Big Data is all the rage across many enterprises. The potential payoffs are compelling. For example, research out of MIT found that firms leveraging Big Data achieve, on average, 5-6% greater productivity and profitability than their peers. McKinsey calls Big Data a game changer for sales and marketing along with other areas of the business. But like anything else, getting the most out of data means knowing what not to do. Companies looking to extract value from their terabytes of data should make sure they avoid the following 6 mistakes.
Boiling the ocean
Big Data can be big work. It is easy to burn through a lot of time and cost finding insights that don’t materially address major business challenges like getting closer to customers or improving operational performance. One way to ensure value is to ask research questions whose answers will directly impact key corporate goals. This focused method also enables your company to ‘learn as they go’ and develop quick wins that justify further effort and investment.
Only considering data in silos
In many organizations, the majority of data resides in functional areas or business units — not in an enterprise-data warehouse. Only analyzing siloed data reduces your chances of finding key insight that can affect a company because you are limiting the number of variables and quantity of data under consideration. However, bridging these silos is easier said than done; organizational and data issues may hinder an enterprise-wide data mining effort. In other cases, managers often limit their analysis to existing digital data. This approach could miss out on insights that are discovered when analog data (such as social feedback and qualitative research) is ‘datafied.’
There are good reasons why Big Data resembles science. The analytics can be challenging and methodological errors are not uncommon. “There is significant risk of the analytics being wrong,” says Neil Seeman, founder & CEO of global online data collection firm, The RIWI Corporation. “Systematic bias can easily slip into enormous data sets. Or, not understanding unknown bias in the data set results in false conclusions. Case in point was the early analysis of large HIV data sets; this did not consider the influence of intravenous drug use.”
Focusing on cause instead of correlation
Understanding precise cause and effect is difficult and impractical. What’s more actionable is uncovering correlations — patterns and associations that help predict what will happen next time. Managers should be mindful of looking for and expecting data perfection. For example, the shelf life of market-based insights could be measured in days or even hours. Often it is better to quickly make decisions with 80% confidence in the data than to wait for perfection farther out in the future.
Disregarding qualitative knowledge
Data analytics can deliver many insights but it often cannot tell the entire story. Take the drivers of consumer behaviour as an example. It is difficult to comprehend what drives action without looking at qualitative research tools like behavioural psychology or anthropology as well as expert opinion. These tools should be used to fill in knowledge gaps.
Forgetting about instinct and creativity
At a certain point in the future, the leaders in each sector will have comparable Big Data capabilities and access to the same data. To wit, the Open Data movement is making terabytes of the same data available to everyone. Like other innovations, the greatest returns will flow to those who use the tools and methodologies in the most creative way. As well, management instinct will continue to play a key role in setting Big Data priorities and figuring out how to combine disparate information into more powerful conclusions.
A dangerous assumption made by some companies is to think your entire team should be made up only of credentialed “data science” experts. According to Seeman, “Experience in this fledgling field of data science often eclipses the value of fancy degrees from prestigious universities. What are needed are curiosity seekers with demonstrable experience in pattern recognition and exploiting data for real value. In my case, everything I learned about Big Data came through experimentation, failure, and asking really dumb questions — through efforts to solve the problem of how to collect a unique data stream from every country and territory in the world.”
Big Data is still in its infancy. The first cases studies are still being written. Of course, success will be a product of a strong top-down mandate, having sufficient resources and working with competent partners. At the same time, managers should be mindful of hamstringing themselves by not following a common-sense and continuous-learning approach to project design and implementation.