Barbara Galiza is a mobile growth expert who's worked across HER, Microsoft, WeTransfer, and Mollie. She joins the Price Power Podcast to discuss conversion events, attribution, and how subscription apps can optimize their ad performance through better signal engineering.
The 3 Key Events for Ad Optimization
When optimizing subscriptions on Meta and Google, focus on three key events:
- Sign-up: The first PII (personally identifiable information) touchpoint. This is your best match rate with ad platforms.
- Trial Start: Users who initiated a trial. This is a strong purchase intent signal.
- Subscription Purchase: The actual conversion. Only send this when the transaction is confirmed.
Don't overthink it. These three events give you a clear funnel and enable the ad platforms to optimize effectively.
Speed Matters: Fast Events Win
When you send an event to Meta or Google, speed impacts match rate. Events that fire quickly (within milliseconds of user action) are matched more effectively to the ad platform's user database.
If your subscription purchase event fires 2-3 seconds after the user completes payment (due to server delays or SDK latency), you're losing match rate. The ideal is sub-second event delivery.
This is why client-side events (fired from your app immediately) often outperform server-side events (sent after server confirmation). Yes, you get some false positives, but the improved match rate often makes up for it.
The Dual Event Pattern
Here's a pro tip: send both a client-side event (immediate, optimistic) and a server-side event (confirmed, accurate).
The client-side event gives you fast match rate. The server-side event provides accurate attribution. Both serve a purpose in the ad platform's optimization algorithm.
Value Scoring and Predictive Ranking
Beyond the event itself, the value you pass to the ad platform matters enormously. This is where signal engineering comes in.
Simple value scoring: send the user's predicted LTV or a point value (e.g., $50 for an annual subscriber, $5 for a monthly). This tells Meta and Google not just that a conversion happened, but how valuable it is.
Predictive value ranking goes deeper: instead of static values, you use machine learning to predict each user's actual LTV based on their behavior before conversion. This gives ad platforms the richest signal possible for optimization.
Unified Master Event Pattern
At scale, apps often send dozens of events to ad platforms. A cleaner approach is the unified master event pattern: send a single comprehensive event with different parameters for each conversion type.
For example, you might send a "Purchase" event that includes these parameters: eventType (trial_start, subscription, etc.), value, currency, userData, and other contextual data. This reduces event noise and makes it easier for ad platforms to optimize.
PII for Match Rate
Personally identifiable information (email, phone number) dramatically improves match rate, especially for smaller apps with limited event volume.
Include hashed PII with your events when possible. Make sure to hash it (SHA-256 or MD5) before sending. This helps the ad platform match your users to their user graph more accurately.
Incrementality vs. Attribution
Attribution tells you which ad got clicked. Incrementality tells you whether the ad actually caused the conversion. These are different things.
Attribution can be gamed or misleading (e.g., last-click attribution gives all credit to the final touchpoint). Incrementality requires controlled testing: running test and control groups to see if an ad truly moves the needle.
For subscription apps optimizing ad spend, think about both. Attribution is useful for diagnostics. Incrementality is what actually matters for ROI.
Advice for Smaller Apps
If you're under $1M MRR or getting fewer than 100 conversions per day:
- Focus on the three core events (sign-up, trial, purchase)
- Make sure events fire quickly and include hashed PII
- Use simple value scoring (not complex ML models)
- Test one ad platform thoroughly before splitting budget
- Wait for sufficient sample size before declaring winners
As you scale, you can layer on more sophisticated techniques like predictive value ranking and incrementality testing. But start simple, nail the basics, and iterate from there.
Resources
- LinkedIn: Barbara Galiza
- Get help with your attribution: FixMyTracking.com
- Check out past episodes here: PricePowerPodcast.com
