Voices

Notes from people who had to take a score back to work

These are not campaign quotes. They refer to specific Cloudpacketbase modules. One of them is deliberately mixed. Read them as field reports, then look at the longer case studies below.

Instrumented Session Quality made us stop merging crash-free sessions with task completion. Finance still wants one number; we now send a one-pager with three columns and a footnote they actually read.

Araya Ph., product operations, Bangkok
★★★★☆

Latency Budgets for Interactive Products finally gave our platform team language that design would accept. The percentile workshop was dense; I would not recommend it as a first Cloudpacketbase course if you have never touched telemetry.

Verified learner · Singapore-based client in ride-hailing

I came for Cohort Replay Analysis because our library was 40,000 clips and nobody watched them. The sampling log template is now mandatory before anyone books a “highlights” meeting.

Mikael, Kuala Lumpur

Voice of Session was the first time our CX transcripts and our event stream sat in the same document. It was uncomfortable. That was the point.

D. Sorasak, research manager

The Bangkok Field Lab day on our own checkout traces was more useful than the lectures, though the room was warm and the Wi-Fi on Soi Charansnidvongs dipped twice. Bring a dongle.

Anonymous client in grocery delivery
★★★★★

Observer Desk suited me while on parental leave. I worked through the score charter alone. I still want a live critique before we ship internally.

Siriwan J. · Hat Yai

Case notes

Two longer accounts

A wallet app that retired its “session health” ring

A Bangkok consumer-finance squad arrived with a circular gauge that mixed crash rate, time-on-task, and a survey snippet. After Instrumented Session Quality they split the gauge into three traces and wrote a charter that defined a “clean session” as: no fatal error, primary task completed or explicitly abandoned, and no blocked PII in the replay. Six weeks later they still did not have a single replacement number. They did have a weekly quality review that engineering attended without being summoned.

What did not go smoothly: their capture vendor could not export rage-click timestamps with the fidelity the charter assumed. They documented the gap instead of inventing a workaround score. That limitation paragraph is now copied, with permission, into our flagship FAQ as a teaching example.

Sampling a logistics replay library without theatre

A regional logistics platform had reviewers watching whatever clips the tool surfaced as “frustrated.” Cohort Replay Analysis taught them a stratified sample by corridor, device class, and time of day. The first month’s report was boring: most “frustration” tags were map-tile delays on a specific Android webview, not driver error. They cancelled a planned training programme and opened a platform ticket instead.

Reservation from their analytics manager: the workshop assumes you can query an event warehouse. Teams still living entirely inside a replay UI will need extra setup before the method pays off.

Browse the programmes they mention