Est. Bangkok · Lat Phrao studio

App analytics for product desks

Forecasts that respect how people actually open the app.

Logiceventbase teaches usage forecasting from event data: weekday seasonality, messy Firebase names, and the quiet hours when Thai users actually return. You leave with a model you can defend in a roadmap meeting.

See the flagship cohort
Laptop showing an analytics dashboard with trend lines
63 days Median time from first lab to a forecast a PM will actually use
11Graduating Bangkok & remote cohorts
2,180Practitioners who finished a live forecast
71%Still running the same model family six months later
8 weeksFlagship Retention Horizon desk

Flagship

Retention Horizon sits at the centre of the studio.

Printed charts and a laptop on a wooden desk

8-week live cohort

Retention Horizon: forecasting usage from event streams

Build a 28-day usage forecast from raw events, not from a vanity DAU export. You learn priors, holiday effects in Thailand, and how to explain a miss without hiding behind a confidence interval.

Informational fee: THB 24,900. No checkout on this site — enquire if a seat is open.

Open the syllabus

What you take home

Three habits that keep a forecast honest.

Name the event before the model

Most “broken forecasts” are unnamed screens and duplicated purchase events. We start by rebuilding a slim event dictionary you can actually maintain.

Seasonality before sophistication

Bangkok late-night usage and weekend grocery apps do not look like US SaaS. You fit weekday and festival effects before you reach for a library.

A miss you can narrate

Every lab ends with a one-page note: what moved, what you ignored, and which product lever would change the next four weeks.

Studio method

You work on a live series, not a toy CSV.

Bring a sample of your own events if you can. If you cannot, we issue a masked Thai super-app extract with the same ugliness: missing days, renamed screens, double-fired logins.

  1. Observe the week

    Plot raw daily actives against calendar, not against a smooth trend. Mark Songkran, payday Fridays, and store-listing spikes.

  2. Choose a horizon you will defend

    Seven, fourteen, or twenty-eight days. Longer than that and we treat the number as a scenario, not a forecast.

  3. Ship the memo, not the notebook

    A PM should be able to read your output without opening a cell. If they cannot, the lab is not finished.

From recent desks

People who sat through the labs.

All reviews

The Retention Horizon weekday lab caught a Saturday dip we had blamed on “seasonality” for a year. It was a silent crash in the checkout event after a store release. We still argue about priors in stand-up, which I did not love at first.

Mira K., growth PM · Bangkok fintech
★★★★☆

Session Depth & Feature Adoption Models is quieter than the flagship. I wanted more live critique in week two. The adoption curve worksheet, though, is still on my desktop.

Client in grocery delivery · rated on the studio desk

Journal

Notes from the Lat Phrao whiteboard.

Open the journal
Line chart printed on paper beside a pen

Forecasting12 min

Why DAU forecasts fail when you ignore weekday seasonality

A flat seven-day average will look confident until Friday payday hits. Here is the check we run before any model is allowed near a roadmap.

Read the note
Bangkok city skyline at dusk

Thailand9 min

Bangkok product teams and the late-night usage spike

If your “global” model assumes evenings peak at 19:00, you will miss the 22:40 commute home on BTS nights.

Read the note

Bring a messy export. Leave with a horizon.

Tell us which series you care about. We will say honestly whether Retention Horizon, a shorter desk, or a private Atelier is the better fit.

Write to the studio