If you don’t know how users find and install your app, you can’t have an effective user acquisition (UA) strategy. The attribution models you pick determine which marketing channels get credit for an install, which directly controls where your budget goes. Get it wrong, and you’re misreading your campaign data and burning ad spend. With so many models out there, picking the right one is a serious challenge.
Key Takeaways
- Stop relying only on last-click. Use a multi-touch attribution model like Linear or Time Decay to see the whole user journey.
- Put a quarterly review on your calendar to check if your attribution model still lines up with real user behavior data from your analytics platform.
- Pipe your Mobile Measurement Partner (MMP) data into your CRM and in-app analytics to connect acquisition cost to post-install user value.
- Run incrementality tests to prove a channel’s real impact, which moves you from just seeing correlation to proving causation in your UA.
1. Define Your User Journey Stages and Key Conversion Events
You can’t pick a model until you’ve mapped the user’s path from first seeing your ad to becoming a paying customer. The install is just one stop. You have to think about post-install events like their first purchase, starting a subscription, or using a key feature. A key event for a game could be finishing the tutorial or hitting level 10. For an e-commerce app, it’s probably the first purchase inside of 7 days. Get your marketing and product people in a room with a whiteboard to hash this out, it’s far more effective for clarifying these touchpoints than messing with a spreadsheet.
Pro Tip: User journeys are messy and almost never linear. People see an ad on one channel, click a different one later, and might see three more before they finally install. Your journey map has to reflect that chaos.
2. Understand Core Attribution Model Types
Every attribution model hands out credit in its own way, which will completely change how you view your campaign results. Here are the most common ones you’ll see:
- Last Click (or Last Touch): This gives 100% of the conversion credit to the final marketing touchpoint before the install. It’s simple and common for direct response campaigns. If someone clicks a Google App Campaign ad then installs, Google gets all the credit. Simple.
- First Click (or First Touch): The reverse of last click. This model gives 100% credit to the very first touchpoint that introduced a user to your app, making it useful for seeing which channels are driving initial awareness.
- Linear: This model splits credit evenly across every touchpoint in the path. If a user saw a Facebook ad, clicked a search ad, then installed from an influencer post, all three touchpoints get 33.3% of the credit.
- Time Decay: This model gives more credit to touchpoints closer to the conversion. A touchpoint from yesterday gets more weight than one from a week ago, often following a half-life where a touchpoint 7 days prior gets half the credit of one from 1 day prior.
- Position-Based (U-Shaped): This model gives the first and last touchpoints the most credit (usually 40% each) and splits the remaining 20% among all the touchpoints in the middle. It values both the introduction and the final push.
- Data-Driven Attribution (DDA): You’ll find this in platforms like Google Ads and Meta Ads Manager. DDA uses machine learning to look at your actual conversion paths and assign credit based on how much a touchpoint really influenced the outcome. It’s the most advanced model because it’s tailored to your specific account data. It’s not surprising that a 2024 eMarketer report found enterprise marketers are adopting DDA 18% more year-on-year. They see it as more accurate.
Common Mistake: Just using Last Click for everything. It’s easy, but it consistently starves your awareness channels of the credit they deserve for starting the user journey, causing you to under-invest in top-of-funnel work.
3. Configure Your Mobile Measurement Partner (MMP) Settings
Your MMP, whether it’s AppsFlyer, Adjust, or Singular, is your command center for mobile attribution. It’s where you set your core logic, including attribution windows. When you log into your MMP dashboard and go to “App Settings” or “Attribution Settings,” you’ll see these key options:
- Click-Through Lookback Window: This sets the time limit for a click to get credit for an install. Common settings are 7 or 30 days. If you set it to 7 days, a click that’s 8 days old won’t get credit for an install today.
- View-Through Lookback Window: This does the same for an ad impression (a view without a click). This window is much shorter, usually 24 hours or less, because view-through can easily over-credit passive views.
- Re-engagement Window: This applies to existing users who come back after being inactive. It dictates how long a click can re-attribute a lapsed user to a new campaign.
- Attribution Priority (for multi-touch): Some MMPs let you set tie-breaker rules. For example, you can tell the system that a click always wins over a view, or a paid touchpoint always wins over an organic one.
When you’re starting out with UA, a 7-day click-through window and a 24-hour view-through window is a solid baseline. This setup gives users a reasonable amount of time to make a decision after a click without giving too much unearned credit to simple ad views. Of course, these are just defaults, you’ll likely need to tweak them once you have data on how long your specific user journey takes.
Pro Tip: Turn on your MMP’s fraud prevention suite immediately. Bogus installs and clicks will wreck your attribution data and cause you to make some terrible budget calls. Keep an eye on your install validation rates every single week.
4. Implement Multi-Touch Attribution for Deeper Insights
Your MMP’s last-touch report is just the starting point. To get the whole story, you need to go deeper by connecting MMP data to other tools or using the advanced reports your MMP already has. Most modern MMPs have multi-touch reporting. In AppsFlyer, for example, you can go check out “Cohort Reports” or “Path to Install” to see the chain of events, and Adjust offers similar views in its “Funnels” section.
If you really want to get your hands dirty, export the raw attribution logs from your MMP. This firehose of data includes timestamps for every click, impression, and install, which you can then dump into a data warehouse like Google BigQuery or Amazon Redshift. Once it’s there, you can run your own SQL queries to build custom models. Want a simple linear model? Just write a query that finds all touchpoints for a user and splits the credit evenly. This kind of custom work lets you build a model that perfectly matches your business logic, like giving more weight to a channel you know has a bigger impact.
