Most mobile UA teams think the algorithm is something you set and surrender to.
Itai Kafri doesn't. He engineers the signal.
Itai leads product growth at Voyantis, a predictive LTV and signal engineering platform for mobile apps. He has been in mobile user acquisition since before Meta introduced oCPM, back when advertisers controlled every bid manually. Since then, he has spent years figuring out what actually replaced that control.
The answer is not the model. It is the signal strategy built on top of it.
In this episode, Itai breaks down signal engineering for mobile UA: how to send predictive LTV signals to automated channels like Meta Advantage Plus, Google UAC, and TikTok in a way that actually moves performance. Why timing matters more than accuracy. Why the channel needs rank, not dollar precision. And the three questions every growth team should answer before investing in a predictive LTV media strategy at all.
If you run paid user acquisition for a mobile app and you are thinking about PLTV, value-based bidding, or tROAS optimization, this episode is the most practical breakdown of signal engineering available.
Video Chapters:
00:00 Introduction 00:36 Pre-2014 vs post-2014: how automation changed acquisition 02:42 The seven-day conversion window ceiling 04:11 Why marketers and CFOs define a "good user" differently 07:00 Predictions, confidence, and the timing tradeoff 11:11 What actually feeds a prediction: the marathon analogy 16:26 Why proxies like "level 5" break down 17:42 How much data you need to build a model 20:15 Value-based bidding and the accuracy misconception 23:12 Managing outliers without breaking the strategy 25:11 Real client story: Google Search Partners going haywire 28:11 Whales vs outliers, and why hyper-whales backfire 32:03 Signal engineering vs RTB 34:05 User-level data, privacy, and reading the user mix back 38:57 How Black Friday and seasonality hit signaling 40:48 Why casual gaming is tricky for PLTV 41:16 Three questions to ask before investing in a prediction model 45:32 Where to find Itai and Voyantis
Topics covered:
- Signal engineering for mobile user acquisition - Predictive LTV and value-based bidding on Meta, Google, and TikTok - Confidence thresholds, latency trade-offs, and signal timing - How automated channels use signals: Advantage Plus, UAC, PMax - When D7 ROAS is already your best signal and PLTV will not move the needle
Podden och tillhörande omslagsbild på den här sidan tillhör
Shamanth Rao. Innehållet i podden är skapat av Shamanth Rao och inte av,
eller tillsammans med, Poddtoppen.
