It’s no secret that data is making the speed of business advance dramatically. With the convergence of cloud, mobility, social platforms and the emergence of IoT, companies are forced to look outside and combine in-house data with partner and third-party data to make the kinds of quick decisions that help them stay relevant in the market.
Simply put, data-driven decision making is becoming the #1 priority for businesses across all industries. A solid command of data science helps companies build “sense and respond“ systems that take data from within the enterprise as well as from outside, and create a smart, contextualized view of both customers and business situations.
Competing Concerns About Catching Up
At Mindtree, we see that beneath this reality, companies have concerns on two major fronts: the business side, and the IT side. Business teams want a cohesive strategy and quick build out of analytical models that can aid customer retention, cross sell and up sell, product recommendation, and so on. But IT teams are focused on challenges like integrating heterogeneous data sources, measuring data quality, addressing data security and identifying the right set of machine-learning algorithms that can deliver the best outcomes.
For example, companies want to identify the right set of offers/promotions in real-time based on customer behavioral patterns and past transactional data—a perfectly reasonable business need. On the technical side, this requires real-time event processing capabilities through streaming techniques, applying machine learning models, actuating a decision engine to identify the right set of promotions, and real-time campaign execution—all with the ability to measure the effectiveness of this automated yet personalized offer.
Meanwhile, both teams have concerns about the size of the investments necessary, what ROI might look like and how it should be measured, and about their lack of a proper data science implementation strategy amid an explosion of big data tools.
Need of the Hour: Data Science Accelerator
What modern businesses need is a data science accelerator that not only provides an analytical decision engine, but also serves as innovation sandbox that allows for experimentation and evolution. There are some essential features for a framework like this to succeed:
- An engine that can rapidly ingest heterogeneous data sources (poly-structured), and supports polyglot persistence with multiple data compression techniques.
- An abstraction layer on top of the polyglot persistence, to ensure consistent read/write patterns across different databases.
- Multi-tenant platform functionality, to ensure data isolation across LOBs.
- Readymade industry-specific machine algorithms to jumpstart analytical capabilities.
- Real-time personalization and targeting.
- Support for hybrid cloud, where idea validation and proofing happens in a public cloud while production deployments come from on-premise infrastructure.
Pointing this out is all well and good, but what is the roadmap? What’s the right implementation strategy? What are the key pillars for success?
It may seem like a lot, but it can be done. This blog took some time to point out the “what”—be sure to check out the next blog for the “how.”