Reviews

What people actually changed after the studio

These notes refer to specific modules from Proxy Servicehub programmes. They are not a rating widget. One of them includes a reservation, because the work is uneven when your release train is slow.

★★★★☆

“Consent-aware flags in week three finally matched our reject button. Before the studio the button only hid the banner.”

Nicha Sawat · Android engineer

I came for the PDPA Analytics seminar and stayed for the dictionary critique. The retention-window homework used our own event names, which made legal review shorter than the last vendor DPIA.

Anan · Chiang Mai

“Cookieless Funnel Architecture was denser than I expected. The server-side join lab is the piece I still open on Mondays.”

Client in grocery delivery
Person working on a laptop with notebooks beside the keyboard

Case note · payments app

Cutting 41 event properties without losing activation

A Bangkok payments team entered Privacy-First Mobile Telemetry with a 90-property Android payload and a dashboard that nobody in support trusted. During the inventory week they discovered 19 properties were copies of device fields already available in crash logs, and 12 more were never queried.

The on-device aggregation module was the turning point: they rolled session length and screen class on the device, then shipped daily counts. Activation (first successful transfer inside seven days) stayed within two points of the old identified curve. What disappeared were city-level heatmaps the growth lead had requested but never used in a launch review.

Limitation they still live with: the studio sampling plan assumed a steadier install rate than a promotions-heavy calendar. They now recompute thresholds before large campaigns. Faculty flagged that gap in week six; the team wishes they had brought traffic seasonality on day one.

Product team working together at a long table with laptops

Case note · education app

First-party funnels after dropping a third-party pixel

An SEA education product used a web-to-app pixel that failed PDPA review. Two PMs and one backend engineer took Guided Bench together. They replaced the pixel with a first-party click identifier that expires in 24 hours and is hashed before it reaches the warehouse.

Attribution for paid web traffic became coarser: they report campaign family, not creative. Enrolment conversion is still visible at the week grain. The privacy officer signed the new purpose map in one sitting because every property had a deletion clock written next to it.

K. Preecha, head of product: “The Mentored Residency critique of our deletion job caught a silent retry that would have re-ingested old rows. That hour was worth more than the recorded Open Studio lessons we watched beforehand.”

“Office hours on hashing vs sampling saved us from shipping both and doubling the SDK. We picked sampling for cohorts and left hashing for opt-in research exports.”

Yui Tanaka · data engineer · enrolled in Guided Bench

Platform review · Privacy-First Mobile Telemetry

★★★★☆

Clear artefacts, slow release assumption. If your org cannot ship an SDK build in under two weeks, budget extra time around the on-device aggregation module. Faculty were frank about that. I would enrol a release engineer next time, not only analytics.