Knowledge Base

How to Compare Your Downloads and Revenue with Estimates in Competitor Intelligence

When getting started with Competitor Intelligence, it’s good to get your bearing and understand how estimates in your specific corner of the app market match to actual downloads and revenue.

Appfigures uses sophisticated models and millions of data points of training, to estimate how many downloads iOS and Android apps get all around the world. We do that while maintaining a strict privacy policy where private data (that's what you see in your own reports) is only ever used to help calibrate our estimation models and is never used for anything else.

Our models are calibrated to produce estimates with a very low error rate, but as with all estimates, they fluctuate based on various things—from the day of the week to how popular the app is.

We measure error rates using the industry-standard MAPE, or Median Absolute Percent Error, and have an error rate between 5% (best case) and 25% (worst case).

Such low error rates mean that our estimates will get very close to real-world numbers, close enough to answer important questions to help you grow your downloads. And with our privacy-first approach, will never compromise your private data in the process.

If they don’t, beyond the 5% - 25% error range, there are a few other things to keep in mind when comparing:

  1. Estimates don’t cover all countries. In some cases, we can't provide an accurate estimate for an app in a specific country. This can happen because the app isn't popular enough in the country, our estimation models aren't able to meet the low error rate threshold we have set, or because we don't cover the country altogether.

    We're actively improving our estimation models and extending their reach, so they can cover more apps in every country and also so they span more countries, while providing accurate data you can rely on.

    When comparing your actual downloads and revenue, it's best to compare them at the country level. This way, you can see how estimates across different countries match, instead of comparing totals.

    ? Check out the full list of countries we currently support.

  2. Estimates don’t cover the full date range. Our estimation models see apps that have some downloads, where some depends on the country and category the app is in. It could be as low as one download, or as high as 100. When an app doesn't reach that threshold, it won't have an estimate for the day in that country.

    We're working hard to have our estimation models count every download, but it's always recommended that you look at the trend in countries where your numbers are too far off and look for gaps.

  3. Downloads for paid apps aren’t estimated. If your app is a paid app, at this time, we won't have an estimate for it. We've spent a lot of time calibrating our estimation models for free apps and apps with in-app purchases so that we can get very accurate estimates. Very soon, we'll be doing the same for paid apps.

  4. Revenue estimates don't include the store's fee. Revenue estimates are after Apple's and Google 30% (or 15% for subscriptions) fee, so they aren't what the customer spent but rather what the developer earned.

    We do this on purpose so you can compare your revenue to competitors and don't have to think about gross revenue and the store's fee.

  5. For Apple: Estimates are for iPhone downloads only. For iOS apps, our estimation models are currently estimating downloads on iPhones and not on iPads. This means that apps that are available for the iPad will show less estimated downloads. It also means iPadOS apps will not have estimates.

    We're working on training our estimation models to handle iPad downloads and will be bringing those into the Competitors report soon + show a breakdown by device.

    This is not the case for Android apps, where our estimates include both handheld devices and tablets.

We'd love to hear how your data aligns with the estimates, and if there are specific countries you'd like to see us cover, so please Get in touch.