Guide
Every frame below is the real component running on demo data for Pulseboard, a B2B analytics startup with a week-two retention problem. Follow the story from the brief to the roadmap, then open the tool.
The loop
Every tool reads the same brief and hands its output to the next step. Write the product down once, then move work left to right. Chat sits across the loop and pushes things between steps.
Write the brief once. Talk it through.
Find pains and hypotheses.
Rank what is worth building.
Prove the numbers hold.
Plan and deliver it.
01 · frame
One brief. Every tool reads it.
Who uses it, what they finish, what exists, what you refuse.
11/11 fields
Brief health
Sharp product and pricing story. The gap is between activation and the paywall: the brief says week-two retention is the constraint, but nothing in the funnel brings a team back after the first ranked list.
Day one, and again whenever a fact about the product changes. If you keep re-explaining who the user is, the brief is not done.
The founder picks “Head of Product” and “Series A SaaS”, drops the landing link, adds four funnel screenshots. Brief health flags the gap between activation and the day-7 paywall and sends her to Problems.
02 · frame
A product partner who read the brief.
I read the Pulseboard brief. Strong ICP and pricing; the weak spot is week-two retention — trials start, few teams come back.
Before we pick a fix: do you know why week two is empty? Is it that the second standup never happens, or that the board has nothing new to score?
Then that is the first move — 40 answers is enough to cluster. Run them through Problems, then we decide between a Slack nudge before the second standup and a “re-score last week” prompt on the board.
Thinking out loud, deciding what to do next, or turning a vague worry into a concrete task. Chat opens the conversation with what it sees in the brief.
Chat spots that the constraint is pre-payment retention and that 40 exit answers exist. It proposes clustering them first and specs a Slack nudge as a backlog task.
03 · discover
Hypotheses for a stated pain.
Six hypotheses for the week-two problem, constrained by the brief.
1 liked · taste is learned from what you keep
Open the workspace to a diff: which scores moved, which items shipped. Gives the second session a reason to exist.
Risks
Counter-bet
Ship the Slack digest first — it reaches people who never reopen the tab.
Thursday evening: top three ranked items and one open question, posted to the team channel.
Seed the board so onboarding step 2 is never blank. Score five imported items in the first session.
Two standups inside the trial instead of one. Cheap to test, but hides the retention problem rather than fixing it.
One-click ranked plan for stakeholders. Makes the output visible outside the tool.
Everyone scores blind, then reveal. Turns the ranking into a meeting ritual instead of a PM chore.
The problem is clear and the solution space is open — or the backlog is empty and the team is arguing from taste.
From “teams score once and vanish” the tool proposes a Monday diff view, a Slack digest, Jira import, and a longer trial. Challenge flags the longer trial as hiding the problem.
04 · discover
Raw feedback → ranked pain clusters.
Teams score once, close the tab and forget. No trigger between session one and the paywall.
What to do about it
Onboarding ends on an empty board. PMs with a Jira backlog do not want to retype it.
What to do about it
There is a pile of reviews, interviews or exit surveys and nobody can say what they add up to.
Nineteen of forty mention “never opened it again”. The blank onboarding board is second. Stakeholder visibility is real but small.
05 · decide
Score it, then stop arguing.
Several candidates, one sprint. The team is ranking by who spoke last.
Team scoring mode has the highest impact but effort 5 and confidence 2 push it to the bottom. The Slack digest tops the list on effort alone.
06 · model
CAC, LTV, payback — live.
Workable but fragile — fix the leaks
Move the inputs
ARPA $80/mo at 82% gross margin. The full tool models tiers, plans, rebill curves and three scenarios.
Before scaling a channel, changing price, or promising a payback number to anyone.
LTV:CAC looks fine at 3.9x, but 5% churn stretches payback to 14 months. The recommendations point back at retention — the same bet the rest of the loop made.
07 · model
Go / no-go, not p-value theatre.
Bayesian · 14 days · 2,140 trials
Trials with the Thursday digest came back for a second standup more often, and the posterior clears the threshold. Ship, then watch week-two return for two more cycles.
P(variant beats control)
97.5%
Week-two return: 31% → 35%
What the numbers say
Move the variant
Same Beta–Binomial engine as the tool. Peeking-safe, so you can add data and re-run as the test goes.
Before a funnel experiment to size it, and after to decide. Also when someone wants to stop a test early.
Week-two return climbs from 31% to 44% with the Thursday digest. Probability of beating control clears 97% and holds, so the test ends early with a clear call.
08 · ship
Work, with the decision attached.
Monday “what moved” view
Onboarding step 2 renders blank on Safari
Slack digest before standup
Interview 5 lapsed trials about week two
Something is agreed and needs to become a spec the team can pick up.
The RICE score, the cluster quote and the chat thread all sit behind one card. Six months later nobody asks why it was built.
09 · ship
Lanes × months, with dependencies.
Week-two retention above 40%
Sequencing a quarter, or showing stakeholders when things land without opening the board.
Digest this month, the Monday view next, Jira import over two months. Team scoring mode waits until the retention review milestone.
The trial includes all six tools and the full feature set. After seven days, Pro is $29.99/mo or $299/yr.