What is campaign optimization? A guide for app growth teams

Nov 22, 2026|5 Min
User acquisition
Game optimization

What is campaign optimization?

Campaign optimization is the ongoing process of improving a marketing campaign's performance based on the results it's generating. The goal is always the same regardless of channel or method: spend budget more efficiently over time by learning what works and doing more of it. In mobile user acquisition, that means using install, in-app, and revenue data to teach machine learning (ML) models what a valuable user looks like.

This guide focuses on paid user acquisition (UA) for mobile apps, where the process is driven less by manual tweaks and more by machine learning models that predict which users are worth bidding on.

Campaign optimization vs. one-time setup

Optimization is a continuous loop, not a launch task: measure how users respond, learn what predicts value, adjust targeting and bids, and repeat. A campaign that isn't optimized after week one hasn't failed; it just hasn't started the loop yet. The loop improves as it gets more and better data.

How privacy changes affect optimization signals?

Privacy changes across platforms and regions have made it harder to know exactly who converted and why. Frameworks such as Apple's App Tracking Transparency (ATT) give users a choice about whether an app may track them. According to AppsFlyer, the ATT opt-in rate has climbed to roughly 50% of users globally*, about 10% higher than when ATT launched.

That is a meaningful, growing pool of consented, addressable signal, but it also means a significant number of iOS users still opt out, leaving optimization engines to work with a mix of deterministic and aggregated or modeled data.

Making the most of the signals that remain, by giving the model data it can actually use, has become increasingly important for UA teams.

Consent comes in more than one form. ATT opt-in governs access to the iOS device identifier. On Android, users can reset or opt out of the advertising ID, and consent rules in regions such as the EU apply on any platform. Separately, developers can choose to share consented first-party data, which this guide covers below. Higher opt-in on either can add usable signal, but they are different mechanisms.

For opted-out users, ad networks lose the link between an ad and a specific install, so attribution becomes aggregated. The setup checklist later in this guide is how you keep events, attribution, and first-party context available despite this.

Which metrics matter?

Before diving into how optimization works, it helps to be precise about what it's optimizing for. Different goals call for different metrics.

Key metrics in campaign optimization

Which campaign metrics matter?

Metric
What it measures
Revenue generated per dollar of ad spend
When to optimize for it

Campaigns focused on monetization and profitability

What it measures

Cost to acquire an install

When to optimize for it

Early-funnel growth or volume-driven campaigns

What it measures

How often an ad prompts a click

When to optimize for it

Creative testing and top-of-funnel engagement

What it measures

Share of users who return after install (for example, on day 1, day 7, or day 30)

When to optimize for it

Long-term value and churn-sensitive campaigns

What it measures

Total value a user generates over their lifetime

When to optimize for it

Campaigns optimizing for durable, high-value users

What it measures

Cost to acquire a user who completes a chosen post-install event, such as reaching a level

When to optimize for it

Campaigns that need engaged users

It’s common to monitor early signals like conversion rate and CTR to steer budget while a campaign is still learning, then shift weight toward ROAS and retention as it matures.

How campaign optimization works: signals in, performance out

An optimization engine is a prediction machine, and its ceiling is set by the signals you feed it. Better, more complete data can strengthen optimization inputs over time. A data signal, in advertising terms, is any piece of event or behavioral information the model can learn from (an install, purchase, session, or click). It tells the engine something about which users are valuable. Two campaigns can run on identical optimization engines and get very different results in part because one is feeding it a richer, more relevant signal than the other.

Most explanations of campaign optimization skip straight to tactics: raise this bid, test that creative, shift this budget. Those levers matter, but they operate downstream of the signals. If the model doesn't know which users actually purchased or stuck around, bid changes are less informed. The signal is the input; performance is the output.

These signal categories are complementary rather than competing. Attribution and conversion data, in-app behavioral signals, purchase and monetization events, and even gameplay data each capture a different slice of the same user journey.

The signals that drive optimization

No two signals teach a model the same thing. Here's how the major categories map to what they improve:

How each data signal enriches the model

Enabling data completeness

Signal

Install events

What it teaches the model
Which channels, creatives, and audiences drive volume

Outcome it can improve

CPI, top-of-funnel efficiency
Post-install and in-app events
What it teaches the model
What users do after the install (onboarding, key actions)

Outcome it can improve

Early engagement, funnel drop-off
Purchase and revenue events
What it teaches the model

Which users generate revenue, such as through in-app purchases or ad revenue

Outcome it can improve

ROAS, value-based bidding
Engagement and retention signals
What it teaches the model
Whether users stick around and return

Outcome it can improve

Retention, long-term LTV
Privacy-preserving attribution signals (e.g. SKAdNetwork, or SKAN, on iOS)
What it teaches the model

Aggregated, privacy-safe signal when deterministic data is unavailable

Outcome it can improve

Optimization when deterministic signals are limited

What it teaches the model
Permissioned behavioral and player data from the app itself

Outcome it can improve

Richer optimization signals for model learning

A checklist for activating your data signals in Unity Vector

The table below shows the main data sources that can help power Unity Vector, the AI foundation of Unity Ads. Each signal helps the optimization system fill a specific gap. Putting them together creates a richer data foundation for Unity Vector to learn from.

