What's Crucial- Big Data or Big Insights?

Ujjyaini Mitra, Analytics & Data Science Leader & Mentor - Driving Data to Decision, Viacom18

Ujjyaini Mitra, Analytics & Data Science Leader & Mentor - Driving Data to Decision, Viacom18

One of the most promising jobs in the current days is “Analytic Translators”—who has no Statistics or Computer Science background but has the skill to read data, understand and appreciate data and translate the Analytic results to actionable business insights. Why?

Over the last decade, the world has seen a data revolution. There was a time when nations fought for natural recourses like petroleum, but the time has come when the war will be fought on data, which is why data is the new oil. Companies and nations having bigger ownership of data will rule the world. But data alone can’t make you king unless you know how to harness the power of data. Big data is like a gold mine. You know gold is there, but unless you mine the land you don’t get gold ores. Once you mine the gold ores, you melt them to get the pure gold, and these golds, when shaped into ornaments, are ready to be sold or used. It’s the same with Data. From the fast-flowing stream of data which are also coming from a variety of sources in various shapes and forms, we need to know how to mine and dig out the ores. Without right technology you can dig out ores which might have a lower concentration of gold that will reduce your profitability, similarly, if you dig out data which are of no use or less use, that won’t serve any purpose. So, beyond big data technology, it’s highly important to bring out insights which business can act upon. These are called actionable insights. Such insights could come from some simple analyses or through deep learning technologies. Irrespective of how complex or simple the techniques are, insights that drive bigger business ROI are appreciated more. At the same time, an outstanding model if can’t be adopted by business has zero value.

To give an example: Most mobile apps track user’s location. A Food tech app tracks user location every time one orders food or browse for restaurants. Every day on an average 300,000 users orders food. They have information for every single user transaction— which restaurant, which food items ordered, when, with how much quantity and how much one paid for the food. This is Big Data.

They analyzed the data and found that there is 30 percent of their user bases who order from premium restaurants. This is a descriptive analysis based on historical data. They do a consumer segmentation and found this base has three segments – a) Who mostly place lunch orders from a specific location (50 percent of the base), b) Dinner order during weekends (35 percent base) and c) Who order both weekdays and weekends (15 percent base). This is a deeper analysis. They do further geo-location and text analysis on the location and delivery address details of weekday consumers and find that those are mostly corporate areas, which means they order office lunch, while those ordering for dinner mostly come from residential locations. This is big data analysis.

Now their business development proposes to launch subscription scheme for premium customers, where if one pays certain amount they will be part of ‘premium club’, they get delivery charges waved off, plus buy one get one free on all food items from partner restaurants. They want to know what kind of subscription they should work on. This is business case.

Based on the Big data analysis, they then suggest that one single subscription won’t work, given 50 percent people order corporate lunch, there must be a Weekday only custom subscription. Best to launch a Weekend only subscription pack also. This is Actionable insights from the analysis done.Without this point the big data or the big data analysis had no value to the business.

Big data is challenge until you have right technology to store it and process it. It, however, can be tamed through new technology of cheap cloud storage, cloud computing, parallel processing in Spark cluster. Even after deploying the right big data infrastructure more than 40 percent organizations mention that they haven’t seen much value out of it. Why?

• They have not received right insights driven from the data

• Data-driven decision making has not been adopted at an enterprise level

• No clarity what kind of questions could be answered through Big data

• There have been more analysis-paralysis

• Vision may be too short-term ROI gain

• Data Science projects focuses to accuracy goal than adoption and ROI goal

Analytics is not a magic wand. As soon as infrastructure is ready,many people believe, soon it will start churning deep insights. Unfortunately, it’s never that way. Most cases it takes minimum 18-24 months before Big Data team can start producing tangible insights. Can business wait so long? Unfortunately, in the VUCA digital world everything changing so fast, when analytics team completed a whole project the reason of project might have already gone by. So, it’s very important that business teams actively participate in prioritizing the Big Data exercises.

Analytics team should be built with complementary skill sets. People from various backgrounds – Mathematics, Statistics, Computer Science, Engineering, Economics, Econometrics, Management, Psychology should join. Team must have few Data translator or Story teller at senior level. Team should be distributed into three fundamental groups:

Group 1: Data Stuarts - Who works on multiple dashboard, report, visualization exercises and support business with ample information to understand how we did until yesterday.

Group 2: Data Turbines - Who works on quick 80-20 analyses. Not all requirements are long term and quick data driven insights would be good enough for the business team to take an immediate action.

Group 3: Data Astrologers - Who will work on longer term predictive and prescriptive model building. These are long term sustainable requirements. Such projects have longer turnaround time, but when deployed produce big values.

Group 4: Data Researchers - I call this Aspirational group. This is in-house lab. This team does not work on any immediate business requirement, but mostly focus on projects which are future visions. They work on innovative projects, which will bring next level competitive advancement for the business. If such a lab cannot be established in house, one can collaborate with research institutions, where PhD or Post-Doctoral fellows along with best research experts can work on such visionary projects and publish joint papers or IP or Patent.

The single measure of success should be ‘Adoption’ of analysis to start with. Only after few successful adoption, one should measure them by monetary value addition. Because without adoption no analysis can generate any value to business.