This beginner’s guide to retail analytics will help you get the most out of the data you’re probably already collecting.
When people shop online, they know they’re sharing personal data to be collected and used for marketing purposes. And yet we’re surprised by how many companies fail to do even the bare minimum of analysis with the data they’ve painstakingly gathered. By implementing just a few basic analytics techniques into your marketing practices, you can quickly gain insight into your site’s performance, your customers’ satisfaction, and the effectiveness of your marketing campaigns. And if you take your analytics a few steps further, you can actually predict what your customers want and gain specific recommendations about how best to market to them — whether you have hundreds of customers or millions of them.
Here, we’ll show you the basic building blocks of retail marketing analytics. The insights gathered from these techniques can help you optimize business strategies and increase revenues. We’ll show you which metrics matter the most and how they can be used to measure the success of a retail enterprise. We’ll also shed some light on what’s possible once you establish the basics: the more advanced analytics techniques that deliver predictive insights and recommended actions. Your crash course starts now.
Cost per Acquisition
Cost per Acquisition (CPA) is a vital metric that gives insight into the cost of acquiring a customer. Consumers typically don’t arrive at your site by chance; rather, they have seen or heard about your brand somewhere — whether via a referral by friends or family, on social media, in search engine results, or through display ads. Calculating CPA is important to gauge a marketing campaign’s impact.
The CPA metric is determined by dividing the amount of money spent on marketing your brand, including any discounts offered as well as the time and cost of support, by the number of customers who made purchases. For example, if a merchant spends $1,000 on an ad campaign that brings in 50 purchasing customers, the CPA is $20.
CPA = $ amount spent on marketing
# of purchasing customers
To obtain a more precise model and gain the most value from CPA analytics, retailers could analyze the cost of acquiring customers in specific channels, such as social media, organic search, paid search, mobile search and display campaigns. This way, retailers can identify the specific channels and campaigns that perform best and adjust strategies, including the financial and time resources put into each of these areas. Down the line, you can use CPA along with a multiple linear regression algorithm to determine a trend against various predictor variables such as marketing budget. You can also apply nonlinear optimization techniques to maximize the effectiveness across all acquisition channels, improving return on investment.
Lifetime Value (LTV) is focused on keeping customers rather than acquiring new ones. This metric helps determine the true value of repeat customers. LTV not only shows how loyal a customer is, but also the economic value of the customer to a business. Retailers should determine LTV for their entire customer base or specific customer segments. Here is the formula for calculating LTV:
LTV = (average # of purchases per year) (average order alue) (average lifetime of customer)
Many retailers assess the LTV of a customer segment in specific increments such as 6, 12, or 18 months. Weighting factors could be incorporated into this model, such as the length of time a customer has been on record or the number of opportunities the customer has had to make a purchase.
A possible opportunity for using the LTV metric is to deploy a nonlinear model such as polynomial regression, in which the relationship between the predictors and the response is nonlinear. This would tell you, for instance, whether a customer's value changes with the season. Then you can use that information to increase marketing campaigns during low activity seasons.
If it turns out your LTV is low, it could be due to a high Churn Rate. By calculating this metric, retailers can discover how many one-time customers they have. To determine churn, retailers can start by identifying a cutoff date, so they can consider customers who have not made a purchase in that timeframe — a condition defined as a churn. The best way to select a cutoff date is to take into consideration data such as an e-commerce site’s repeat purchase rate. For instance, if most loyal customers make repeat purchases within 60 days, any customer who has not made a second purchase within that period is considered churned. The formula for determining churn rate is:
Churn Rate = # of one-time customers
000000total # of customers
By identifying churned customers or customers who are about to churn, merchants can take steps to win them back. For example, merchants can segment customer groups and email promotions to encourage purchases. Moreover, merchants who notice their churn rates increasing over time can take steps to improve customer loyalty, whether it be through more personal marketing strategies or implementing more rewarding loyalty programs.
It is also possible to predict when customers are about to churn, using a logistic regression machine-learning algorithm, for example, allowing you the opportunity to take some sort of action to prevent it, like offering a special discount.
The X Factor of Customer Loyalty
The ability to create emotional bonds with your customers is the X Factor that drives customer loyalty. It is a calculated value that characterizes a metric that correlates with higher spending levels and lifetime value. Essentially, the X Factor determines the relative range of customer loyalty and what behaviors or channels influence that loyalty. Here is one possible formula for X Factor:
X Factor = (# of purchases over past 12 months) (total sales)
00000000months since last purchase
To calculate a more rigorous X Factor, retailers can add weights to related purchase events such as the number of product categories purchased from, the number of items purchased in an order, or non-purchase metrics, including page views, cross-channel interactions, and social media activity. By scoring these classes of events, retailers can build loyalty scores for their customers that can help with marketing initiatives. For instance, a retailer could offer a promotion for free overnight shipping to all customers within a certain score range in an effort to further influence conversions and thus increase their LTV.
Although quantifying the X Factor is useful to see a relative range of values, you can go a step further by pairing the X Factor metric with a simple probabilistic classifier like Bayes to predict whether a customer is loyal. This allows you to get a yes or no answer for each customer (loyal or not loyal). And such a specific and concrete answer can lead to specific and concrete marketing actions, such as rewards for loyal customers or knowing which customers are not worth the cost of trying to attain loyalty.
Building a Retail Analytics Web App
One method of exposing the above metrics in a convenient way for management decision making is to build a simple Web-based application. Here is a recipe for developing one or more apps to track metrics over time:
- Use the R statistical environment to access data in your marketing department data mart or enterprise data warehouse. You can easily do this using the sqldf package for R.
- Perform some exploratory data analysis to discover factors like cutoff date for determining the churn rate described above.
- Define a set of R functions to calculate the metrics like LTV over a selected time increment (number of months).
- Use a linear regression model to detect trends, or build a decision tree classifier to detect likely churn.
- Use the Shiny tool to expose your R scripts and plots on the Web for management to interact with and make decisions.
Intelligent retail analytics can — and should — become an integral part of your company’s decision-making process. By digging deep into your data, you can maximize revenue and improve performance of your e-commerce initiatives. The key is to know what data to look for and how to use it effectively.
Whether you’re just dipping your toe into analytics or are ready to take your business (and your data) to the next level, there’s a lot more to learn. Our free paper, “Transformative Analytics: The next level in predicting and shaping consumer behavior,” shows you additional advanced analytics techniques and explains how they can not only help you monitor your marketing strategies but guide future strategies for improved sales and revenue.
Daniel D. Gutierrez is a Los Angeles–based data scientist working for a broad range of clients through his consultancy AMULET Analytics. He’s been involved with data science and Big Data since long before it came in vogue, so imagine his delight when the Harvard Business Review deemed “data scientist” as the sexiest profession for the 21st century. He is also a recognized Big Data journalist and is working on a new machine-learning book due out in later this year.