Part 3: Bringing your data together

Once your stack is optimized and running, you need a reliable way to know what's actually happening across it. The problem is that ad monetization data doesn't live in one place: your mediation platform has its dashboards, individual networks have their own, your telemetry system has its data, and your UA sources have theirs. Each gives you a piece of the picture, but none gives you the whole thing.

The solution is a centralized monitoring point that pulls from all of these sources and puts them side by side. It can be a dashboard built with tools like Data Studio or Tableau or you can build something entirely custom, especially in the age of AI. The important thing is that you can see all relevant data points as a whole picture. 

What you're bringing together:

• Mediation platform data  (eCPM, impressions, revenue, fill rates by ad unit, network, country, and platform)

• Your telemetry data (ad lifecycle events, user sessions, engagement metrics, ad opportunity tracking)

• User acquisition data (install costs, user quality metrics, LTV projections)

  
  

This combined view surfaces correlations that individual dashboards won’t show you. However, it does require a proper backend and storage solution to make it work. By using MDN11 you have a lot of telemetry data already available to you, so you can track your ad lifecycles - one can easily get an insight into the journey of an ad from a request to an impression through a unified and simple way. 

Overall, bringing your data together can feel like a big investment, depending on where you are in your monetization journey, but a worthwhile starting point is even just pulling key data into Google Sheets and analyzing it there, tracking your main metrics that we’ll discuss down below.  

Bringing your data together

Metrics worth tracking

Beyond the fundamentals (ARPDAU, eCPM, impressions, fill rate), some more nuanced indicators become useful once your monitoring setup is more mature.

The percentage of DAU that has seen at least one ad tells you how effectively your inventory is actually reaching your users, broken down by format. It's a direct read on how successfully you're monetizing your daily active base.

Impressions per session ratio shows how many impressions you're generating per user session. A rising ratio alongside falling retention is a signal you might be pushing too hard on ad density.

Time to first ad tracks how long after launch a user sees their first ad. If it’s too short, you're serving ads before users have properly engaged with your app. Conversely, if it’s too long, you're leaving money on the table from your most engaged sessions.

Ad opportunity rate measures the percentage of ad opportunities where an ad could have been shown but wasn't, due to no fill, timeouts, failed requests, etc. While fill rate tells you how often a network fills your requests, ad opportunity rate tells you how much of your designed inventory went unused.

Revenue per second of ad time is a slightly more advanced metric. If you're tracking ad duration via telemetry and have access to ILR data, you can calculate the actual return on the time users spend watching ads. This is useful for understanding whether longer ad formats are genuinely worth the user experience cost.

Catching problems early

Active daily monitoring has its limits as you scale. At some point, you need automated alerting for anomaly detection, either built in-house or through a third-party integration.

The kinds of things worth setting alerts for:

• ARPDAU dropping more than 10% against your 7-day average

• Impression volume dropping similarly

• Significant divergence between your telemetry data and ad network reporting

• Ad opportunity rate drops that might indicate something broke in a recent app update

Early detection is the point of all of this. Problems that go unnoticed for days or weeks do real damage, but automated alerts can help with something humans are bad at, namely spotting slow, gradual trends across longer timespans. Small month-on-month drops that wouldn't trigger alarm on any given day can add up to something significant, and a well-configured alerting system will catch those where manual review would miss them.