Setting Up and Optimizing Mobile App Campaigns: A Practical Guide

This isn't a platform walkthrough — you can find those in Meta's, Google's, and TikTok's own documentation. What I want to share is the logic behind campaign planning: the reasoning that drives setup decisions, budget choices, and how to read early data before you've spent too much money learning the wrong lessons.

These are the patterns and considerations I use when setting up and optimizing mobile UA campaigns in practice.

Starting New Campaigns: Where to Begin

If you're launching campaigns on a new app or a fresh ad account, start with install optimization. This is the most efficient way to help the platform's algorithm begin learning who your users are.

Install campaigns help the algorithm understand which types of users are most likely to download your app. As those users install and engage, the system collects in-app event data — signups, trial starts, subscriptions, purchases. Once enough events accumulate, you switch to optimizing for those higher-value actions.

Jumping straight to in-app event optimization without sufficient data usually leads to underdelivery or wasteful spending. The algorithm doesn't have enough context to find the right users. Meta recommends at least 50 optimization events per campaign per week (ideally 10–15 per day) before making that switch.

So the progression is: launch install campaigns, collect event data, then shift to in-app event optimization once you have enough signal.

Optimization Goals and Targeting Logic

Your optimization goal directly affects who sees your ad and what you pay.

Install optimization means broader audience delivery. You'll typically see lower CPMs, higher CTRs, and lower CPIs. This is best when your goal is volume and early learning.

In-app event optimization means narrower audience delivery — the platform is trying to find users who will take a specific action, not just install. You'll see higher CPMs, lower CTRs, and more expensive CPIs. But if you're acquiring users who actually convert, the higher cost per install is worth it.

On targeting: unless your product serves a very specific niche, avoid overly detailed targeting (by interest, behavior, or demographics). Detailed targeting usually drives up CPMs and limits delivery. Broad targeting lets the algorithm learn faster and find relevant users at lower cost.

Here's the key insight: your creative does the targeting for you. What your ad looks like, what it says, and how it's presented naturally attracts the right users without needing manual audience filters.

Budget Planning

Budget decisions should follow your campaign goal. Without enough budget, you don't give the system enough data to learn, and campaigns drift without direction.

For install optimization, aim for roughly 50 installs per day per ad set. For in-app event optimization, target 10–15 events per day per campaign.

If your budget can't support that daily volume, focus on fewer campaigns and ad sets rather than spreading budget too thin. It's better to fund one strong test than several weak ones that never exit the learning phase.

At the start, don't use bid caps. Begin with lowest cost bidding to understand your natural CPI and CPA. Once you know your numbers, you can experiment with target bids or minimum ROAS strategies.

Understanding the Learning Phase

When you launch a new campaign — especially on Meta — it enters a learning phase that lasts up to 7 days. During this period, the system is testing different delivery combinations to find the best performance. You need at least 50 optimization events for the campaign to exit learning. Results during this period are often volatile and unreliable.

Once learning is complete, I look at the last 2–3 days (today plus two full days) to track recent trends. If your app shows weekly seasonality — users engaging more on weekends, for example — daily-level analysis can be misleading. In those cases, weekly windows give a more realistic picture of how campaigns are actually performing.

Spotting Issues Early

You don't need to wait a full week to see if something's broken — especially on Android, where data arrives in real time.

Here's what I watch for: Low CTR combined with high CPC usually means weak creative or messaging that doesn't resonate with the audience. If your CPC is approaching your target CPI, your actual CPI will almost certainly be too high — the math doesn't leave room for the install conversion step. No installs or events after 24 hours on Android means something is fundamentally off, likely a creative or targeting problem. On iOS with SKAN, expect data delays — wait 3–5 days before drawing any real conclusions.

These early signals help you quickly pause underperformers and reallocate toward elements showing promise, before you've burned through significant budget.

Making Campaign Adjustments

Campaign changes should be made carefully, especially during the learning phase. Frequent or aggressive changes reset the algorithm's learning and make performance harder to read.

Keep changes to once every 24–48 hours. Limit budget or bid adjustments to 20–30% at a time. Too many changes too quickly will destabilize campaigns that might have been working fine with a little patience.

If performance looks unclear, wait for a full 7-day view before making any big decisions. Reacting to day-to-day noise is one of the most common and expensive mistakes in campaign management.

The Bigger Picture

Running paid UA campaigns is part logic, part pattern recognition, and part structured experimentation. The more you understand how the system works and why you're making specific decisions, the better your chances of running campaigns that actually scale — even if your budget is limited at the start.

Start focused, test methodically, and let the data guide your next move.