How we plan, run
and learn.
A walkthrough of the operating system that sits underneath Road to Launch: what the pieces are, where each one lives, and what we need from each of you.
We have a plan. We do not yet have the machinery.
Road to Launch says what has to be true by September, and the scorecard tells us every Tuesday how close we are getting. What sits between them is the actual work, and today each of us runs that part our own way. This is the bit we are unifying.
Work lives in heads.
Who is doing what this fortnight is knowable only by asking. Nothing rolls up to “is Reviews engine on track for September?”
What we learn evaporates.
We run a test, we read the result, and six weeks later somebody proposes it again. Nothing catches it.
We settle things twice.
A choice gets made in a meeting and lives in whichever document it was made in. That document gets rewritten each phase, and the reason we chose it goes with it.
None of this is a tooling problem. It is a missing set of shared words and a place to put each one. That is the whole of what follows.
Four layers, each one smaller than the last.
Read it downward and it answers “what am I doing today?”. Read it upward and it answers “and why does that matter in September?”
An initiative is not a sprint.
This is the one word we were quietly using for two different things, and it is the collision most likely to resurface in this room. So we have settled it.
A workstream with an owner and a September outcome.
It spans several sprints. It is the thing the Strategy Map tracks and the thing the room asks “is this on track?” about.
Reviews engine → Trustpilot live and climbing toward 150 reviews.
The prioritised chunk of work we commit to next.
One objective, a committed set of cards, a start and an end date. It is the progress we are committing to make on one or more initiatives over the fortnight.
Get the review request flow live and measure reply rate on the first cohort.
Initiative is the word from the pre-read, so the pre-read wins. Nothing in here uses “initiative” for a two-week thing, and nothing uses “sprint” for a six-week one.
Four places, and only one of them is new to you.
The databases.
Everything with a row and an ID. Under Growth & Marketing, seven databases, four you already use.
The scoreboards.
The L10 scorecard, live and editable by anyone with the link. The team dashboard for the outcome numbers.
The context.
Our warehouse numbers plus competitor reviews, customer interviews, Slack and the business docs. It answers questions, it is never where work is recorded.
The skills.
Ten commands you type into Claude. They do the moving of work between the databases so nobody is hand-filling forms.
The scorecard is already live and already yours. Nothing here replaces a tool you use today.
Four words we are going to use precisely.
Each one gets its own database because each one behaves differently. The test for which is which is the sentence form.
What we concluded.
A claim you can agree or disagree with, rated on how good the evidence is.
Sufferers avoid cold drinks rather than manage the pain.
What customers say.
Their words, kept verbatim, because that is language you can put straight into an ad, an email or the website.
Like biting into an ice cube.
What we chose, and what it rules out.
Not something you agree with, something you comply with or breach.
Not a log of everything, a place to go back to why. We can always reverse one, with a new row that supersedes the old, never a quiet edit.
Anchor at £24.99, discounts earned through volume.
A problem, not a task.
Raised for IDS at the L10. Solving one creates cards, it does not become one, which is what keeps the list short enough to rank.
Google is not showing on Shopping and we do not know why.
The L10, every Tuesday.
90 minutes, same shape every week. Most of it is not status, it is solving.
| 10 min | Scorecard read. Each owner takes us through their number, its RAG and a line of commentary. No solving yet. Anything red or amber drops onto the issues list. |
| 10 min | Card updates by owner. On track or not. Not a walk through the board. |
| The rest | IDS on the issues list. Identify the real problem, discuss it once, solve it with an owner and a date. |
| 2 min | Rate the meeting out of ten. If it was not a ten, say what would have made it one. |
Solving an issue creates cards.
One problem routinely produces three or four actions across three or four owners. If the issue simply became a task, we would lose that, and the list would grow until nobody ranked it.
So the issues list stays a list of problems, short enough that we can always argue about which one is worst. A new skill, s3-ids-issue, files one in the right shape, so raising a problem takes a minute.