Data checklist for Unity Vector

Recommended data sources to enrich Unity Vector's optimization capabilities

MMP integration

Configuration

Organic indicator

Recommended setup

Turned on
How to connect
Enable organic install tracking in your MMP’s Unity partner configuration settings

Configuration

Unattributed events

Recommended setup

Send unattributed install and post-install events (including organic)
How to connect
In your MMP’s Unity partner settings, enable sending of unattributed and organic events

Configuration

Postback window

Recommended setup

Set to Lifetime

How to connect
In your MMP’s Unity partner settings, set the event postback window to Lifetime

Configuration

Event sharing

Recommended setup

Send installs, sessions, purchases, and ad revenue
How to connect

In your MMP’s Unity partner settings, enable sharing of install, session, purchase, and ad revenue events

Unity Dashboard

Configuration

Google Analytics

Recommended setup

Google Analytics linking
How to connect

Configure in the Unity Dashboard

Configuration

Android license key

Recommended setup

Provide license key (Android-specific requirement)

How to connect
Configure license key from your Unity Dashboard

Configuration

Developer Data Framework

Recommended setup

Unity Ads permission and consent layer
How to connect

Configure in the Unity Dashboard

Unity SDKs and Editor

Configuration

Unity Ads SDK

Recommended setup

In place and updated to the latest version

How to connect

Install via CocoaPods or manually for iOS

Use the following directions for Android

Configuration

Unity IAP SDK

Recommended setup

Updated to the latest version
How to connect

Configuration

Diagnostics

Recommended setup

Data collection/source
How to connect
Configured in Unity Editor

Before sharing data with Unity, make sure your privacy notices and any required user consents cover that sharing.

The more data sources you connect, the more information you’re giving to systems like Unity Vector, which can use the information to help make optimization decisions that inform better outcomes over time. For example, when you add MMP data such as installs and purchase events, the engine can get a better understanding of your acquisition, conversion, and player value signals. At the same time, this data alone doesn’t tell the engine much behavioral context about how players engage with your game. Providing an additional source like Google Analytics, can add that context for optimization.

Campaign optimization for mobile app user acquisition

Optimizing for user acquisition in mobile apps differs from web or PPC in a few structural ways. The chain runs from install through post-install events all the way to in-app purchase and retention, and much of that chain is mediated by mobile measurement partner (MMP) postbacks rather than a single platform's native pixel. Apps also generate a high frequency of behavioral signals, sessions, level completions, in-app actions, that a website typically doesn't produce at the same density. That density is an advantage if it's captured and fed to the optimization engine, and a missed opportunity if it isn't.

Mobile UA is where the signal-to-outcome relationship can be especially direct and measurable, because the game itself generates so much of the signal. This is the layer Unity Vector is built for since it can translate installs, in-app events, and revenue signals into UA campaign decisions.

Common campaign optimization mistakes

Some of these are familiar, others are specific to how signals get mishandled:

  • Optimizing to the wrong event. Optimizing for installs when your business depends on revenue trains the model to find any user, not necessarily the most valuable ones.
  • Truncating postback windows. Short windows can look efficient in reporting while quietly starving the model of the late-funnel data that retention and LTV optimization depend on.
  • Resetting during the learning phase. Judging a campaign's success before the model has had time to learn, then restarting it, resets the clock without fixing the underlying problem.
  • Chasing vanity metrics. Optimizing for CTR or install volume because they're easy to see, rather than the outcome the business actually needs.
  • Treating it as a one-time setup. As covered above, optimization works best as a continuous

Set up the right data foundation

Getting campaign optimization right starts with connecting and activating the signals your optimization already depends on. Learn how you can set Unity Vector up for success.

Explore the Developer Data Framework

Frequently asked questions

*Source: Appsflyer, AppsFlyer Data Shows Market Adaptation to Privacy-First Ecosystem Four Years After ATT, https://www.appsflyer.com/company/newsroom/pr/post-att-growth/, 2025