Tom from Bristol. Winter intake. The readout practice was the part I did not want and the part I reuse. The warehouse modelling was correctly out of scope, which annoyed my data team for a fortnight.
Flagship programme
Cohort Signal
Twelve weeks to stop averaging the install base. You will write an event contract, defend a cohort window, and practise a two-minute readout the rest of the company can actually interrupt.
At a glance
Who it is for
Product managers, founding analysts, and engineering leads who already have a dashboard and no longer trust the blended retention number. You should be able to ship or request an event. You do not need a statistics degree.
Informational fee · £1,840 · Atelier Circle seat
No payment is taken on this page. Write to the studio to join a listed intake. VAT is added on the invoice where United Kingdom law requires it.
Request a seatYou will be able to
Learning outcomes
- Draft an event contract with owners, properties, and a kill date for deprecated names.
- Define a cohort by first qualifying action rather than by install timestamp alone.
- Choose an activation window and write the exclusion rules in plain English.
- Plot retention without mixing platforms that do not share event names.
- Separate new-user decay from resurrected-user noise.
- Deliver a two-minute spoken readout without opening a full dashboard.
- List what the chart cannot claim, including sampling and missing background events.
- Hand the artefacts to a colleague who was not in the room.
Twelve weeks
Modules
- 01
Paper before pixels
You describe one live app (or a public stand-in) and the meeting where measurement currently collapses. No tools this week.
- 02
Event contracts
Owners, properties, allowed values, and the difference between a user property and an event property when both are tempting.
- 03
Identity and devices
Anonymous to known, shared tablets, and why staff accounts destroy civic-app cohorts.
- 04
First qualifying action
Install is not a cohort. We pick the action that means the product was actually used.
- 05
Activation windows
Twenty-four hours versus seven days, and how a weather app on Saturday is not the same product as on Tuesday.
- 06
Retention without theatre
Classic N-day, unbounded, and range retention. You will be asked to retire at least one pretty line.
- 07
Decay and resurrection
People who leave and people who were never in. Mixing them is how boards get false hope.
- 08
Platform split
iOS and Android synonyms. Funnel Grammar is a sister clinic; here we only split enough to stop lying.
- 09
Board grammar
Two minutes spoken, one page written, one limitation list. Guests from the Signal Room sit in.
- 10
Counter-metrics
What would have to be true for your recommendation to be wrong. We practise saying that out loud.
- 11
Handover
A colleague who missed the course must be able to maintain the contract. You write for them, not for us.
- 12
Public readout
Each student presents. The grade is whether a stranger can repeat the claim. Alumni hours begin after this week.
Tutor
Dr Miriam Pell
Miriam led product measurement in two United Kingdom scale-ups before founding the studio. She still marks the event contracts herself. Visiting engineers join weeks two and eight. She does not teach paid-media attribution and will send you elsewhere if that is the real brief.
From this circle
Reviews
The week we split first-session completers from the rest, the activation debate in Leeds finally had a shape. I still think week nine is too short for people who hate public speaking, but the artefact survived the board pack.
Questions we actually get
FAQ
Does Cohort Signal cover paid-media attribution or SKAN?
No. This is a real limitation, not a humble brag. We stay inside product analytics: events, identity, cohorts, and retention narrative. If your brief is campaign postbacks, we will decline the seat rather than pretend a module exists.
Do I need my employer’s production data?
Preferable, with a written permission note. Individuals may use a public app pattern that Miriam assigns. You will still write a contract as if you owned the product.
Is the fee refundable?
See the refund page for windows and non-refundable items. The syllabus PDF, once issued, is not refundable on its own.
How is this different from Retention Atlas?
Atlas is eight seminars on decay maps for teams who already have a trusted event dictionary. Cohort Signal is the longer literacy, including the dictionary and the spoken argument.