Sub Club by RevenueCat · 2025-08-20
Signal Engineering: Strategic Data Filtering for Better Ad Performance — Thomas Petit, Independent Consultant
Consultant Thomas Petit on signal engineering: first make ad-network event totals match internal data within ~5-10%, then send qualified trials (device, age, onboarding answers), then engineered revenue values (predicted month-13 value, amplified spread, API-served values), sent within day one and above ~10 events/day/campaign. Also: use AI for creative analysis, not just production; keep exploring new concepts; app-to-web suits big brands; hybrid monetization only after pricing wins slow.
Play the episode23 tactics
Numbers mentioned
| Subject | Metric | Value | Quote |
|---|---|---|---|
| personThomas Petit, ad spend managed over the past decade | ad spend Ad spend managed with a positive return across hundreds of clients | nine digits approximate | “Over the past decade, Thomas has worked with hundreds of clients and helped manage nine digits in ad spend with a positive return.” |
| benchmarkCreative testing hit rate for paid social | count Roughly 50 creatives produced to find one winner; the winner takes 90%+ of spend | 50 creativesper winning creative approximate | “the stats out there are like roughly, you have to produce 50 creatives to have one that is actually a winner, and this winner is going to take 90+ percent of the ad spend.” |
| benchmarkShare of ad spend going to the single winning creative | share Winner-takes-all spend concentration on the winning creative ('90+ percent') | 90% approximate | “the stats out there are like roughly, you have to produce 50 creatives to have one that is actually a winner, and this winner is going to take 90+ percent of the ad spend.” |
| benchmarkAcceptable gap between conversions an ad network's events manager receives and internal counts | share Discrepancy between events received by Meta Events Manager / Google Goals and internal analytics: ~5% is fine, more than 10% is a problem | 5–10% approximate | “If I'm seeing more than 10% discrepancy here, it's going to be a problem.” |
| aggregateAd accounts the guest audits | share Typical discrepancy he finds between events reported to the network and internal counts in broken setups | 30–50% approximate | “But in so many cases, I'm seeing 30% or 50%, and this this is going to be a a major problem that I need to fix before seeing anything else.” |
| App the guest ran growth for in 2017 | conversion rate Trial -> paid for Google Ads users, vs ~40% on organic and Meta | 15% illustrative | “we would have like, let's say, 40% trial-to-paid conversion on organic, on Meta, like, I mean, small differences, and then we'd have like 15% on Google.” |
| marketNewest Android flagships (Google Pixel, latest Samsung Galaxy) as a share of the Android market | share These devices convert about as well as iPhones; the rest of Android converts terribly | 5–10% approximate | “Like, they only represent 5% to 10% of the market, and the rest have a terrible conversion rate” |
| Thomas Petit client that scaled from 0 to eight figures on a couple of platforms | revenue Growth trajectory while scaling on a couple of platforms, before performance broke | 0 to eight digits in a couple of years approximate | “went from 0 to 8 digits in a couple of years, like very beautiful trajectory.” |
| benchmarkMinimum optimization events per campaign for ad networks to optimize well | count Guest's rule of thumb from the commonly cited 50-100 events per week; below 10/day/campaign things go wrong | 10 eventsper campaign /day approximate | “the number that's floating around is somewhere between 50 and 100 per week, so I simplify in my head and say, below 10 event per day per campaign, things are going to go a little bit wrong.” |
| benchmarkOptimization event threshold per campaign as commonly cited | count | 50–100 eventsper campaign /week approximate | “the number that's floating around is somewhere between 50 and 100 per week” |
| benchmarkRevenue threshold below which app-to-web checkout is not worth trying (App Store Small Business Program, 15% fee) | revenue Below $1M (15% store fee) the guest thinks almost nobody should try app-to-web unless already web-native | USD 1M/year approximate | “Below 1 million, I mean, when you're paying already 15%, I don't think like anybody should really try it, unless you're very web-based from the beginning.” |
| benchmarkMaximum delay before sending an optimization event to ad networks | time to milestone Signals sent after the first 24 hours are nearly useless for network optimization | 24 hours estimate | “I'm entirely convinced that whatever comes after 24-hour is is useless for the platform to optimize towards” |