Get Appy: A data analysis of the Global App Economy
By Bryan Pon, Caribou Digital
The global app economy continues to grow rapidly, driven by increasing smartphone adoption, higher-speed wireless networks, and behavior changes in how users engage with digital content.
Apps attract significant attention in part because they represent the first truly global market for digital goods, which can in principle be produced anywhere, distributed at almost no cost, and consumed wherever there is a network connection. Yet the app market, like all markets, is a socially constructed system with policies, architectures, and intrinsic biases that govern participation and outcomes.
This research sought to investigate these dynamics through a descriptive study of supply side participation, value capture, and international trade in the global app economy, spurred by questions such as, Who is successfully making apps? Who is making money, and in what markets? How do the structure and design of the app stores affect value capture and trade?
Where are the developers?
In order to understand who is benefiting most from the structure of the app markets, we conducted a geographical analysis of developer participation worldwide. The results show that despite claims of the digital app economy as a level playing field, the most successful developers are geographically concentrated, primarily in the largest cities of the wealthiest countries, but with important exceptions.
Image: Top 20 countries by total developers.
To better understand any hot spots or app development “clusters” we also located developers at the level of the city. In this analysis, the major cities of East Asia and the San Francisco Bay Area (including Silicon Valley) top the list in absolute number of developer. But measuring on a per capita basis, using the population of each metropolitan area, reveals a few standouts where app development seems highly concentrated: the San Francisco Bay Area is still tops, with 5x the concentration of developers as the average city, followed by Helsinki and then Hong Kong. Other high-performing cities Figure 6 with at least 20 developers were Gurgaon (India), Taipei, Minsk (Belarus), and Tel Aviv. Among the major Western European capitals, London performed best, but Paris and Berlin showed essentially average concentrations of developers.
Visualizing regional trade and value capture
Using a series of Sankey diagrams, we show for each region the flows of apps into the markets and the flows of estimated revenue back out to the country of the developer. These visualizations allow us to easily see several dimensions within the data, including the producer diversity in each market, domestic market share, and total exports. Because the charts are organized primarily geographically, we can also see regional trade (or lack thereof) in a much more direct fashion. Perhaps most importantly, these charts also include the value capture side of the equation, and therefore show not only which countries are making and exporting lots of apps, but which countries are earning lots of revenue.
And by tracing the flows, it becomes apparent when there is a large differential between the two—e.g., few apps but lots of revenue—revealing especially strong or weak performance by producers from a country. For example, in virtually all regions and markets, Finland earns more revenue than expected given the number of apps it produces, which is primarily due to the very high ranking Supercell products. While China is a larger producer and typically earns a concomitant amount of revenue, we can see that in the English-speaking regions it earns relatively little revenue given the number of titles it exports to those markets, indicating that it has more low-ranking apps compared to other regions.
As shown in the example below, the charts are to be read left-to-right. The left column is the origin country of the app developer, with the width of the flows representing the number of apps that were “exported” to the national markets shown in the middle column. The width of the flows on the right side represent estimated financial value being captured by each country, based on the rank of the app and rank of the national market (for details, see section “Estimating value capture”). For better clarity in the diagram, we omit flows of apps or revenue below a certain threshold (~1%), and we set all the national markets to the same size (the nodes in the middle column are not sized to represent the total value of the market). This latter modification sacrifices the ability to visually compare total value from different markets (e.g., U.S. vs. Brazil), but enables us to represent all markets in the same series of charts.
Image: App market share and estimate valued capture in Southeast Asian markets
The app store analysis is built on an original dataset of top-ranked apps and their developers across 37 national markets. For each national market, a snapshot was taken in June 2015 from the App Annie website of the 500 top-ranked apps in both the “Top Grossing” and “Top Downloads” categories, for both the iOS and Android platforms, resulting in 2,000 app records per national market (Ghana and Tanzania have slightly fewer because there weren’t 500 apps in the Top Grossing category). The result was approximately 74,000 records, which corresponded to 21,539 unique apps and 11,644 unique developers. For each developer a manual online search was conducted to identify the city and country location; for larger firms with multiple offices, the headquarters location was selected. This resulted in 8,441 developers with both a location of production and a location of consumption, and enabled us to trace the flows of apps from the country of origin to the country of consumption. For categorization purposes, we use the World Bank county income classifications, though we consider China part of the high-income economy group.
We supplemented this quantitative data with a limited number of developer interviews and survey responses. The outreach was targeted specifically at developers from non-Top 10 markets in order to better understand their perspectives on the opportunities and barriers they face in the app economy. Although we eventually contacted thousands of developers via email, we saw a minimal (~1%) response rate, resulting in only 60 responses to the online survey, and 7 semi-structured interviews conducted over Skype. While the qualitative responses cannot be considered representative of the sample, we do use some quotations in the discussion section to provide context and add the voice of developers.
There are limitations to this analysis. First, the intended scope includes only the most successful developers in the largest markets on the two dominant platforms, and therefore cannot capture all revenue or commercially successful developers. China is especially opaque to this analysis because it has banned Google services, including Google Play, meaning that while China is an enormous market for Android, the vast majority of app downloads and revenue occur via 3rd-party app stores. As a result, the number of developers from China and other countries where Google Play is not dominant (e.g., India) are under-represented in the analysis (though iOS data for China should be valid). Secondly, because we find location data for only 72% of developers, there is a significant margin of error with the geographic analyses. This introduces at least a few potential forms of bias: First, the population of developers that could not be located is evenly distributed across the 37 markets except for China, Japan, and South Korea, which have a higher ratio of un-located developers. Because this is primarily the result of the language barrier, we assume that most of these un-located developers are domestic, and therefore, for these countries, the number of developers is likely underrepresented. Secondly, there are more un-located developers with lower-ranked (less popular) apps, primarily due to smaller, independent developers not having a formal online presence. This skews the results towards larger and more-established firms, though we don’t find evidence of a geographical bias to the location of smaller vs. more established developers. And finally, because the app store data is not a time series, the results represent a snapshot and cannot capture changes over time.
This piece has been edited for clarity. Explore the project further here.