Ledger 07 · Oldtown Of Aigas

Place this week’s sessions beside a cohort that actually looks like yours.

Softmatrixhub is a small studio for Benchmarking App Usage Trends. We teach product and data pairs to pick peers, set windows, and write a ledger a sceptical director can follow.

Open the flagship atlas

Illustrative weekday session minutes · media apps, GB

Peer set P50 P75 Delta vs you
News, 0.4–1.1m MAU 6.8 11.2 +1.4
Catch-up video, ad-funded 18.1 31.0 −4.6
Utility, daily habit 3.2 5.1 +0.3
Your last complete week 5.4 9.7 —

Current programmes

Four desks, one shared rule: no comparison without a peer definition.

Laptop with charts used while building a usage cohort atlas

Flagship · 6 weeks

Usage Cohort Atlas

Build a versioned set of peer cohorts, weekday baselines, and a one-page board ledger. Rowan Hale critiques live.

Syllabus
People reviewing a session-window worksheet on a laptop

Workshop · 2 days

Session Window Lab

Stop treating a 30-minute timeout as a law of nature. You leave with three documented windows and the trade-offs written down.

See listing
Notebook beside a retention table printout

Studio · 4 weeks

Retention Ledger

Week-n tables that survive a release train. Includes a module on when not to publish a number at all.

See listing

From recent atlases

What people actually used after week six

The Usage Cohort Atlas made us drop a “category average” we had been quoting in investor slides. The peer set we built is smaller, and the product conversation is quieter.

Amelia K. · product lead, Edinburgh

Session Window Lab was useful, though the homework assumes you can query raw events without a two-week ticket. We borrowed a replica and still finished a day late.

Client in grocery delivery

Studio facts, not rounded folklore

How the work usually looks from the Highland desk

18atlas cohorts since 2023
73product–data pairs taught
£1,140median programme fee last year
11 wkstypical wait for the next atlas
Workshop table with printed usage tables and laptops

Method in brief

A peer set first. Charts second.

Most “industry benchmarks” collapse news apps with utilities and call the average a trend. We start by writing who is allowed in the set, which weekdays count, and what a session even is.

Only then do we plot deltas. The custom page on Benchmarking App Usage Trends walks through the same order we use in class.

Read the method page

What the studio insists on

  • Named peers Every comparison names the apps or the inclusion rules. Anonymous “category medians” do not survive critique.
  • Windows in writing Session, week, and “active” are defined in a short RFC before anyone pastes a screenshot into Slack.
  • A ledger, not a score You leave with a one-pager: your P50, the peer band, the delta, and a sentence on what would falsify the story.

Field notes

Recent writing

Code editor beside a coffee cup during a session-window write-up

12 March 2026

Session windows that lie

A 30-minute timeout can invent “engagement” on a podcast app and hide it on a banking one. Same number, different physics.

All notes

If the next board pack still quotes a nameless average

Write to the studio with the category you think you sit in and one number you currently defend. We reply within three working days, usually with a question rather than a quote.

hello@softmatrixhub.digital