Baltasound studio · App Analytics

Read the runtime.
Not the slide deck.

Most product charts describe a wish. We teach teams to sit with the event stream until retention, paywalls, and onboarding stop contradicting each other.

View the programmes →

Analytics charts on a dark monitor, with a time series and a conversion panel visible — From a working board, not a pitch: weekly active readers versus claimed “engagement”.

63: product squads coached since 2019, mostly UK and nearby European teams.
14 days: typical length of an instrumentation audit before a cohort programme begins.
3,412: named events reviewed in client taxonomies last year — duplicates included, on purpose.
7 tutors: all still shipping analytics inside live products; none of us teach from a retired playbook.

In the current season

Three rooms, not a catalogue

We keep the roster short. Each programme is built around a failure we see in App Analytics every month.

A wall of small dashboard tiles in muted colours

Signal Cartography

The flagship. Eight weeks on event grammar, funnel honesty, and a map your warehouse can actually keep.

Two colleagues reviewing product metrics on a laptop in a quiet office

Retention Forensics

A four-week clinic for teams whose D1 looks healthy and whose D30 has quietly collapsed.

A workshop table with notebooks and a projected funnel sketch

Attribution without Theatre

How paid, organic, and in-app prompts share credit without inventing a last-click morality play.

Laptop on a wooden desk showing a spreadsheet of cohort columns

How we teach

A number is a sentence. We ask who wrote it.

App Analytics is often treated as a layer you install. In the studio it is closer to editing: you name what happened, you refuse synonyms that hide drop-off, and you leave a paper trail a sceptical finance partner can follow.

Students arrive with vendor fluency and leave with a taxonomy they can defend in a Tuesday standup. We work in Amplitude, Mixpanel, GA4, and warehouse SQL, but the craft is the same in each: decide what an event is allowed to mean, then stop adding events that mean nothing.

The Signal Atlas is the method we use on every engagement. Capture, name, cohere, decide. If a dashboard cannot point to a decision, it is interior decoration.

The studio, in full →

From recent rooms

Voices, not slogans

Signal Cartography forced us to retire screen_viewed as a success metric. Painful for a week. After that the paywall funnel finally matched what billing already knew.

Priya N., product analytics lead · Leeds

I came for Mixpanel shortcuts. I got an argument about property dictionaries instead. That argument is now in our onboarding doc, which is more useful than the shortcuts.

Client in grocery delivery

★★★★☆

The taxonomy week ran slower than I wanted — we spent two sessions on viewed_paywall versus paywall_impression. I still think one of those sessions could have been an email. I also know our board pack no longer has two different “activation” numbers, which is why I would send a new hire.

Tom Hesketh · staff engineer, Edinburgh

Longer notes and two case studies →

Journal

Recent writing

Short pieces on App Analytics habits we keep seeing in United Kingdom product orgs.

Your DAU line is a weekend story in disguise Why a Monday spike is not a product miracle, and how to stop presenting it as one.
Event names that survive a reorg A naming rule that still works when the growth team is renamed twice in a year.
When not to instrument The cheapest analytics decision is often a missing event.

What you leave with

A vocabulary the warehouse can keep Properties with owners, not a graveyard of “we might need this”.
Funnels that survive a finance question Drop-off stated as a behaviour, not as a missing pixel.
A habit of refusing vanity If a chart cannot change next sprint, it leaves the wall.

Enquire about a seat

Read the runtime.Not the slide deck.