Big Data and Analytics lessons for credit unions can come from some unlikely sources. Consider the contest between U.S. and European weather-prediction models. The European Centre for Medium-Range Weather Forecasts (ECMWF) is widely acknowledged to be superior to the U.S. Global Forecast System (GFS). While the GFS has been improved since 2012 when it predicted Hurricane Sandy would not make landfall, the European model is still considered to be the better weather forecasting tool.
The ECMWF has the edge for three reasons. First, the model is run on a supercomputer with a ten-fold performance advantage over the U.S. hardware. Second, the ECMWF divides the atmosphere into ten square mile cells each with 137 layers.The improved GFS uses 8 square mile cells but with only 64 layers. Finally, the European software runs a variety of complex weather simulations that have consistently delivered better forecasts.
This may seem unrelated to credit unions, but there are several lessons they can learn from the battle of the weather prediction titans.
Hardware Horsepower Matters
When sizing hardware for a Big Data and Analytics project, don’t skimp on processing speed and disk space. Even if many years of internal data can be easily handled on planned hardware, consider the possibility of large amounts of external data that can be integrated with legacy data. In many cases, the biggest success stories of Big Data and Analytics have come from such integration. Estimating data volume from this perspective will drive a much bigger number. Whether on-premises or in the cloud, having the right hardware to accommodate these volumes will allow the credit union to take advantage of more Big Data and Analytics opportunities.
Smaller Grain = Bigger Gain
One of the advantages of the ECMWF model is a smaller level of data granularity. The smaller level grain allows a much deeper level of analysis. The parallel in the credit union world is capturing data at the transaction level. Too often, credit unions launch a Big Data and Analytics initiative only to discover the ability to drill down through the data is limited. This frequently happens when data is aggregated to improve performance. When users can drill to the transaction level there is a greater probability that analytics will yield high-confidence results. Performance risk is mitigated by following the first lesson.Wanted: Powerful Analytical Tools
The right hardware and the right data can be wasted by inadequate analytical tools. Using an Excel pivot table on a large integrated dataset might work but limitations of the software prevent deeper analysis. In evaluating analytical tools, credit unions need to consider multiple levels of users. (See our recent blog, 5 Steps for Making Analytics Work). Analytical tools need to fit a range of needs and user capabilities. Data scientists and front-line Member Service Representatives require tools that fit their respective job responsibilities.
Weather forecasters are constantly working to upgrade their ability to predict tomorrow’s weather. In the same way, credit unions need to learn these lessons to get the most out of their Big Data and Analytics initiatives.