Home / Programmes / Atlas
Usage Cohort Atlas
A six-week live programme for product and data pairs who need a defensible peer set before they quote a trend. Places are capped at eight teams.
What you leave with
A versioned atlas: inclusion rules, three named peers or a rule-based set, weekday and weekend P50/P75 for session minutes, a D7 note that states its window, and a one-page ledger with a falsification sentence.
You do not leave with a warehouse audit, a tracking plan, or a vendor score. If your events are a mess, we will say so in week one and you may need to pause.
- Write inclusion and exclusion rules that a new hire can apply next quarter.
- Separate weekday habit from weekend binge without pretending they share a mean.
- Present a delta without dressing it as a ranking.
- Know when a release train has poisoned the comparison.
Modules
-
Peer physics
Who is allowed in the set. Category labels from app stores are treated as gossip until you write a size band, monetisation model, and geography rule.
-
Windows and “active”
Session timeout, backgrounding, and the difference between opened and used. You document three windows; one becomes the house default.
-
Weekday versus weekend ledgers
Habit products and binge products cannot share an unlabelled weekly average. You split the week before you plot a trend line.
-
Release trains as confounders
A paywall test in week 3 is not “seasonality”. You mark trains on the ledger and, if needed, withhold the number.
-
Sampling and coverage
SDK sampling, logged-out users, and tablet traffic. You write a coverage footnote the board is allowed to see.
-
The one-pager
Public critique. Rowan marks inflated peer sets in red. You revise once. The second draft is the artefact.
Instructor
Rowan Hale
Rowan spent nine years in product analytics for UK media and utility apps, including a stretch where a “category benchmark” deck had to be withdrawn after a board member recognised the unnamed peers. She now runs critiques from the Aigas desk and still refuses to ingest production event streams for students.
Questions people ask before week one
Do you connect to our warehouse?
No. That is a real limitation of this programme. You query your own data (or a replica) and bring aggregates to class. Softmatrixhub never receives raw event tables. If you cannot run a session query by week two, the atlas will stall and we will say so plainly.
Can one person attend without a data partner?
Yes, if they can already pull session minutes and D7 for a defined window. Solo product managers who rely on a locked BI queue usually struggle in week two.
Is the fee refundable?
See the refund notice. Places are limited; the rules are written there rather than negotiated in email after the first critique.
Will you name other students’ apps in class?
No. Peers in examples are either public figures from research notes or synthetic bands. Your atlas stays yours.
From people who sat the atlas
Week three’s weekday split stopped us averaging a Sunday spike into a “growth story”. The ledger we took to the director was two numbers and a caveat, which was the point.
★★★★☆
Clear on peer rules. I still wanted a template SQL file; Rowan would rather you write the window yourself. Fair, slightly annoying if you are on a deadline.