Skip links

Harnessing the power of Big Data in Retail

More and more companies are migrating to data-driven models in order to survive; in other words, they are basing their decision-making on data.

The term that defines a great volume of data is BIG DATA, also called data intelligence. The size of the volume that we handle nowadays is impressive:

  1. Every minute 300 hours of video are uploaded to YouTube.
  2. More data has been created in the last 2 years than in all of human history.
  3. About 40,000 Google searches are performed per second.
  4. Each person generates about 1.7 megabytes of new information per second.
  5. In the next 5 years, there will be more than 50,000 million smart devices connected worldwide, all developed to collect, analyze and share data.

The beginning of Data Science was closely related to industries that needed high precision as well as technology projects, such as the ones we find within the Fintech sector to minimize risk investments or avoid fraud in debt operations. However, Big Data is applicable and necessary for any company that works with data: from retail, education, or even the healthcare industry, as recently seen with macro analysis and forecasts of the COVID-19 pandemic.

For the FMCG sector, the Data-Driven model is not an exception. One of the great advantages of big data for this sector is the improvement in forecasting demand. 

Accuracy in forecasting demand is essential to establish a production plan, and resources. As I usually emphasize to my students:

“Any deviation in demand forecasting is expensive, even more than they seem”.

If we have made a pessimistic forecast, below its possibilities, and we end up having stock breakage, customers, manufacturers, and distributors get harmed. I like to call it the triple loss: the manufacturer loses, due to the direct loss of sales opportunity, the buyer loses due to the bad experience of not finding a desired product, and the distributor also loses since the consumers, by not finding the desired product, can replace not only the product but abandon their entire shopping cart completely, thus losing the total amount of the ticket and the customer loyalty.

If otherwise, due to an optimistic deviation in the demand forecasting we overstock, we will face other inefficiencies equally unpleasant or even more serious than the previous ones: stocking is expensive. Most food products are perishable, even some need extreme cold temperatures to be preserved, making stocking and storage even more expensive and consequently increasing inefficiencies and carbon footprint.

The more variables we manage to parameterize, the closer we will be to minimizing deviations between forecast and actual sale, we’ll have fewer incidents and we’ll be closer to a triple win: consumer, manufacturer, and distributor.

Not so long ago, demand forecasts were built based only on internal data from historical sales. Now we have the opportunity to enrich these forecasts by taking into account external variables directly related to the demand for our products.

For example, for a pizza maker, it is important to consider the schedule of the football league matches as we know there is a close correlation between the cross-consumption of both products. For a soup manufacturer, it will be relevant to include the weather forecast in their big data, since with the rise of temperature, their product will be less attractive and the demand will decrease accordingly. Finally, for a coffee maker, it would be interesting to include variables such as restrictions of the hospitality industry across the country. With consumption on the go and in restaurants reducing, the coffee consumption at home can increase and thus positively affect the sales forecast within the retail industry.

The need to fine-tune the demand forecast to avoid inefficiencies affects not only mass consumption but also the new business models that have appeared: Companies that offer food delivery services must anticipate how many employees need to be available or companies that offer on-demand cab-hailing services such as Uber or Cabify, need to know how many drivers are working each day and in which time frame. Data such as weather forecasts, public transportation schedules, or scheduled events can be useful data to their business.

The new data-driven models allow us to match all series of variables, minimizing deviations and avoiding extra costs for manufacturers, consumers and distributors.

What is really attractive about Big Data is not the amount of data that it contains but what organizations do with this data. It is estimated that only 0.5% of all existing data is actually used and analyzed, so the question is no longer the potential or the permanence of this new revolution but rather whether we will know how to adapt to the necessary speed.

Natàlia Planas

Director of the Postgraduate course in Big Data at the University of Girona.
B2C Marketing Nespresso at Nestlé.


Leave a comment