Despite all the buzz around Data Science, a lot of retail companies today are still struggling to become data-driven. Working in the retail data analytics field for the last 5 years, I noticed that many retailers seem to think their hands are tied because of the limited amounts and types of data available to them: “I would like to be more data-oriented, but I need real-time sales data to do that.” “If collecting individual customer data wasn’t so expensive, I would be doing a lot of retail data analytics right now.” Due to this mindset, many companies become DRIPs: Data Rich, but Information Poor. Many companies have yet to realize the wealth and potential of the data they already have. Having more data is not always the solution — it is what you do with it that counts.

Having limited data is a big roadblock to achieving a data-driven culture, but it shouldn’t limit managers from making more informed decisions. With the right knowledge and application of Data Science and Analytics techniques, there can be a lot of quick wins for retail companies.

 

On Expanding Customer Basket Size

There are two ways retailers earn — by getting new customers or by getting existing customers to buy more. For businesses, a bigger basket size is also a form of validation that the customers are loyal to the brand. One way to expand average basket size is through market basket analysis, a technique used to look for associations among products by exploring customer transactions. By analyzing which items are purchased together, the following questions can be answered:

  • Which products should go together on store shelves?
  • Which products can be bundled together to boost sales for a less popular item?
  • Which products can merchandisers or field agents push to customers through smarter recommendations?

For example, companies will be able to see if customers who buy the shampoo product of a particular brand also buy the corresponding conditioner product from the same line. Some associations can even be surprising like the classic data urban legend about beer and diapers. The legend goes that a store owner discovered that male shoppers who buy diapers also tend to buy beer. By placing beers close to the diapers section, he was able to boost the sales for beer.

With the right infrastructure and resources, a market basket analysis can even be elevated to become a recommendation engine. These are models that power targeted digital ads highly customized to a customer’s preference based on past purchases, location, or interests. If only non-real time transactional data is available, a market basket analysis is the next best thing.

 

On Price Optimization

Pricing can be a sensitive topic for marketers. Increasing it means taking a hit on volume while decreasing it means taking a hit on margin. The most ideal scenario then is increasing the price without making a significant impact on volume. Many retailers with e-commerce presence adopt dynamic pricing, which involves changing prices in real-time according to current data on website traffic, competitor prices, demand, etc. You will see this in action with eBay and Amazon. However, dynamic pricing requires real-time data that is difficult to acquire. One alternative for retailers is a price optimization model that considers publicly available information. These kinds of data can be collected from the internet or from the field and can be combined with internal data on sales and advertising:

  • Competitor prices
  • Weather
  • Holidays or events
  • Inflation
  • Product attributes (size, brand, color)
  • Store location

The only caveat here is that manual data collection will require more time and effort, especially if the company does not have web scraping capabilities. Statistical methods might help in inferring how much data should be collected and how efforts can be minimized to ensure the data being collected is representative of the overall picture. If data collection is still not possible, a price optimization model based on just internal sales and advertising data will still enable data-driven decision making. 

Cluster analysis can also be used to infer pricing trends and elasticity. A simple clustering model can be used to find segments across products based on factors such as location, season, and product attributes. Are there products that can take a price increase because they do well in premium locations? Are there low-priced products with similar attributes to a higher-priced product? 

 

On Retaining Customers

Companies that have the means to track customer transactions, such as loyalty programs or online e-commerce accounts, can easily predict which customers are likely to churn by looking at patterns of churned customers. With a churn re-engagement strategy, companies can win back customers and prevent churn.

If you’ve ever received an email that goes like this- “We miss you. Here’s a coupon code you can use on your next purchase”-chances are you’ve been deemed as a likely churner. Re-engagement emails like this one from Missguided, a UK-based fashion retailer, are examples of reactivation strategies that target specific customers who are not likely to come back and shop again. The strategies are based on attributes such as last log-in date, length of time in between log-ins, time spent on browsing, count of items in the shopping cart, etc. Through a churn model, retailers can implement preventive measures to engage users at risk.

Example of a marketing email offering a special coupon encouraging customers to shop again
Here is part of an email from Missguided, a UK-based retailer, encouraging a customer to shop again using a special coupon code.

In the retail landscape, a customer churn model can be developed by considering the following:

  • Basket size and value
  • Days since last purchase
  • Days in between purchases
  • Day of visit (Weekend or weekday)
  • Categories of purchases
  • Seasonality

For companies that rely on offline transactions data from POS systems, attributing purchases to individuals may pose a challenge. It might be worthwhile to consider introducing a membership program or another similar strategy to capture customer information. Notice how SM Advantage users get text blasts on promos and discounts from time to time. Aside from knowing your customers, there is also great value in being able to reach them. Measures to engage retail customers who are likely to become inactive can easily be implemented with a membership program in place.

 

On Expanding Customer Segments

Another benefit of collecting customer information is having greater visibility on customer segments. Companies spend huge amounts of money to identify market segments. Product development and campaigns all depend on which customer groups are undervalued and untapped. However, in order to make sound decisions, retailers must also complement market research with methods in retail data analytics. Most of the time, company-specific customer segments do not really align with industry-wide market segments.

Internal customer segmentation can easily be done by analyzing customer and transaction information, such as:

  • Age
  • Income
  • Transaction Data
  • Purchasing Behavior

By doing this, retailers will be able to uncover behaviors that are specific to their customers. For example, a company would be able to design better products or campaigns if they know that 60% of their customers are young professionals who go to the store at least twice a week and have a minimum income of Php 60,000. Typical market segmentation may capture age and income, but rarely does buying behavior get considered as valuable data.

It might be hard to do a customer segmentation analysis for companies that do not have membership programs. One alternative is to do random surveys on customers during store hours or through social media platforms.

 

How Can You Get Started?

There are many ways to tap the potential of already existing data. A lot of the examples given above require transaction-level data that are usually already collected by companies to track aggregated sales. However, if even this information is not available, retail companies can consider other cost-effective ways such as surveys and manual collection. 

What’s most important to take note of is that all data, no matter how granular or detailed, have value. Businesses don’t need fancy AI models and bots to get started with Data Science. With a team of Analytics experts, Retail Data Analytics can be done with data that are already accessible. For companies without in-house Data Science capabilities, consulting with companies well versed in extracting value from data would be a good starting point to becoming more data-driven.

The DRIP culture is a thing of the past. Companies already have the power to achieve more — they just need to unleash it.


Justine Guino is a Senior Data Analyst at Amihan Global Strategies. She specializes in helping our clients unlock tangible value from data. When not working, you’ll find her planning for her next travel adventure. Connect with her on Linkedin (https://www.linkedin.com/in/justinecg/).