As the API Ecosystem grows, at 3scale we often asked where growth is fastest and how to characterize certain trends. While this is challenging to do, sites like Programmableweb provides great data of many public APIs and regularly feature trend analysis (see John Musser’s excellent “Hot or Not” Presentation Gluecon for example).
To dig deeper though, we decided to try to classify a snapshot of 1000 of the latest APIs recently registered on Programmableweb, with some surprising results.
The study took the latest 1000 APIs registered in the Programmableweb directory prior to the 2nd of May 2012 and analysed them in various ways. Hence the sample stretches from the 17th of January until the 2nd of May 2012. The directory is now over 6000 APIs and those latest 1000 were added in just 90 days – so it’s recent snapshot and should tell us something. The APIs were classified in a number of dimensions ranging from the size of the company launching the API, the sector it operates in and the business model behind the API.
The most interesting finds were …
It’s not just web startups!: A common perception of APIs is that the majority of APIs are run by web technology startups. However, in the sample, only 17% of the companies providing APIs were web technology startups by our definition of less than 3 years old with a primarily digital / web based product).
The long tail is still missing in action: A second assumption is that APIs would likely “mirror the web” in terms of distribution by the size of the company providing them – mirroring a long tail distribution of few large companies providing APIs and many tiny companies. However, from the data it appears that APIs are primarily concentrated on medium sized and small companies, as well as with a few large organisations with many APIs. The true long tail isn’t quite there yet.
Mobile “Only” is rare – but Multi-channel is huge: Thirdly, with mobile being such a driving force in the current evolution of the Web we expected to find many APIs which primarily focused on powering mobile applications – however only 10.7% of the APIs in the sample could be said to have a pure mobile focus. There were though many of the rest (another 51% of the total) which clearly targeted both mobile and non-mobile re-use. This result is likely to be somewhat skewed since many private mobile APIs are not to be listed on Programmableweb to date, however it still seems to suggest that we are entering a “multi-channel” world rather than a purely mobile one.
Business Models are getting there: A final common perception is that API business models are still in flux and there are no clear patterns emerging. In the sample there were a diverse range of business models and the majority of APIs launched had clear business value to the provider – either in terms of direct monetization or indirect collateral benefit. It also seemed there was convergence to a few common models.
APIs from companies of all Sizes
The chart below shows the estimated size distribution of the companies launching APIs in the sample set (based on data from Crunchbase, the company’s web sites and general web search). Interestingly the distribution does not follow a clear long-tail distribution as the Web does. In fact:
Large companies are well represented: launching significant numbers of APIs.
There is a “missing” long tail with companies of medium and small size dominating API launches and not the much longer tail of relatively few micro / very small organisations (the number of micro-organizations is similar to the number of “small” organisations when in a long-tail distribution it would be much greater).
This seems to indicate that to date it still requires a certain size of company – 10 employees upwards and more likely 50 employees upwards to launch an API, rather than an explosion of APIs from very small organisations. This tendency is likely due both to technical constraints (it remain technically challenging to launch an API, despite platform and tool advances) and business constraints – APIs are likely to appeal primarily to organisations with standardized products and services to distributed but not yet to those with niche bespoke / custom products.
Not only for Web Technology Startups
The study also separated those considered web technology startups from others since this category was expected to be one of the most significant sources of APIs. However, only 17% of the organizations were identified in this category.
The definition we used was less than 3 years old and with a business focused primarily on web/digital product or distribution. While this misses out some others that might well be considered startups in other spaces we wanted to figure out how many companies were in the Tech Startup ecosystem specifically since these together with a few giants, these tend to dominate the much of the API tech news.
Not all of the startups identified were small – some were already organisations of significant size – having raised multiple rounds of capital in their 3 years.
Company Size by Business Area
Figuring out which companies do what was also interesting and a few clear patterns emerged from the sample:
Very small API providers (35%) tended to providing novel APIs as extensions to SAAS applications (APIs are now becoming a standard part of the SAAS Armoury!), but a surprising number provide primarily infrastructure style services (29%) such as SMS resale or tools / utilities.
Small and Medium API providers also focus on similar areas, but also add Data Services (20%).
Large and very Large organizations overwhelmingly (54%) released APIs focused on infrastructure services, with payment and ecommerce services a distant second (18%).
Governments primarily focused on Data APIs that covered both raw information services in categories ranging from Seismic Activity and Weather data to government spending – secondary to these were APIs in areas such as compliance, procurement and tendering.
Mobile and Multi-channel
One of the biggest growth drivers for APIs is certainly mobile – with mobile apps increasingly needing to call back to network services to deliver the right user experience. To understand how much of a driver mobile was we also classified the APIs as to whether they had an exclusive (or near exclusive) focus on delivering for mobile, the objective of delivering for both mobile and non-mobile or (finally) being primarily for fixed Web re-use.
The figure below shows the results of this split. Surprisingly, less than 11% of the APIs in the survery seemed to have a pure mobile focus. However the significant majority (62% total) were either pure mobile or mixed both Mobile and Non-mobile targets.
The trend is even starker if Scientific APIs (125 in the study) are factored out since these all fell into the non-mobile category – in this case 70% of non-scientific APIs addressed mobile to some extent. This indicates that mobile is of very significant driver, but multiple channels are even more important.
Business Models Diversity
The business models applied by many of the APIs have shifted to provide clear value to both the API provider and the user of the APIs. Figure 3 shows the split in the sample between primary business models. The business model categories in the classifications were defined as:
Direct API Monetization (is-API): API service in which fees are applied directly to API usage, either to make calls to the API or for the immediate effects of API calls (such as SMS sending, cloud infrastructure operations or pay per query search services).
Product Extension APIs (Product Addition): these APIs provide additional features for customers of an existing (generally browser based) application as additional features. The value to the user is a new access to their existing data, the value to the API provider is potentially monetary (API access requires higher tiers of service) or indirect (the API increases service stickiness).
Product Projection APIs: these APIs enable third party partners, resellers and application developers to build on top of an existing product and act as channels or enhancers. The value to third parties is increased functionality or revenue share and the value to API providers is the creation of new distribution channels and a richer experience for their users.
Non-Profit: general free to use APIs for wide re-use or with a donation system to support the maintenance.
Government: APIs are provided by public agencies and by governments at no cost to the user but with rate limits and restrictions imposed (although some government APIs in certain cases are paid APIs for volume usage or re-use for commercial gain).
Some APIs also applied mixed business models, but in most cases the focus was clear. Suprisingly, the business models applied for commercial APIs split almost evenly between the first three models (26% focusing on customer enhancements, 23% on direct monetization and 20% on reseller style APIs) – indicating the emergence of clearly distinct value propositions. The remaining APIs were relatively evenly split between Government, Scientific and Non-profit models.
The sample size of the study is small and no doubt some artifacts are accentuated or muted as a result, our classifications also might not always have been spot on, so we’d advise against taking mission critical business decisions based on it. The sample represent approximately 1/6th of the overall programmableweb directory at the time of writing however, so perhaps provides some useful hints to current trends.
Over the coming weeks, we’ll try to dig down into some of the elements of the study to tease out patterns.
Follow us on twitter and tweet out your thoughts!
Thanks to Programmableweb for the hard work in maintaining the directory, feedback on some of our initial data and also their own regular trends analysis!
Update: there’s a comment thread on hackernews: http://news.ycombinator.com/item?id=4079615
Data refers to the 1000 APIs registered in Programmable Web between the dates of 17/1/2012 and 2/5/2012.
Company Size data was captured from crunchbase, press articles, google searches and company sites.
Company Business Model information applies to the API in question, companies may have different Business Models for different APIs.