Retail Analytics: More Than Just Numbers
An ever-increasing number of retailers are searching for undiscovered opportunities using data science and machine learning. Massive volumes of data in storage networks are accumulating daily for retailers to explore, but without the right tools and methods to understand their meaning, it’s all just noise. Choosing the right high-performance computing technology, developing the best processes to handle and prepare data for analysis, and creating expert algorithms to unlock hidden secrets are just some of the challenges retailers face. But perhaps the biggest challenge of all comes after the data is scoured and the analysis is complete, when the business interpretation and action plans are initiated.
Many retailers find that integrating new analytical tools, methods and results into their historical business acumen, culture and processes is a very tricky step. It's not uncommon for machine learning and artificial intelligence-driven results to be dramatically different from those of “human-driven” methods and tools. Traditional analytical methods too often include bias, limitations of corporate and industry orthodoxy. Machine learning and AI focus purely on data sets, while data scientists create the algorithms, prepare the data for analysis, and verify the results. Then comes the challenge of interpreting the results into business action.
One strategy to overcome the chasm between data science and the traditional retailing world is to embed retail staff members within the data teams. Data scientists certainly know their trade, but may know little of how analysis outcomes might be interpreted as meaningful and actionable or as precision “guesswork.” In the same way, it can be hard for retail business teams to accept and digest results if they don't understand how the data sources and algorithms were used to arrive at any given answer. With an embedded team member approach, a retailer can be better educated on the data science methods and understand the results. Getting accustomed to this new process helps retailers build confidence in the results and comfort in executing the recommended plan.
This process can also be inverted by embedding data scientists within the business teams. If there's a secondary goal to help acclimate retail business teams with institutionalizing analytics, getting the team used to an embedded data scientist as a team member may be a smarter strategy. Likewise, if there's a similar goal to get a team of data scientists well-accustomed to retail’s broad range of challenges, behaviors, influences and language, this approach is more appropriate if creating bias is recognized and addressed.
Another challenge retailers must overcome is convincing decision makers that complex data science-derived results are realistic, executable and sustainable. Even if the numbers are clear and easy for a data scientist to understand, they can sometimes be difficult to explain, especially when the results propose a radical departure from traditional business behavior and thinking. Often, a well-scripted, simplified walk through the analytics process and results are the best way to steer decision makers.
Data visualization is also a powerful tool that has become very useful in the business community to explain the results of complex analytics. Dynamic visualization allows recalculation and immediate presentation of results from “what if” scenarios, surely to be asked by decision makers.
Retailers are starting to gain traction in creating more meaningful changes to their businesses by embracing and operationalizing analytics, but it doesn't come without a few obstacles. To unlock the full potential of their data, retailers need to adopt strategies that will help their business teams better understand the analytical processes, and their data scientists to better explain it.
Paul Burel is the director of the retail solution portfolio at Fujitsu America, Inc., a company that develops business technology solutions.
