Android App Conversion Setup - What Am I Doing Wrong?

davidsearch

Running two campaigns for a utility app with a single lifetime subscription ($1 max). Indonesia: conversion campaign optimising for a closest event (event_x) rather than 'Purchase' because of low purchase data - launched a day ago, got around 15 installs, zero event_x triggers. India: install campaign, over 250 installs over two weeks, event_x fires but zero 'Purchase' events. App is freemium with limited features, only about 10 purchases came from Reddit, not ads.

Intuition says low purchase volume shouldn't be the sole factor with 100+ installs and an active paywall. But I'm burning budget with no subscription conversions. Should I switch Indonesia's event to 'Purchase' despite the data sparsity? Or shift India to event_x? I know these markets have lower IAP culture, but zero conversions feels off.

What's the smarter play here - keep optimising for the closest event until we have enough purchase signals, or go straight to 'Purchase' even with a tiny conversion count? Also, any tactical tips for a newbie scaling with limited conversion data? Cheers.

weekendhustle

Check tracking first. Most "0 conversions" problems are actually broken event imports. Confirm event_x and Purchase are both firing in Firebase → Google Ads, and test a purchase yourself to verify.

$1 is probably too low. Sounds backwards, but the algo needs a value signal to find paying users vs freeloaders. Even at IN/ID install costs ($0.30–$0.80), $1 leaves you no margin. Try $2.99–$4.99 still cheap locally, 3–5x more headroom.

Switch India to tCPA on event_x. You have enough volume now (250+ installs, event_x firing) to move off install optimization. Don't jump to Purchase until you have 30+ purchases in 30 days UAC won't optimize properly below that.

For Indonesia, the strategy is right but make sure event_x is actually predictive of purchase. If it's "app_open," it's basically install optimization in disguise. You want something like "viewed_paywall" or "initiated_checkout."

Also: set tCPA at 2–3x target initially, give it 7–14 days to learn without touching it, and remember UAC needs ~10 conversions/day to optimize at all.

Honest take: 250 installs / 0 purchases at $1 cold UAC in India isn't actually unusual. Your Reddit conversions were warm, problem-aware traffic totally different intent. Paid UAC for cheap utility apps in tier-3 markets is brutal even with perfect setup. Might be worth raising price and leaning into organic/content for those geos, saving paid for higher-LTV countries.

wordweaver

Speaking from painful experience: keep trial starts and purchases in separate buckets, or Google Ads will muck up your optimisation something rotten.

ctrjunkie

Whew, tracking conversions for a new Android app with zero data is basically throwing darts in the dark. using event_x as a proxy makes total sense when you've got nothing else to work with.

Around 16 installs in a day isn't bad for the Indonesia geo, but yeah if those installs aren't kicking off any conversion events yet the algorithm's just guessing on purchase optimisation with zero signal. that's a fast way to burn budget.

The smart play early on is to pick a conversion event that actually happens - app open, tutorial completion, that first meaningful action in a utility app. doesn't have to be the final subscription, just something frequent enough to give the algo a clue. once you're getting 15-20 of those consistently, then you can test switching the optimisation target closer to actual purchases. jumping straight to subs with no history is basically lighting money on fire for most new apps.

Keen to hear - how long was your utility app live before you launched these campaigns? Were there any organic installs or user feedback to help decide which early events to track first?

weekendhustle

Honestly, I wouldn't optimise for Purchase yet if there's barely any purchase data. Using an intermediate event first is the right move, but if even event_x isn't firing much, that usually points to a funnel or paywall issue rather than an ads problem.

250 installs with zero purchases doesn't automatically mean the countries are bad. It could be weak onboarding, the paywall shown too early, an unclear value proposition, or low‑intent traffic from install campaigns.

I'd focus on improving activation first. Track things like app open → key action → paywall view → trial or purchase attempt. Also test different paywall copy and screens before scaling ads harder.

Right now you probably need better conversion signals and funnel validation more than more traffic 🙂

ranktracker

Tracking trial starts and completed purchases separately is a real nightmare - subscriptions seriously muddy what Google treats as the primary conversion. I've been caught in that exact trap before, where the optimisation kept chasing the wrong event and everything went sideways fast. For cleaner optimisation, be ruthless about which single event you feed into your main campaign.

emmareach

hit the exact same dead zone with cheap installs in those markets. turned out my event_x was completely worthless. swapped it for a later funnel signal and bumped the price to $3.99 - within a week actual purchases started showing up.

davidsearch

No trial starts here either - freemium with limited features and a one-time lifetime payment. That's actually a tough conversion model for App Campaigns, especially with <20 purchases.

10 purchases sourced organically (Reddit) means your pixel data is lean and likely biased toward self-qualifying users. Optimising for purchase right now will stall the campaign - Google needs ₁₅-20 conversions per week per ad group to exit learning phase. At your volume, you'll get stuck in "limited by conversion volume" hell.

Better play: set up a secondary in-app event that maps to real engagement - something like "completed level 3" or "used premium feature trial". Optimise for that event to drive traffic volume that actually mimics buyer intent. Once that event hits ₅₀/week, layer in a purchase optimisation as a separate campaign or ad group.

Also, check your attribution window. For one-time payments, a 7-day click + 1-day view is fine, but if your users take weeks to decide, you're losing data. Push that to 14-day click, maybe even 30-day if the app has a long consideration cycle.

What's your current bid strategy - tCPA on the purchase event? If so, switch to "maximise conversions" with no target until you hit that volume threshold. Let the algo learn on its own.

marcuthread

Same issue here. Split events into trial_start, high_intent (like hitting the paywall), and purchase - then only optimised for purchase once I had 50+ actual buys. Firebase + GA4 helped me see exactly where people dropped. i ended up leaning on a community monitoring tool after using Sensor Tower and AppFollow to figure out how people talked about similar paywalls and pricing before I even touched the ad setup.

davidsearch

41% is decent, but you can still squeeze CPA efficiency. Screenshots often mismatch the search intent that converts-test a benefit-driven set against your current ones. What's your average CPC? That tells you how much room you've got.

wordweaver

tell me about it. nothing worse than throwing ad spend at an algorithm that's basically blind. If conversions are still thin on the ground, optimising for purchases is just burning cash. stick to higher intent engagement first - let the data pile up before you ask Meta to go shopping

davidsearch

Exactly. The real pain point is separating high-intent users from the noise of low-quality installs. Have you considered dropping install-level conversions and moving to an in-app event for lead form submissions? That single change cut my CPA by about 40% on Android. what attribution model are you using?

wordweaver

Oh honey, urgency filtering is the bare minimum. If you're not prioritising someone screaming "fix my app" over some rando who just tapped an ad by accident, you're basically lighting your budget on fire