Measuring the Effect of Localization on App Sales

Posted on November 11, 2013

tl;dr

If you want to cut to the chase, you can just read my conclusions, then share your own results.

Introduction

Over the years, I’ve repeatedly heard people say that localizing an iOS app is important for maximizing the app’s audience and ultimately the app developer’s revenue. Apple urges developers to localize their apps every year at WWDC, and similar sentiments are often shared in conversations among developers whenever they get together. As the theory goes, people like to interact with their iOS devices in their native language, and they are therefore more likely to purchase an app if it’s available in the their native tongue. That’s all almost certainly true, but the key thing that no one ever talks about is the numbers. Does localizing an app really result in more sales? If so, how big of a difference does it make? In which countries is localization the most critical? How long will it take for the investment in localization to pay for itself? These are all questions that can’t be answered without some hard data to work with. Unfortunately, that data is hard to find.

Data on the sales impact of localization is hard to find, I think, because developers are reluctant to share too much financial information with the world. And rightfully so. But I realized after my own recent experience localizing an app, that the information that’s most valuable to other developers doesn’t have to be quantified in dollars and cents.

What follows below is an analysis of the sales data I collected after localizing one of my apps. This analysis is an attempt to collect and share data that I think will be valuable to other developers, but it’s also an attempt to model a process that generates useful data without revealing too much information about revenue. My hope is that the data will be useful to others, and that other developers will use a similar method to share their own experiences with localization. By sharing solid numbers, it gives all developers the opportunity to make decisions that are based on facts instead of guesses.

Background

In this study, I’ll be looking at just one of my apps, Benjamin for iPhone (hereafter, just Benjamin). Benjamin is a niche productivity app which sells for the relatively high price of $9.99. As a result of its small audience and high price, there are rather few daily sales – especially outside the United States. Benjamin has been available in the App Store for about a year and a half, and until very recently it was available only in English. As a part of a major update (version 2.0, released June 25, 2013), I decided that I should move beyond my English-only strategy and localize the app for other language markets.

To understand the language choices I made in localizing Benjamin, you first need to understand a little about the app and its audience. Benjamin is a task manager based on the FranklinCovey method of time-management. FranklinCovey is one of the world’s largest providers of time-management training, drawing their clients primarily from white collar workers in fields like healthcare, education, and other corporate environments. According to their own reports ((FranklinCovey Investor Relations)), FranklinCovey does business with 90% of the Fortune 100 and 75% of the Fortune 500. Their bread and butter are multinational corporations. As a result, their training is conducted around the globe and in many different languages. The multinational character of FranklinCovey’s operation was an important consideration in my decision to localize Benjamin. FranklinCovey conducts training around the globe, so it’s reasonable to expect that Benjamin has potential customers around the globe as well.

With this in mind, I decided to localize Benjamin into five additional languages: Spanish, German, French, Japanese, and Simplified Chinese (which is used in the People’s Republic of China, but not in Taiwan and other Chinese speaking countries). These languages were chosen because the countries in which they are primarily used have all widely adopted the iPhone and iPad, and they all have a large base of people trained in FranklinCovey time-management – with the notable exception of Simplified Chinese. FranklinCovey has only had a presence in China since 1998 ((FranklinCovey China)), but it’s a booming market with enormous growth potential. I decided to include Simplified Chinese primarily as a prospect, so that my app would be ready as the Chinese market grows and matures.

Assumptions

As much as I wish it were not the case, I don’t have perfect information about my customers. I don’t know where they come from, how they found my app, or even which languages they speak. Consequently, there were a number of assumptions that had to be made in performing this analysis. I list the major ones below so that you can decide for yourself if they are valid.

Geography and Language

Apple doesn’t provide any information about the language used by those who download apps from the App Store, but it is possible to determine from which country’s store a customer purchases an app. Unfortunately, geography is not a perfect analog for language. One only has to look at the growth of the Spanish speaking population in the United States to know that’s true. And don’t even get me started on linguistically confused countries like Belgium and Switzerland! But despite it’s imperfect nature, some general assumptions can be made regarding geography and language. In this study, I count the United States, the United Kingdom, Canada (my apologies to the Québécois), Australia, New Zealand, India, and Ireland in the English language group. I count Spain, Mexico, and all of Central and South America (with the exception of Brazil) in the Spanish language group. France is counted as the sole member of the French language group, Germany is counted as the sole member of the German language group, Japan is counted as the sole member of the Japanese language group, and the People’s Republic of China is counted as the sole member of the Simplified Chinese language group.

Promotion

I did my best to garner attention for the Benjamin 2.0 update, which hopefully affected sales positively. Although my efforts were global (since they were on the Internet), all my promotional materials and press were in English. My assumption, then, is that if different language groups have benefitted unequally from my promotional efforts, then English language sales have benefitted most from that inequality.

Companion App

A new app, Benjamin for iPad, was launched at the same time that version 2.0 of Benjamin for iPhone was released. Benjamin for iPad was a long-requested product from my established user base, and has been well-received in the App Store. It’s reasonable to think that the availability of an iPad version has contributed to the sales of the iPhone version, especially since they sync with each other and are designed to be companion apps. But since Benjamin for iPad is localized into the same languages as Benjamin for iPhone, my assumption is that any effect that Benjamin for iPad has had on the sales of Benjamin for iPhone have been felt by all language groups equally.

