Readout Craft
Flagship. Six modules on metric trees, peeking, variance reduction, and a readout a staff engineer will actually argue with.
Autumn desk · Beeswing & remote GB
Appstreamgrid is a small school for people who already ship app features and still watch lift evaporate in the next release. We treat A/B Test Analytics for Apps as a writing and measurement craft, not a dashboard tour.
These are not launch quotes. They name the module that actually changed a weekly ritual.
The assignment lecture in Readout Craft made me reopen three iOS tests we had treated as gospel. Two of them had leaked users through a shared paywall. The readout template is now on our wiki; it is slower to fill, which is the point.
I still dislike CUPED. The guardrail module, though, stopped us shipping a checkout change that looked fine on conversion and ugly on refunds.
Short programmes. One flagship with a full module list. Fees on the pricing page are informational; nothing here is a checkout.
Flagship. Six modules on metric trees, peeking, variance reduction, and a readout a staff engineer will actually argue with.
A four-week desk on crash, refund, and latency rails that should never be optional in an A/B plan.
What to do when the weekly active base will not give you a clean 5% lift, including sequential looks you can defend.
A/B Test Analytics for Apps is mostly unglamorous: bucketing bugs, metric definitions that drift after a rename in the event taxonomy, and stakeholders who want a number by Friday. Our teaching starts there.
You will write a pre-analysis plan, name the primary metric before looking, and practice saying “this test is inconclusive” in a room that wanted a ship decision. That sentence is a skill, not a personality trait.
Longer pieces, not recaps of the syllabus.
Why the Friday screenshot of a dashboard is not a stopping rule.
A working tree for trial, paid conversion, and refund without drowning in vanity nodes.