something interactive to try before committing.** A live demo, a sandbox, a calculator, an interactive tour. Anything that converts reading into doing before an address is requested. 5. **Three testimonials, chosen from ten or more people you contacted.** The selection ratio matters more than the count. Three quotes chosen from three people are the three people who replied. Three chosen from twelve are the three that were most specific. 6. **An aggregate social-proof counter.** A number of people already on the list, already in the beta, already using the earlier version. If the number is embarrassing, it is too early for the counter and you leave it out rather than inventing it. 7. **Frictionless registration that names the security fear out loud.** Whatever the objection is in your category, put the answer next to the button rather than three clicks away in a policy page. 8. **Email capture that asks two or three interest-priming questions.** Which of these three problems is yours, what do you use now, how many people are on your team. This does two jobs at once: it segments the list for week eleven, and answering a question about a problem makes the problem more salient to the person answering. ### Week 3. Locate the audience and build the channel table For each segment from week one, find where those people already gather and read. Named communities, named publications, named newsletters, named search phrases. Then complete the channel table from step 2, because you now have the basis column. ### Week 4. Instrument everything, before spending anything Analytics installed, events defined and verified with a real end-to-end test conversion: page view, signup, activation, purchase, refund. Verify in a real browser session on both Windows and macOS, since a tracking script blocked by a default browser configuration on one of them is a category of bug nobody finds later. This sits at week four because it must precede the week five message tests and the week eight spend. If you already have working analytics, do it in week one instead. A campaign that ran before tracking existed produces a number nobody can act on, and it cannot be retrofitted. ### Week 5. Message tests, and first contact with the press Two things, and the second is the offset. **Split landing pages to test messaging.** Different framings drawn from the different segments, judged on signup rate. This is a test of which problem statement pulls, not conversion-rate optimisation of a checkout that does not exist yet. **Begin relationships with writers and moderators, with no ask in them.** Build a list of twenty to thirty named people who plausibly care about the segment, and make contact that contains nothing you want from them. This is the non-obvious ordering in the entire plan: the pitch happens in week eleven, six weeks later. The mechanism is not politeness. A pitch from someone a writer has exchanged two messages with gets opened; a pitch from a stranger in launch week competes with every other stranger in launch week. ### Week 6. Search foundations Technical basics, the pages that target the phrases found in week three, and the internal linking between them. Search is in the plan at week six because ranking has a lead time measured in months, which means the honest expected contribution for search inside a 90-day launch window is small and long-tail. Put the small honest number in the table rather than the large hopeful one. ### Week 7. Content plan Decide what gets published, on what cadence, for which segment, and where it will be distributed. Content that starts publishing in launch week contributes nothing to launch week. ### Week 8. Paid plan and long-tail outreach Write the paid plan, run the readiness arithmetic before committing a budget, and begin outreach to the long tail: smaller newsletters, niche communities, individual practitioners with modest but exactly-matched audiences. The long tail is here rather than at week eleven because it has no embargo and no news value to protect, so it can start early and compound. ### Week 9. Social and content advertising Put money behind the content that already performed organically, to audiences built from the segments. Advertising content rather than the product at this stage builds the retargeting pools that launch week will need, which do not exist yet and cannot be created retroactively. ### Week 10. Launch timeline and early access Write the hour-by-hour launch day sequence, with owners. Give early access to mid-tier voices in the segment: people with a real audience of the right shape rather than the largest audience available. They have time to actually use the product before launch, which is what produces a specific opinion rather than a repost. ### Week 11. Pre-launch email, and the press ask Warm up the list with two or three emails so the announcement is not the first contact in three months. Then execute the press pitch to the people first contacted in week five. This is the week the six-week offset pays. ### Week 12. Launch Community posts open, list emailed, media contacted, aggregator submitted. Order the one-shot moves so a smaller one fires first and the funnel is watched for an hour before the biggest one goes. Somebody is on support rather than on the launch. ## The validation gate underneath all of it **Collecting email addresses is not validation. Money is.