News & Blog

May 12, 2015Blog

Bananas, Bananas, Bananas! The Importance of Real-World Context when Analyzing Data for Business Insights

Bananas, Bananas, Bananas! The Importance of Real-World Context when Analyzing Data for Business Insights

Shawna Thayer, Senior Director, discusses the value of context when looking at a bundle of data.

Background

A few years ago, I was working for a US-based, national grocery retailer and our analytics team was given new business questions from top management:

  • How could we leverage our internal customer data to select products for cross-promotions that were particularly appealing to our best customers?
  • What products should be featured together?
  • What weeks were most important to execute the cross-promotions?
  • How could we drive incremental sales with our top-tier (i.e., most profitable) customers?

This was a great opportunity for our Insights team to shine. This was also a huge opportunity for business application, so we were excited to jump right in. But the timing was not great – my analytics team was swamped in other projects, resources were stretched, and deadlines were pending. If we were going to do this project, it would have to be fast, and we’d need help.

At the time, we worked with a variety of data analytics providers and this seemed like the perfect project to delegate to our overseas vendors. Our partners all had very speedy data processing capabilities (crucial given this was a sku-level analysis) and the business questions were relatively straight-forward. Given our tight timelines, we split up the work across a few different vendor teams and decided to combine the insights at the end of the week to provide recommendations  to our promotions team.

The answer came back as “bananas! bananas! bananas!”

Later that week, as my team reviewed the different outputs generated from our partners, we saw a common theme emerging from the work – Bananas! One vendor provided a matrix of SKUs that were commonly purchased together, and bananas seemed to be a strong fit with everything from fresh meats to laundry detergent. Another vendor provided a week-by-week breakdown of the most-commonly-purchased SKUs to help us focus our promotion planning on key business cycles. Sure enough, bananas popped to the top of those weekly lists as well. A third group indexed products purchased by our top-tier customer segments in Q3 against the total shopper population and found (you guessed it) that our best customers were even MORE likely to buy bananas than the other groups. The studies all came up bananas, bananas, bananas!

Now, anyone familiar with American household diets and purchasing patterns will not be surprised.  Bananas are heavily consumed, spoil quickly, and are amongst the most basic of staples in every American shopping cart. Because of the popularity of these items, they are bound to show up in any sku-level analysis. Therefore, these analytics teams did not produce any actionable business insights. Promoting bananas along with batteries or laundry detergent wasn’t going to do a thing to boost sales.

The Importance of Context

So, we learned a valuable lesson in businsess analytics: Understanding the context of the question is crucial for providing true insights from your data. In this case, many of our partners were not familiar with American habits and trends of banana consumption. Since our analysts did not realize that nearly every grocery shopper in the US makes frequent banana purchases, they did not realize that the “insights” highlighting a banana opportunity just proved what my team already knew – bananas are a staple.

Lessons Learned

It’s very important to work with a partner that has regional context for your business to ensure that insights are relevant for your market. Market dynamics change frequently and there is no substitute for understanding the fundamentals of a local market when pulling insights from data.

We also learned the folley of splitting up the full context of a business question across teams that did not interact with each other. The approach of “divide-and-conquer” was highly inefficient for us. We had three different teams replicate the same fundamental error independently. Our vendor partners had intelligent and curious analysts, who, given a broader context, would have not wasted time tracking down correlations to banana sales. Taking a holistic approach to any business question results in fundamentally better insights. Project management across analytic and research projects, and working with collaborative partners is the key to comprehensive insights and actionable recommendations.

Latest news & blog posts

January 22, 2018
January 20, 2018
See all posts