Cisco Blogs

What the Supermarket taught me about Big Data

October 25, 2012 - 4 Comments

Author’s Note:  I have no kids.  I have friends with kids, who used to be in diapers.  The kids were in diapers, not the friends.  I’ve changed a few in my day, but not nearly as many as my friends have. And yes this has some sort of relevance to this story…


In every trade show or conference there’s someone talking about Big Data.  They talk about algorithms, CPUs, memory, software stacks, cabling, racks, ROI, TCO, nodes, names, federation, centralization, organization until you get “the pitch.”  I’m not really interested in the pitch for why someone’s product is better than the other, I’m more interested in the “What is the Problem that you’re trying to solve?”  This to me gets to the root of Big Data,or  the consolidation of a set of diverse data sources with a multitude of data types for which you’re attempting to determine relationships and patterns amongst it. Phew. Got it?

Me neither, but I like to think in examples and this is where it dawned on me in the grocery store.

I was walking through the local grocery store, a place with a very high number of items. In fact, in 2010, the average grocery store carried over 38,000 different items.  In some areas of the store, the placement of these items is logical.  Dairy is on the back wall so that it takes a very short trip from a refrigerated truck, to a refrigerated room which often opens into the milk cases.  Restocking the shelves seems to be a process of just lifting the 1Gal jugs off a pallet/shelf and putting it into the case, without having to wheel it out into the store.  Others not so much… I’ll get to one of them in a bit.

The store also contains seemingly unrelated types, sure they carry food and other edibles, but also flowers, hardware, cooking pans and yes, diapers.  I get it, it’s a one stop shopping location.  The problem the supermarket faces isn’t about “how do I store all these items on the shelves”,  but “how can I arrange the items such that buying item #1 puts the customer in the right mindset to buy item #2?”

So I continued in my shopping tour of the store and came down in aisle which had diversity and quantity in it, but I didn’t readily see the pattern.  There was a relationship between items on one side of the aisle, but I was having trouble seeing the relationship between both sides of the aisle. On the right side was flowers, balloons and greeting cards.  Never seen it fully stocked, as the only time I visit it is at 5:45pm, on my way home, on February 14th, and stand there looking at a 99% empty shelf with the 20 other doofuses that waited until the last minute, and pray “it’s the thought that counts” will save my procrastination. But I digress. All these items were pretty easily related.

The other side had baby food, diapers, baby wipes and other infant related items for which I had no idea what they were for (See Author’s note above).


So here we are, on one side of the aisle we have baby products and the other side we have cards, balloons and flowers.  What is the relationship between the items on the left vs. the right?  Seems like a problem for Big Data.  It’s one thing for a human to look at these and come up with a relationship, but it’s another to be able to feed this into a automated system and come out with a relationship.

So how does this fit into the definition?

  • Consolidation: One stop shopping
  • Diverse Data Sources: Different types of items
  • Multitude of data types: 38,000 items…
  • Relationships and Patterns: What do the items on the left and right side of the aisle have in common?


So what is the relationship between diapers, baby food and flowers/cards?  While this is solvable with a machine system and algorithms, what makes this interesting isn’t the Petabytes of data or racks of servers, but in uncovering the relationship in all that data.

In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.


  1. From Enterprise Architecture perspective the business problem you are trying to solve is to maximize revenue by optimizing investments in the infastructure and the placement of resources. Infrastructure investment such as hardware gear & stuff will take you only that far . The real fun is how to slice and dice the data , correlate it and makebit meaningful , that’s where the software comes in play. In other words there’s no choice but to move intelligence up the stack , the gear becomes secondary.

  2. I do believe the conclusion that those items go together could be done with big data. What the stores do is evaluate all the receipts for peoples’ purchases. At this point, they aren’t concerned with the location of the items. What these stores have probably discovered is that husbands could be in two situations that complement each other. 1) They were sent to the store for diapers and baby food because they ran out at home. Usually when there is not a continuous supply of baby staples it means something probably happened such as wife is having a breakdown/toddlers are driving her crazy. So when the husband is told to pick up diapers, he might buy her flowers/card/chocolate all of the above.
    2) He might be wanting to show up with a peace offering, and what better way to show you care, then to remember that you are low on diapers and buy. The trick is, some husbands won’t remember unless they turn around and it smacks em in the face. This is where the product placement comes in… If enough receipts show flowers/cards/diapers/baby food being bought together, and some receipts with just one or the other, then the store needs to do a better job placing their goods to be easily purchased. I’ve been on both those situations and another where I walk up and down the store for a long time looking for those items.
    3 cheers for Big Data!

  3. The idea is when you (husband) drop by the grocery store to pick up diapers and baby food, don’t forget to pickup a thank you card and flowers for your wife. I don’t think a compuer/software can figure this relationship out through data crunching. Needs human intervention.

    • Sukento, Thanks for replying. Do you think we could eventually create algorithms to take into account human nature?