** A list of 4,000 addresses is evidence that 4,000 people found a page interesting, which is a different proposition from anyone paying. So set a customer-count validation threshold before week twelve, and set it against your own time. Decide the number of paying customers at your intended price that would make the next six months of work worth it, then check whether pre-orders, paid beta places or deposits reach a meaningful fraction of it before launch week. Sell to a niche audience first, because a threshold reached inside one segment is a repeatable finding and the same number spread across eight segments is noise. And ask every buyer two questions in the confirmation email: why they bought, and what else they would buy. The first gives you launch copy written in customer vocabulary, which outperforms anything written internally. The second is your roadmap, ranked by people who have already paid. ## Post-launch: five pillars, written before launch A plan that ends on launch day discards most of the value, because a launch produces a spike that decays within days and the question is what remains. ### 1. The affiliate programme, and the payout rule Choose one structure and state it: a flat amount per sale, a higher rate for new customers than for renewals, or a recurring percentage for the life of the subscription. **The rule that costs real money if you get it wrong: the payout period must be at least as long as the refund window.** If you refund for 60 days and pay affiliates at 30, you pay commission on sales that later reverse, and clawing it back is unpleasant, slow and damages the relationship with the affiliate who did nothing wrong. Set payout at refund window plus a few days of settlement. Recruit from the people who left you the best reviews, because they have already demonstrated both enthusiasm and the willingness to write. Put the affiliate signup inside the product and in the account-creation email sequence, not on a page in the footer that only existing affiliates ever find. ### 2. Community, with three requirements Written guidelines published before the first member joins. Something exclusive that exists only there, which can be access, early builds or direct contact rather than discounts. And a feedback loop that visibly changes the roadmap, meaning shipped changes are attributed publicly to the request that caused them. A community with no visible causal link to the product is a support queue with worse tooling. ### 3. Reviews, on multiple platforms Three specifics, all of them counterintuitive. **Ask on more than one platform.** A profile with forty reviews in one place and none anywhere else reads as managed. Spread across the two or three surfaces your buyers actually check. **Phrase the request as sharing a story, not leaving a review.** "Tell us what you were trying to do" produces specific, usable prose. "Leave us a review" produces four stars and no sentence. **Ask within roughly 48 hours of signup**, while curiosity and novelty are at their highest, rather than after thirty days when the product has become furniture. Then run an escalating sequence of two or three emails rather than one, and point each at a short step-by-step page rather than explaining the process inside the email itself, because a long email with six numbered steps is a page nobody scrolls. One constraint that is genuinely dated: several major review platforms prohibit incentivised reviews and will remove them and flag the profile. Verify the current policy of each platform before offering anything at all in exchange, and treat any advice on this point older than a year as unreliable, since these policies have been tightened repeatedly through 2024 to 2026. ### 4. Recurring revenue and upsells The best upsell is more of what they already like. For a capped product, the highest-converting expansion is raising the cap for the people already pressed against it, because the need is demonstrated rather than assumed. Identify who is at 80 percent of any limit and make the next tier one click away. For people who arrived after a launch offer closed, run a deliberate late-arrival conversion: acknowledge that they missed it, and offer the annual plan instead. This converts a group that would otherwise be lost to timing, and it does it without reopening the original offer, which would devalue it for everyone who acted. ### 5. The second campaign, bigger and sweeter Plan a second campaign that is visibly better than the first. It does two jobs. It gives everyone who missed the launch a reason to stop waiting, which cures the residual regret in the audience. And it gives your affiliates, community and reviewers a second event to promote, which is when the pillars above start compounding rather than decaying. ## The decision rule: is this a plan or a wish - **The table sums to the goal and every line has a basis in something measured.** Proceed. Set the kill criteria per channel now. - **The table sums, but two or more lines rest on unmeasured rates.** Proceed, and mark those lines as assumptions in the plan document itself. Schedule the check that measures them before week eight, since that is the last week the budget can be reallocated. - **The table falls short of the goal.** Lower the goal, fund the buyable line, or get a commitment this week. State which of the three you chose in the plan. - **One row carries more than half the total.** Rewrite it as two rows with separate owners, or accept the concentration explicitly in writing, including what happens if that row returns nothing. - **You cannot tell, because no conversion rate in the chain has ever been measured.