APIs (Application Programming Interfaces) are everywhere. We use them every day, whether we are aware of it or not. They are basically the bridges that enable one piece of software to talk to another - Uber’s app to Google Maps, for example, or Amazon’s Alexa to music-playing services or train timetables, or Ebay to your PayPal account or Spotify to your Facebook friends.
The APIs run by the tech giants like Amazon, Google and Facebook are perhaps the best known. And there are increasing numbers of APIs designed for businesses such as Stripe, which supports the easy and rapid creation of new payment systems. Essentially, the idea behind APIs is to allow developers to save time and money by using apps and systems that others have already developed. For Uber, as an example, this has meant being able to avoid having to develop their own maps by piggy-backing on Google’s existing work.
The analytics economy and APIs
APIs are essential for the analytics economy which evolves rapidly with API economy and adopts the use of analytics in our daily lives. The integration of APIs into our decision support systems would require a series of processes. We would start with asking the question: what do we need to know? (I want to learn exactly how much money will be left in my bank account in three months), then call the API to collect and provide the required data (here are my daily transactions for the last 2 years), then call another API to run analytics (run the statistical forecasts for an accurate estimation), and finally make your summer plans in advance based on your budget! Since, like any other tech development, APIs are designed to be in use for certain applications, there are likely to be points in the analytics lifecycle where they are significantly more reusable and efficient and even essential. But what are these points?
The early use of APIs in analytics was to draw in new data for analysis, from a wide range of possible sources. Using Uber as an example again, it uses an API to draw data from Google Maps to enable it to pinpoint users and cars. The huge advantage of APIs is that they can turn almost any input into data, including speech and images. Google’s Cloud Translation API can even manage inputs in different languages. For the analytical use of APIs, the requirement of data preparation is unlikely to be going away. As big data becomes even more ubiquitous, being able to draw down data from more (and sometimes irrelevant) sources seems likely to become a distinguishing feature of new analytics models. Analysts always need to remain aware of the issues of ‘garbage in, garbage out’, but it is also true that some of the most valuable insights are obtained by extracting information from new or different data sources.
What’s more, data preparation may not be the most exciting work, but it is both vital and time-consuming. Improving its operation using APIs could generate huge efficiencies in analysts’ time. Given the global shortage of data scientists, this wouldn’t be insignificant for any company wishing to rely on data to support decision-making.
Getting insights to consumers
Besides the data preparation, APIs are also starting to play an important role in getting conclusions and insights to consumers and end-users. The output from models can sometimes be quite technical, so an API that helps to deliver it into a more consumable, user-friendly form is likely to be very popular. It is not hard to see that APIs for data modelling and data visualisation could turn out to be essential tools for data scientists interested in making their work accessible and presenting it clearly and in a format that can easily be understood by business users. As an example, the utility companies are embedding analytics into their systems as a service since the energy becomes more of a commodity with low profit margin and it gets more important to differentiate yourself with the value-added-services to keep up in a competitive market.
The increase in APIs enabling speech and voice recognition such as AT&T Speech, also opens up some very interesting options for presentation to end-users. For example, conversational interfaces and AI-driven chatbots may be used to respond to demands of people in the executive level or tactical non-techy end users by providing a human-like interaction while extracting the insight from data. This adds a whole new layer of interest and complexity to the term ‘citizen data scientists’.
Predicting the future
Crystal ball-gazing is a risky business, however, I feel confident in saying that the use of APIs will continue to expand in analytics, as the users enjoy the benefits more and more.
Suggested read: You might also be interested in my colleague's thoughts on How will APIs change the role of data scientists?