I used module four’s layer labels in Amplitude the same week. The argument about session length moved to an appendix and stayed there.
Jonah Pell · product analytics, Bristol
Flagship programme
Experiment Readout Studio
A seven-module studio for people who already ship app experiments and need the Friday review to end in a decision rather than a new chart request.
Learning outcomes
By the last clinic you should be able to open a memo with a ship/hold/iterate/retire sentence, document assignment health before any lift, keep primary, guardrail, and diagnostic layers visually separate, bound heterogeneity so slices cannot rewrite the headline, and present a two-page leave-behind a director can finish on a train.
You will not become a platform engineer, a research scientist, or a facilitator of experiment design sprints. Those crafts sit outside this syllabus on purpose.
Modules
-
The decision the test was hired to make
Reconstruct the original product question from tickets and Slack, then write a decision sentence that does not mention p-values. If the sentence cannot be written, we pause the readout.
-
Pre-register the readout, not only the metric
Lock memo length, audience, and the rule for late-arriving diagnostics. Participants bring one past write-up that drifted in the last five minutes of a meeting.
-
Assignment health and instrumentation
Sample-ratio checks, exposure windows, and what to do when the log and the product analytics tool disagree. The output is a paragraph that either clears the test or stops the story.
-
Primary, guardrail, diagnostic
Label the three bands in your current tool. Practice moving a tempting secondary number into diagnostics without burying it.
-
Heterogeneity without fishing
Pre-declared slices versus exploratory cuts. How to show a regional difference in a GB marketplace app without promoting it to a claim.
-
The memo
Two pages, a recommended action, a kill condition, and a spoken caveat. Tutors mark structure, not whether the experiment “won”.
-
Live readout clinic
A recorded twelve-minute presentation to tutors who were not in your original thread. You leave with annotated notes, not a certificate graphic.
Instructor
Helen Marlowe taught experiment ops inside two UK consumer apps before opening the Great Casterton studio. She still marks memos line by line. Guest statisticians appear in module three; they do not take over the clinic.
Informational fee
Studio Seat is listed at £1,180 including live modules and six months in the alumni circle. VAT is added for UK bookings where due. There is no checkout on this site. Compare Bench Pass and Readout Circle or request a seat.
Questions we actually hear
Do I need a statistics degree?
No. You need access to a completed or in-flight app experiment and permission to share an anonymised readout. We assume comfort with your platform’s default charts.
What will this programme not cover?
It will not teach you to design experiments from scratch, to implement an experimentation platform, or to write production assignment code. If that is the gap, the studio will feel oddly narrow, because it is. We would rather name that limit here than discover it in week four.
Can a whole product trio attend?
Yes, via Readout Circle on the fees page. Studio Seat is priced per person and the clinic time is shared with the public cohort.
From recent studios
The clinic is blunt. Helen stopped me twice for reading confidence intervals before the decision sentence. I still think module two over-indexes on memo length, but the leave-behind is now what our director forwards.
First name withheld · Edinburgh
★★★★☆
Brought a null onboarding test. Left with language that did not apologise for the null. That was the whole point for me.
Client in travel super-apps