Introducing Data Science in Existing Products — When and How?

Data is the new oil

Artificial Intelligence, Machine Learning, Data Science. All these buzzwords are taking the tech world by a storm. It’s no wonder then that a startup or an established corporate may feel intimidated alike by the enormity of resources and a lack of history. I’m sure at least half of the readers may have heard of the below phrase.

Also, many of us would have come across the below chart depicting the exponential increase in data production, leaving a lot to an entrepreneurial imagination.

Everyone wants to be a part of the gold rush and market themselves as a frontier on Data Science. But let’s get a few facts straight before investing precious time and valuable capital.

What do I need?

Fancy terms and fancy talks all around. But do you really need any of these technologies? There are obvious reasons why one should start investing in data. All the big tech companies like Google, Amazon, Microsoft, and countless other smaller players are in fierce competition to become the monopolist in this infant market. Corporates are dominating domains with specific data and trying hard to monetize them for the long run.

There are products built on data, and there are products improved with data. Which one are you? Answer this question carefully and it may go a long way in reducing a lot of potential questions further on.

Driverless cars, facial recognition, speech to text, smart speakers, are just some of the ways in which AI is pervading our imagination quite significantly. Even traditional industries like Banks, Retail, Telecom are using Machine Learning to improve their functioning in multiple aspects. However, the latter set of organizations use data to automate only a part of their products and that part is rarely the major revenue contributor. The accuracy of an ML model keeps improving with time, but that also means that it is fairly low in the beginning to be market-ready. No wonder then, that the established corporates implement ML only in places where the risk is minimal, and which are dying for automation. This leads to the next question that you might have.

Which feature do I start with?

It makes sense to understand how these technologies benefit the functioning of an application, to decide their introduction in a project. Below are only a few common examples to help prove a point, but the actual utilities might be countless. I would leave the reader to research on the bolded terms mentioned here, as the definition borders are fuzzy and the interpretations are varying. But a general consensus could be found in the articles of industry experts like Andrew Ng and others. Now, getting down to applications and use-cases.

If you need to understand the trend or tendency of users to reach a business decision, you could look at how Data Science could convert mundane statistics to a super-powerful decision engine. For example, personalized marketing curated for specific personas, or assessing risk for a credit card consumer.

You could rely on Machine Learning to predict outcomes, or to replace numerous rules and simple logic which may otherwise take a lot of development and testing time to go live. For example, one could perhaps train a model to detect frauds in insurance, loans, and other financial areas just by using history the right way.

Artificial Intelligence could be a value addition at places where you may need human sensing abilities like seeing, hearing, etc. as a part of your user experience, like automated subtitles for a streaming channel, or voice recognition for security.

The stakeholders need to carefully gauge the potential benefits vs cost incurred. Accuracy could prove to be the biggest hurdle sometimes. Also, a typical Data Science project takes its own sweet time, over long development cycles and multiple iterations, sometimes utterly unfruitful, which also needs to be a topic of discussion while reaching a decision. Data is bulky and it requires state-of-the-art storage mechanisms and processing capabilities which also needs money regularly. A collective decision with people representing multiple departments should help.

Great! Anything else?

Yes. You need people who speak the language of data and maintain your systems. With the advent of Data Science came a plethora of skilled professionals working on different aspects of it. Creating a balanced team is important to the extent that it can make or break your product. Like the domains, even the designations don’t have a concrete border, so one needs to assess the skills carefully before bringing a professional on board. Some general definitions of terms pretty popular nowadays are:

Data Scientist: Perhaps the most generic term referring to someone who can clean data, examine it, create models and derive insights.

Data Engineer: Someone who can take care of data storage and accessibility with cost optimization in mind.

Machine Learning Engineer: One who can create models firing up the core of a product.

Business Intelligence Analyst: Someone who could study data to come up with business solutions and derive insights which could drive a business solution.

Big Data Developer: This person can handle the tools and technologies required to play around with enormous amounts of data, in an effective manner.

This tip of the iceberg should help one get started to google the terms and design a path to learn further. All the best!

This article was originally published here.

October 21, 2019

0 responses on "Introducing Data Science in Existing Products — When and How?"

Leave a Message

© Pragmatic Leaders. All rights reserved.
X