Methodology

The easiest way to measure changes in the rate at which an app sells is by comparing its average daily sales over two different time periods. An app’s average daily sales (ADS) is the total sales of an app during a time period divided by the number of days in that time period. For example, if an app has 5 sales on Monday, 2 sales on Tuesday, 7 sales on Wednesday, and 3 sales on Thursday, then its ADS is 4.25 over those four days. A naive comparison of sales growth due to localization could then be made by comparing Benjamin’s ADS from before localization to its ADS after localization for each language group.

The problem with a naive analysis such as this is that it fails to take into account the natural growth of the app’s audience. Looking at the change in ADS for any language group in isolation, it’s impossible to say whether sales growth was caused by localization, or if it was caused by other outside factors. For example, if the ADS of the Spanish language group increased after localization, can the increase be attributed to localization? Or is it attributable to my promotional efforts? Or to the presence of a companion iPad app that makes the iPhone app more attractive to customers? To isolate any growth that was caused by localization, the growth in each language group must be compared to a baseline. In this case, the baseline I’ll use is the English language group, since that’s the language in which Benjamin was originally available.

In this study, two time periods were established. The pre-localization time period was Feb 8, 2013 to June 23, 2013, a span of 136 days. The post-localization time period was June 27, 2013 to October 1, 2013, a span of 97 days. These time periods were chosen because they are ordinary and uninteresting. There’s no particular promotional push that happened during those time periods, and there are no seasonal effects that should have influenced sales. They are representative of typical days in the App Store. Worth mentioning is that the post-localization period begins the day after the last effects of my release promotion efforts can be seen in sales reports.

Sales statistics for Benjamin were collected for both time periods, and broken down by language group. From these statistics, each language group’s ADS was calculated for both time periods, and growth between time periods was calculated for each language group. The difference between each language group’s ADS growth and the baseline (English language group) ADS growth can then be attributed to the effect of localizing the app for that language group.

Results

The pre- to post-localization change in each language group’s average daily sales (ADS) is summarized in the table below. The Change in ADS field represents the percent growth in each language group’s sales when comparing the post-localization ADS to the pre-localization ADS. For example, if a language group had a pre-localization ADS of 10, and a post-localization ADS of 15, then its Change in ADS would be 50%.

Language Group Change in ADS
Chinese (Simplified) 40%
English 11%
French 1,165%
German N/A
Japanese 180%
Spanish 89%

From the table above, you can see that our baseline (the English language group) experienced 11% growth in its ADS after becoming localized. If we assume that this 11% can be accounted for by the effects of marketing, the presence of a companion iPad app, and other effects that are common to all language groups, then we can subtract that 11% baseline from the other language groups’ growth to calculate how much of their growth is attributable to localization. This growth attributable to localization is summarized in the table below.

Language Group Change in ADS Due to Localization
Chinese (Simplified) 29%
French 1,154%
German N/A
Japanese 170%
Spanish 78%

Note: The German language group has its results above listed as not available because of a division by zero error. In the pre-localization time period, there were no sales of Benjamin in Germany. In the post-localization time period, there were 5 sales.

Conclusions

From the results above, it’s clear that localization had a marked influence in increasing Benjamin’s sales in the regions for which it was localized. The Spanish language group in particular shows a growth in sales which I have high confidence can be attributed to the effect of localization.

It should be pointed out that that the number of sales in the German, French, Japanese, and Chinese (Simplified) language groups during the pre-localization time period are very low (in the single digits). Consequently, too much confidence should not be placed in the results for those groups. It took such a small change in sales to wildly influence the results in these language groups that it’s difficult to separate the signal from the noise. My gut (and my insight into the raw numbers) tells me that the there was a genuine improvement in sales that can be attributed to localization in the French and German language groups, though the magnitude of the improvement is unclear. The results for the Japanese and Simplified Chinese language groups, on the other hand, should be disregarded as pure noise, I believe. Sales in these groups were so low that I don’t think any valuable conclusions can be drawn.

So was localization worth the cost in time and money? I would say that it was. The Spanish localization has already more than paid for itself, and the German and French localizations are well on their way to doing the same. There’s an argument to be made that the Japanese and Simplified Chinese localizations were a waste, but my hope is that Japan in particular will become a region of future growth.

Now It’s Your Turn

I know that this was a long journey to take in order to arrive at the somewhat obvious conclusion that “localization improves sales.” The point wasn’t to just show that localization improves sales, though. The point was to demonstrate a way to measure the effect of localization on sales that results in meaningful data that can be shared with others. The method employed above allows developers to calculate and share the effect that localization has on their bottom line, without revealing exactly what that bottom line is. My hope is that other developers will use a similar methodology to publish the results of their own experiments with localization so that our community of developers can accumulate the information needed to evaluate in advance whether localizing for a particular language is a valuable use of our limited time and budget.

If you undertake a similar study of the effect that localization had on your app sales, please publish your results and contact me so that I can link to them.

Note: A version of this article appeared in Issue 12 (Oct 10, 2013) of The Loop Magazine under the title “Localizing Your App.” The version above includes updated sales statistics that weren’t available at the time it was first published in The Loop Magazine.