Reviews
Twenty-nine written notes from 2025. Not all of them flattering.
These are from operators who sat a Packet Servicehub lab. We keep attribution mixed on purpose: some people can be named, some can only share a city, some prefer industry only. Ratings below are individual, not a blended score we print on every page.
Fill Rate Diagnostics caught a timeout we had been calling “soft demand in VN.” The worksheet for SEA hours was the piece I still use on Mondays.
Araya Phong, monetization manager · Bangkok
★★★★★
Privacy-Safe Measurement for Ads is the only session where legal and product stayed in the room without checking out. Dry, specific, useful.
Platform review · in-app course rating
I liked Cohort LTV for Ad-Supported Apps more than I expected. It still will not replace our BI team, and Nalinee said that out loud, which I appreciated.
Minh · Ho Chi Minh City
Client in hyper-casual publishing. Auction vs Waterfall Decisioning stopped us mid-migration. We were about to celebrate bidding before the brief had a single sentence about match rate.
Anonymous client in hyper-casual publishing
★★★☆☆
Mediation Stack Literacy is accurate but slow if you already live in three consoles. I wanted the Revenue Signal Lab first; the catalog order was on me.
J. Patel, UA lead
Office hours from Bang Kapi ran late twice during our Signal Bench cycle. The critique of the brief was still the most useful hour of the month.
Fern · Phuket
Case study
A puzzle studio that had been averaging Thailand away
A mid-size puzzle publisher joined House Residency with a single global eCPM target. Product thought fill was “fine.” UA thought creative was the lever. Finance wanted a smoother curve.
In week two of Revenue Signal Lab methods, the residency log split Thailand from the rest of SEA. Fill was not fine after 21:00 local; it was a timeout cluster on rewarded. The brief that left the house did not recommend a floor change. It recommended a latency budget and a different hour window for UA in that geo.
Thirty days later they wrote back: rewarded completion in Thailand after 21:00 had stopped looking random. They still did not hit the old blended eCPM target, and they stopped using that target.
House Residency · 2025 · name withheld by request
Case study
A bidding migration that needed a quieter ending
An anonymous client in midcore RPG sat Auction vs Waterfall Decisioning while half the stack was already on bidding. The remaining waterfall instances were the only place older geos still filled. The team had a launch date for “full bidding” printed on a slide.
The lab did not stop the migration. It changed the sentence they were willing to say in the product review: bidding would cover new geos first; leftover waterfall would stay until match rate in two countries stopped looking like a cliff. That sentence was less exciting than the slide. It was also the one finance accepted.
Auction vs Waterfall Decisioning · Signal Bench seat
