Week 3 hurt. I had grown fond of a “context_json” blob. Celia made me defend every key as if it were a column we would pay to store. We still disagree about screen names, but the dictionary is now short enough to print.
Flagship course · App Analytics
Instrumentation Studio
A six-week room for people who own the event stream and are tired of discovering, in a board pack, that “signup_success” fires twice.
Who it is for
Product managers who inherited a tracking plan, analysts who are asked to “just add one more event”, and engineering leads who want the analytics request to arrive as a contract rather than a Slack paragraph. You should already ship an app. You need not be a SQL specialist.
We work in App Analytics as it is practised: vendor UIs, warehouse extracts, and the social difficulty of telling marketing that a campaign parameter is not a product event.
What you will leave with
- A written event dictionary for one real surface of your product, with owners and expiry dates.
- A property allow-list and a one-page rule for names (object, action, tense).
- A retirement protocol: how an event is deprecated, aliased, and finally deleted.
- Two critique letters from tutors on your schema, not a certificate with a watermark.
Weeks
Modules
- 01
The map before the warehouse
User jobs, surfaces, and the difference between a log and a product event. You draft a one-page “what we will never count”.
- 02
Names that survive a reorg
Object–action grammar, versioning, and why camelCase arguments waste a Thursday. Critique of anonymised dictionaries from past seasons.
- 03
Properties, and the urge to hoard them
PII, cardinality explosions, and a drill on rejecting “just in case” fields. Includes a short legal note aligned with UK data practice.
- 04
Identity across installs
Anonymous IDs, login merges, and the App Analytics mess created by shared devices. Harun’s guest seminar sits here.
- 05
Quality as a weekly ritual
Volume monitors, schema tests, and a “broken event” playbook you can hand to on-call. No vendor lock-in in the exercises.
- 06
Retirement and the Metric Council
Killing numbers in public. You present a deprecation to a mock council of peers. The polite version of a fight.
Fee, stated without a cart
£1,890 covers six live weeks, Saturday studios, and written critique. Materials are included. Travel to Collynie is not. This page does not take payment; write to the desk if you want a seat.
A Reading Circle seat (notes + monthly salon only) is listed on Fees if six weeks is more than you can give.
From people who sat the Studio
★★★★☆
Clear on taxonomy. Less useful if your app is a thick native client with a chaotic release train — they said so in week 1. I still used the retirement protocol in Q1.
Questions we actually get
Do you teach a specific analytics vendor?
No. Examples wander through common App Analytics tools, but the assessed work is the dictionary and the memos. If you need click-training in one UI, we will say so and decline the seat rather than pretend.
What is a real limitation of this course?
We do not make you a data engineer. Warehouse modelling, dbt, and streaming pipelines sit outside the six weeks on purpose. Several alumni told us they wished week 5 went deeper on automated tests; we keep it to a playbook you can hand to engineering, not a CI course. If your events never reach a store you can query, the Studio will feel abstract until that pipe exists.
Can a whole team join?
Two seats from the same organisation are common. Beyond that, the Private Map residency on the Fees page is the cleaner shape — one dictionary, many voices, one tutor.
Is there a refund if the season has started?
The Refund page sets the windows. After week 2, work already critiqued is not unwound like a shop return.
Seat enquiry
Ask about the next Studio
Tell us what you ship and where the event map hurts. We reply within two working days. No automated urgency, no fake remaining seats.