Come to Tuesday ready. Scorecard numbers in and issues raised before the meeting, with the list already ranked, so the 90 minutes goes on solving rather than filling in. On the weeks a sprint opens, planning runs straight after, same room, same morning.
14 days, Tuesday to Tuesday.
Built around the days we are all actually on, with two team meetings a week at most.
one
two
Two team sessions a week at most, both on days we are all on. The grill is not a meeting: it is preparation brought to planning, it runs Friday to Monday across the seam, and bringing an unprepared one into the room is what makes planning long.
Ten commands, in the order you would use them.
| Plan | |
| s3-sprint-grill | Async. Reviews the inputs, then grills one question at a time until the focus for the next two weeks is settled. |
| s3-to-sprint | Writes the Sprint page. No second interview, it uses what the grill settled. |
| s3-to-backlog | Creates and scopes the cards the sprint commits to. |
| Do | |
| s3-scope-card | Takes a loose idea to a card good enough to commit to. |
| s3-do-card | Executes one card, across any surface we own. Builds, never launches. |
| s3-ids-issue | Files a problem into Issues in IDS shape, ready to rank before Tuesday. |
| Learn | |
| s3-capture-insight | Turns evidence into the voice bank. One survey becomes two or three insights, not 300. |
| s3-distil-learnings | Turns evidence into beliefs. Adds freely, but asks before it weakens a claim. |
| s3-review-sprint | Committed versus shipped, the estimation delta, what we now believe. |
| Improve | |
| s3-growth-audit | Monthly, across every surface. A ranked report that comes out as proposed cards. |
s3-do-card builds. It never launches.
s3-do-card is the one that does the work, on whatever surface the card names. One rule governs all of them, no exceptions, and no per-card setting anyone can get wrong on the day it would spend money.
Lands on a draft theme, with a preview URL.
Campaigns created paused.
Flows built unpublished.
Codes created disabled.
Every card finishes at To Review, and a person flips the switch.
To Review is a status the Backlog already has, so there is nothing new to learn. On an experiment it means the result is waiting to be read. On a project or BAU card it means the work is built and waiting to be published. Which one it is, is obvious from the card's type.
It applies to the harmless things too, a dashboard, a doc, a Notion tidy-up. Slightly redundant on that half, and worth it for having one rule instead of a boundary anyone has to reason about at speed.
Underneath it will sit a skill per surface, build a landing page on Shopify, build a Meta campaign, write the ad copy. How far each of those goes is still to be scoped, and s3-do-card is the layer above them, so the rule holds whichever one it calls.
Rather more than half of this already exists.
- Backlog, with every card carrying an EX-n
- Sprints, with an objective field already on it
- Learnings, with the evidence ladder
- The L10 scorecard, 42 input KPIs, editable by anyone
- The Brain, answering questions across the warehouse and the docs
- s3-experiment-scoping, which becomes s3-scope-card
- Initiatives, so the pre-read's table becomes a live view
- Issues, so IDS has somewhere to point
- Decisions, so a choice outlives the phase it was made in
- Customer Insight, re-grained to one row per recurring thing customers say
- The ten skills, written into the team plugin
The design is settled and written down: a glossary, eight decision records, and this. The build starts after today.
Four things, starting Thursday.
Claim your initiatives.
All 22 owners are blank in the pre-read on purpose. Come and take the ones you want, and say out loud which ones you are not taking.
Bring problems, not tasks.
If something is off track or in your way, raise it as an issue and bring it to the L10. Do not quietly turn it into a job for yourself.
Say when you learn something.
A number that surprised you, a thing a customer said, a test that did not work. That is the raw material, and it is the part that goes missing.
Put your work on the board.
If you are working on it, it is a card. Not because we are counting, but because it is the only way the Tuesday conversation is about the real list.
First sprint opens
Tuesday 11 August.
Closes Thursday 20 August, with the review. The next one opens Tuesday 25 August. Between now and then, the databases get built and the owners get claimed.