Common Mistake: Your analysis and bidding strategies are out of sync. It’s a classic problem: your MMP is reporting last-click, but your internal analysis is based on a linear model. This creates a huge gap between where you *think* credit should be going and what your ad platforms are actually optimizing for, wasting money in the process.
| Feature | Last Click | Linear | Data-Driven Attribution (DDA) |
|---|---|---|---|
| Credit Assignment | 100% to final touchpoint | Equal across all touchpoints | Machine learning based |
| Simplicity | ✓ Simple to understand | ✓ Moderately simple | ✗ Complex, platform-specific |
| Sophistication | ✗ Low | ✓ Medium | ✓ High, leverages account data |
| Awareness Channel Value | ✗ Often undervalues | ✓ Distributes credit | ✓ Accounts for influence |
| Enterprise Adoption Increase (2024) | ✗ Not specified | ✗ Not specified | ✓ 18% year-on-year |
| Common Pitfall Avoided | ✗ Prone to under-investment | ✓ Provides balanced view | ✓ Reduces misinterpretation |
5. Integrate with Ad Platforms for Data-Driven Attribution
The big ad platforms like Google Ads and Meta Ads Manager have their own Data-Driven Attribution (DDA) models. Their machine learning algorithms chew through all the conversion paths on their platform to assign fractional credit based on what actually drove the conversion. This is completely different from what your MMP does, since the MMP’s job is to be an independent referee and catch fraud.
To enable DDA in Google Ads:
- Go to “Tools and Settings” > “Measurement” > “Attribution” > “Attribution Models.”
- Select “Data-driven” for your conversion actions. Be aware that DDA won’t even be an option until you have enough data which is typically around 3,000 ad interactions and 300 conversions in a 30-day period.
For Meta Ads Manager:
- Navigate to “Events Manager.”
- Select your pixel or app.
- Go to “Settings” and find the “Attribution Settings” section. Here you can pick your window (like 7-day click, 1-day view). Meta doesn’t call it “Data-Driven” like Google does, but its ad delivery algorithms are definitely using a similar data-driven method to hit your goals based on the window you set.
You have to remember that a platform’s DDA model is blind to anything happening outside its walls. Google’s DDA optimizes within the Google universe and has no clue what touchpoints a user had on Meta or TikTok. So while these tools are great for optimizing campaigns on a specific platform, they are not a substitute for the well-rounded, cross-channel view you get from your MMP.
6. Conduct Incrementality Testing
Attribution shows you correlation, which ads came before an install. It doesn’t prove causation. To figure out if an ad actually *caused* someone to install, you need incrementality testing. These are basically A/B tests where you show ads to one group and hold them back from a similar control group, often separated by geography:
- Control Group: Users in this region see no ads from a specific campaign (e.g., a new TikTok campaign).
- Test Group: Users in this region are exposed to the new TikTok campaign.
When you compare the install rates between the two groups, the difference is the true, incremental value your TikTok campaign created. This is how you prove causation. Specialized platforms like LiftMetrix or Measured can run these experiments for you. I’ve had clients run these tests and find out that a channel getting tons of last-click credit was actually generating almost no incremental installs, a clear sign it was time to move that money somewhere else.
Pro Tip: Run your first incrementality tests on your biggest ad spend channels or any new channel you’re trying out. The results will probably surprise you and give you a clear reason to shift your budget.
7. Regularly Review and Adapt Your Attribution Strategy
App marketing changes constantly, new ad formats, different user habits, and endless privacy updates like SKAdNetwork and Privacy Sandbox. The model that worked for you last quarter might already be outdated. You need to schedule a quarterly review of your entire attribution setup, from lookback windows to reporting dashboards. Your job during this review is to find gaps between what your MMP says and what your in-app analytics show about user LTV from those channels. For example, if a channel is sending you high-value users but gets almost no last-click credit, you’ve got a problem. It’s probably time to switch to a model that credits earlier touchpoints. You have to be ready to run tests on different models to see how they affect your actual campaign optimizations. If your strategy isn’t evolving, you’re leaving money on the table.
Picking an attribution model isn’t a “set it and forget it” task. It’s a constant cycle of analysis and testing. If you put in the work to map user journeys, configure your MMP correctly, use multi-touch reporting, and run incrementality tests, you’ll stop guessing and start making decisions that actually lead to sustainable app growth.
Main difference: last-click vs. data-driven attribution?
Last-click gives 100% of the credit to the final touchpoint before a conversion. It’s simple but often wrong. Data-driven attribution (DDA) uses machine learning to look at all the touchpoints and assign partial credit to each one based on how much it actually contributed, which gives you a much more accurate result.
How SKAdNetwork on iOS changes attribution:
SKAdNetwork gives you privacy-safe, aggregated install data on iOS, but it strips out user-level identifiers. This lack of granular data makes it difficult to run complex multi-touch models directly. As a workaround, marketers combine SKAdNetwork data with other methods like probabilistic modeling and incrementality testing to estimate campaign performance.
Can I use different models for different campaigns?
Yes, and you probably should. A brand awareness campaign makes more sense with a first-touch or linear model, while a direct-response campaign might still work with last-click. The key is to maintain one primary, consistent reporting view for comparing overall performance so you don’t create total confusion for your team.
What’s a good starting lookback window?
A 7-day lookback window for click-through attribution is a solid starting point. It’s a good balance between giving users time to convert and maintaining accuracy. For view-through attribution, stick to a 24-hour window to avoid giving credit to impressions that had almost no real impact on the user’s decision.
Why connect MMP data with other analytics?
Your MMP tells you where installs came from, but your in-app analytics tell you what users do *after* they install (like making purchases or subscribing). Connecting them is the only way to see the full lifetime value (LTV) of users from each channel, letting you optimize for acquiring profitable users, not just cheap installs.