** Do not build the twelve-week runway yet. Run weeks one, two and four only: segments, a page that converts, and instrumentation. Then buy or borrow a few hundred visitors from the cheapest channel available and measure the visitor-to-signup and signup-to-paid rates directly. A launch plan built on invented rates cannot be evaluated afterwards, so its failure teaches nothing, which is the expensive part rather than the failure itself. ## Worked example, compressed A time-tracking tool for small design studios. Price £12 per month per seat. No existing product in market. Twelve weeks of runway and £6,000 of budget. **Goals.** 600 paying seats within 90 days of launch. Three reviews on independent platforms plus one written mention in a publication that studio owners read, by day 30. **Decomposition.** 600 seats at 20 percent trial-to-paid needs 3,000 trials. 3,000 trials at 4 percent visitor-to-trial needs 75,000 visits in 90 days. **Channel table, with bases.** - Email list, 6,000 addresses at 35 percent cumulative unique click over three sends: 2,100 visits. - Communities and the founder's own audience, measured from previous posts: 3,500 visits. - Launch aggregator, mid-range outcome: 6,000 visits. - Press, one mention: 1,500 visits. - Search, long-tail only inside 90 days: 1,200 visits. - Paid, £6,000 at £1.80 per click: 3,333 visits. - Partners and affiliates, two committed: 2,000 visits. **Total: 19,633 visits against 75,000 required.** The plan produces roughly 26 percent of the goal, which is about 157 paying seats. **Testing the escapes.** Closing the gap with money means buying 55,367 more visits at £1.80, which is roughly £99,700, against a budget of £6,000. Closing it with conversion means moving visitor-to-trial from 4 percent to 15 percent, which nobody does. Closing it with the aggregator means assuming a top-of-day placement, which is a hope, not a basis. **Verdict: the goal is wrong by about four times, and it is the goal that changes.** Restate as 160 paying seats in 90 days, with 600 as the twelve-month target. Move the surplus attention to the two channels that compound but pay outside the launch window, search content and the affiliate programme, and set the affiliate payout at 44 days against the stated 30-day refund window plus settlement. The launch itself is unchanged in shape. What changes is that nobody spends the following quarter explaining a miss that was arithmetic all along. ## Failure modes **Hope marketing.** The goal is a multiple, such as tripling revenue, and the tactics beneath it carry no quantities. Recognisable from the outside because no post-launch analysis can attribute the miss to a line, so the conversation becomes about effort and morale. **The unsummed plan.** Every channel is named and reasonable, and the expected contributions add to a third of the goal. Nobody has ever added the column. This survives review because each individual row looks fine. **Late press.** Reporters are contacted in launch week, cold, along with everyone else launching that week. The tell is a pitch that opens by explaining who you are, which is a pitch to a stranger and reads exactly like one. **Email-list validation.** Four thousand addresses are treated as demand. Launch converts 1 percent of them and the team concludes the price is wrong, when what actually happened is that nobody had ever been asked for money. **Affiliate payout shorter than the refund window.** Commission is paid on day 30 for sales refunded on day 45. Visible from the outside as a reconciliation problem three months in, and as an awkward conversation with your best affiliates. **Single-platform reviews.** Thirty reviews in one place and none anywhere else. Buyers read this as managed rather than earned, and the credibility goal is technically met while producing no credibility. **Split-testing before demand exists.** Weeks are spent testing button colours and headline variants on a page nobody wants to be on. The test is valid and the finding is worthless, because the variable that matters is whether the offer is wanted at all. **The plan that ends on launch day.** No affiliate programme, no review sequence, no second campaign, no follow-up to the people who arrived and did not buy. The spike decays in four days and the quarter is spent wondering where everyone went. ## What this skill does not do - It does not measure conversion rates. Every rate in the decomposition is one you supply or one it marks as an assumption, and an invented rate produces a confident plan that cannot be evaluated afterwards. - It does not evaluate the product, the price or the positioning. A plan that sums perfectly can still be a schedule for launching something nobody wants. - It does not know your real lead times for store review, legal sign-off, translation or print, all of which vary by organisation and country and any of which can be the item that actually sets your date. - It cannot run the plan. There are no owners, no calendar and no dependency recalculation in a document, and after the first slip the dates are wrong until a person fixes them. - It does not do public relations. It schedules the offset between first contact and the ask, and it cannot write the pitch, know the writer, or make the story interesting. - Platform policies inside it move, particularly around review solicitation and affiliate disclosure. The dates given are 2024 to 2026 and every one of them should be checked against current policy before you rely on it.
These skills all ask your assistant to check things against your actual codebase, your actual schema, your actual design system. Locul keeps that context current on its own, from the files you already have, on your machine. Mac and Windows, free to start.