Shopping for hats in London,1942 (Source: Wikimedia Commons)
Ideas for Leaders #651

What ‘First Impression’ Data Reveals About Customers

This is one of our free-to-access content pieces. To gain access to all Ideas for Leaders content please Log In Here or if you are not already a Subscriber then Subscribe Here.

Key Concept

Companies can form a ‘first impression’ of a customer based on the information collected during a customer’s first transaction with a company. This first impression data can help companies predict how often new customers will purchase in the future and how much money they will spend on each transaction. It can also help companies target their marketing campaigns more effectively.

Idea Summary

Acquiring customer data at the point of purchase, including from first-time customers, is not new. However, much of that data for first-time customers often focuses on where the customers came from (e.g., how did you hear about us?). New research by Nicolas Padilla and Eva Ascarza of Columbia Business School reveal that customer behaviour related to a first purchase — for example, whether the purchase was made online, or whether the new customer bought a newly introduced product — can yield extensive data about what new customers will do in the future.

The research was based on a data set from an international retailer that sells beauty and cosmetic products in their owned stores or online. The data covered four years of transactions of more than 13,000 individual customers, beginning with their first purchase.

Padilla and Ascarza began by dividing customers into six different customer types based on their first purchase behaviours. Thus, customers were divided among those who

  1. bought their first product during the holidays (labelled ‘Holiday’ customers);
  2. bought a greater number of products than the average customer (‘Quantity’);
  3. bought a new product that had just been introduced (‘New Product’);
  4. purchased their first product online (‘Online’);
  5. paid higher prices on average (‘Average Price’); and
  6. bought items that were heavily discounted (‘Discount’).

Padilla and Ascarza then observed how the different customer types behaved in the years following the first purchase — specifically 1) how often they returned to buy more products, 2) how much they spent on each transaction, and 3) how long they stayed as customers (attrition in retail is common).

The analysis yielded a wide variety of information. For example, customers who first bought from the company during a holiday season were more likely, in the future, to purchase less frequently from the company, spend less on average on each purchase, and eventually disappear. In contrast, customers whose first purchase was a new product were likely to purchase more frequently than the average in the future, spend higher amounts per purchase, and ‘churn’ — that is, leave the firm — at an average rate.

The next question for Padilla and Ascarza was how these different customer types responded to marketing initiatives. The researchers defined four different types of ‘marketing’ variables — email campaigns, direct marketing campaigns, the introduction of a new product, and holiday shopping — then used statistical modelling to measure how each customer type responded to each of these variables. The following is a small sample of the results:

  • Email marketing campaigns encourage ‘online’ customers to buy more frequently, but not to spend more on average for each purchase.
  • All customers except for ‘online’ customers and ‘new product’ customers were likely to spend more per purchase after they had been targeted by a direct marketing.
  • The launch of new products had the most impact on ‘quantity’ customers, less impact on ‘holiday’ customers, and no impact on ‘online’ customers.
  • Holidays are especially attractive to ‘new product’ customers, increasing their purchasing transactions and spending levels the most.

By taking the predicted future behaviours of the different customer types and factoring in their sensitivity to different marketing campaigns, new product introductions and holidays, Padilla and Ascarza calculated the customer lifetime value (CLV) of the different types of customers. Customers acquired during the holidays, for example, have the lowest CLV. Boasting the highest CLV, but only by a slim margin, are the customers who paid the highest prices for their first purchase (the ‘average price’ customer type).

Finally, Padilla and Ascarza showed how the first impression data could be used determine how best to market to just-acquired customers. Using data from a separate set of 2800 customers (to maintain the integrity of the results), the researchers simulated three different scenarios of future marketing: no marketing, current marketing and first impression-based marketing. From these simulations, Padilla and Ascarza found, for example, that first impression-based marketing was especially effective for customers who purchased online, followed by those who bought heavily discounted products and those who bought during the holidays. Combined with the data from the first phase, the research suggested that email campaigns should prioritize customers who had bought their first products online. Direct marketing campaigns, on the other hand, should prioritize customers who had bought discounted items in their first purchase, or who had first bought the company’s products during the holidays.

Business Application

In the Information Age, companies are inundated with raw data about their customers, and realize that this data needs to be mined, analysed and leveraged to ensure that they keep their customers happy, loyal and active. The contribution of this research is to reveal a new set of data that companies may have overlooked in the past. Companies can now calculate the customer lifetime value of new customers, predict the future behaviour of those customers, and know which customers to target with which marketing strategies based on information that is usually available, but often ignored.

Contact Us

Authors

Institutions

Source

Idea conceived

  • February 2017

Idea posted

  • April 2017

DOI number

10.13007/651

Subject

Real Time Analytics