Actionable Insights From APT's Retail Practice
Header

Does Big Data Have to Be So Big?

February 11th, 2013 | Posted by CGreenbaum in Uncategorized - (Comments Off on Does Big Data Have to Be So Big?)

A recent discussion on Retailwire asks, “Does Big Data Have to Be So Big?” APT SVP Jonathan Marek had this to say on the matter:

Slowly but surely the industry is coming around to the rational point of view on Big Data. Bigness is not a virtue in and of itself. Actionable data is one step closer to the right end goal, because it starts to build in the idea of getting an ROI on data analysis. And here, there is also rightly a Bayesian aspect. Executives must have a deep understanding of how their business works, so that they can rightly assign the right level of belief in what the data is allegedly saying.

Retailers also need to learn that certain methods of analysis create more actionable insights. To take a silly, non-ad example: if I scrape social websites and see lots of customers think I should paint my stores green, so what? Maybe that gives me a hypothesis. But it doesn’t mean there will be ANY return on that investment. If I actually do it in 10 stores and see a statistically significant sales lift big enough to justify the investment, then that is much more actionable. And then, if I can use lots of data (dare I say Big Data?) to sort out which stores or customers show the greatest incremental lift, then I can further target the investment for greater return.

That’s the key. Big Data => cash ROI, not “interesting sounding ideas that seem like they will probably make me money”. Too much of the Big Data space is still in the latter world.

Why Move Beyond Traditional Product Mix Analysis

February 4th, 2013 | Posted by retailblogadmin in Restaurants - (Comments Off on Why Move Beyond Traditional Product Mix Analysis)

Recently, Cosí Inc. launched a “Pop Up” location in Chicago with a new menu designed to reinvigorate a brand that has struggled in recent years. The trial menu includes only 35 items, down from 65 items in the fast casual chain’s traditional restaurants. Though it is great that Cosi is trying to test new menu options without exposing the network to substantial risk, this one-location trial will not yield sufficiently significant results to generate confident rollout recommendations because of the small sample size. Instead, restaurants such as Cosi should use advanced check-level analytics to generate hypotheses about the economic impact of each menu item, and then run scientifically robust tests to either validate or disprove these hypotheses. (more…)