Five Tips on Big Data Optimization for Retailers

If you’re looking for one of the oldest and largest sectors in the economy, look no further than retail. Since the 1990s it has surpassed the automotive industry as the thought leader in operations. In fact, you’d be surprised at just how many business process innovations have come from retail and been applied elsewhere. Take cross-docking practices from Wal-Mart, or e-fulfillment from Dell and Amazon as a couple of examples.

Retail has once again been reinvented over the last decade, mainly due to the fast fashion ideas from Inditex. Ideas that drive frequent changes and quick-response production and distribution for retailers. With this in mind, we find that the retail landscape has become very turbulent, with short life-cycles and uncertain demand patterns. Consider for example the typical demand for a dress at a mass-market fashion retailer. Unless it is a “classic,” demand usually peaks when the product is introduced and quickly declines. Similar patterns are now found in products as diverse as noodles!

To adapt, retailers are investing in technology to monitor changing trends. Besides transactional systems (point-of-sales transactions), measuring store traffic figures, from simple visitor counts to very detailed heat maps and positional records from 3-D cameras has also become common practice. This provides stores with large amounts of data – big data – that can reveal useful and actionable information. For example, in one study with a large retailer, we found that increasing inventory at the store, even beyond the standard service level that the retailer uses, increases sales, because of better displays (see our impact inventory).

It is also possible to establish the impact of external shocks on sales. For instance higher temperatures reduce the number of visitors to the store and may increase or decrease conversion depending on the category. These insights can be used to take better decisions regarding stocking, merchandizing and promotions.

Not only that, there are now sophisticated methods for tracking individual customer behavior which is close to mimicking web analytics. Cookies allow firms to know customer demographics as well as interests from past click history. They help companies such as Google to monetize traffic by matching advertisements to customers’ needs. When it comes to actual stores, new technologies allow retailers to do something similar. Loyalty cards can track past purchase behavior and customize pricing to individual consumers, through coupons for example, successfully implemented at Tesco in the UK or Caprabo in Spain. Unfortunately, this tool is only viable for large players and for loyal and frequent shoppers.

In retail categories where conversion is low, such as fashion, these systems are not that useful, because they only track purchases and are unable to detect or track customer visits without purchase. But nowadays there are alternatives. One option is to track customer smartphones, via Wi-Fi, which allows a retailer to know when and how long a customer visited, and whether he or she made a purchase. This data provides a sort of Customer Relationship Management system for all customers, across stores, which can then be matched with credit card or other types of data.

To navigate the turbulent waters of retail, and effectively use the big data from day-to-day operations, our experience indicates that there are 5 key questions that a retailer should consider, for which we can offer some suggestions:

1. Are you investing in the right technologies? There are many options available in the market, from RFID tags to beacons. It is critical to clearly understand their effectiveness. For example, the low percentage of shoppers with Bluetooth activated on their smartphone currently makes Bluetooth sensors limited in use.

2. Can the data be useful? Storing big data consumes resources – think of Hadoop. At the same time, if a retailer is not going to ever use it, perhaps one should reconsider both collecting the data and storing it. At the same time, there may be future applications that will make the data valuable. Retailers should develop a clear view of what sort of competitive advantages can be developed from big data. For instance, RFID tags can be used to track which products go to the fitting room vs. those purchased. A discrepancy between these two data streams would indicate that there is something wrong with the product fitting.

3. Do you have the right analytics capabilities? Extracting value from big data requires specialized skills: mathematicians, statisticians and computer scientists. These profiles are scarce and moreover they are usually not well trained in business. A good way to overcome this challenge is to partner with a specialist – either an analytics firm or a university. This is a good start that will reveal the potential for a retailer. After this first step, an internal analytics team can be built.

4. Are you translating big data insights into action? Having a strong analytics team is useless unless it can influence and move retail execution. Fluid communication is necessary, where operations challenges are clearly exposed to the analytics groups who respond with suggestions that can be implemented. For example, retailers usually identify store entrance displays as most profitable: what products should be placed there? The best-sellers, to increase their success even more? Or the not-so-good sellers, which need some help from these displays to increase sell-through? This is a question that can be answered with data.

5. Are you willing to experiment? Generating improvements from big data is a continuous process that requires continuously testing the recommendations from analytics. It requires patience to sometimes make mistakes, which are very valuable to confirm or invalidate new ideas. A common framework to do this is to use A/B testing: implement a change in a set of stores or during a few days, and compare the performance with the control group, those stores and days where the standard practice was used.

In the coming years, retailers of all sizes will have the opportunity of improving their business models thanks to the democratization of big data analytics as a service and the introduction of affordable data capture technology at the point of sale. And this is already underway.

Authors: Prof. Víctor Martínez de Albéniz and Álex Herrera (DatActionS)