Google’s BigQuery service allows users to run near-instantaneous queries in massive data sets with billions of entries. The service, which is similar to Hadoop, is often partnered with commercial software packages for deep analytics used in business plans. One of these partners, Tableau, just gave details on a project that helped retailers adjust their business models for huge snowstorms. Tableau and marketing firm Interactions discovered something interesting at the study’s end: Store profits were impacted by the threat of a storm just as much as an actual snowstorm.
The retailers used in the data sets, which were not named in the recently released study, wanted to use big data analytics to tweak the way they handled extreme atmospheric conditions as well as simple inclement weather. According to Giovanni DeMeo of Interactions, the firm wanted to help their retail customers understand how weather events change shopping patterns, and how to better prepare for major snowstorms.
By comparing more than a million data points about consumers’ purchasing habits in-store and running comparisons against specific weather events and the way storms behaved–which meant quantifying how, say, slushy snow effected consumer traffic and days when predicted storms did not occur–Interactions gave their clients more accurate predictions of what unexpected items would go out of stock, how to optimize item assortments, and suggestions on how to increase sales for high-demand items. DeMeo told Fast Company that one of the biggest points of interest in the project was identifying data points related to snow flow and snow depth, and then correlating them with what products customers bought, what time of day they went to the store, and even if they went to the store at all.
The study, which included three years’ worth of data that included 1.4 billion rows of weather and retail information, resulted in some counter-intuitive findings. It turns out that when storms were predicted but didn’t actually occur, customers still bought essential supplies for their homes anyhow. Whether they simply didn’t hear the weather forecast or were subconsciously spurred to buy home items, the result was the same: A non-existent storm spurred emergency shopping as much as an actual storm.
This emergency shopping always tended to occur at the last minute, even when there were storm warnings given a week in advance. According to Tableau’s Francois Ajenstat, sales in certain item categories spiked from 20% to 261% the day before a storm was predicted to arrive. The study also found that sales of items during snowstorms themselves had just as much to do with the day of the week the storm happened on as the weather event’s intensity.
Because having access to these extra analytics means being able to fine-tune buying habits for large retail chains, consultants and retailers are willing to shell out big money for big data. When Fast Company spoke with Google’s Big Data initiative leader Ju-Kay Kwek, he spelled it out for us. “Based on the dollar impact to the retailer, it’s possible to use the understanding of historical patterns of product sales to bulk up on inventory for certain products. Doing this rapidly and and down to the individual location-level is something relatively new, and something that is very useful,” he said.
Findings from the study were presented at Tableau’s European Consumer Conference in London on June 11.
[Image: Wikimedia user Vispec]