Customer data analytics — which include customer data models, segmentation and reporting — form an important aspect of customer marketing.

The Four Stages of Customer Data Analysis

There are four stages to any customer data analysis project:

Data Discovery – planning and gathering information, performing a data audit and presenting results
Exploratory Analysis – performing analyses that will give an immediate look into your customers by identifying trends and segments
Action – leveraging additional models to enhance performance management, explanation of the data and evaluating the impact
Prediction – creation of models that will help explain what will happen based on previous performance and using that information to make things happen/improve

Data Discovery

  • Planning and Gathering Information
  • Data Audit
  • Presentation of Results

Exploratory

  • Decile Analysis
  • Loyalty Tracker
  • Migration and Change Analysis
  • Cohort Analysis
  • Segmentation
  • Campaign and Contact Management

Action

  • Next-best-sell
  • Market Basket Analysis
  • Reactivation Model

Prediction

  • Attrition Model
  • Best Customer “look-alike” Model
  • Customer Acquisition

Stage 1: Data Discovery

Before beginning an exploratory customer data analysis, you’ll want to perform an audit of existing available data. This step will familiarize you with the data environment and allow you to complete initial diagnostics on the data in order to understand the quality and quantity of data available for analysis.

The audit begins with a short planning and information-gathering process where stakeholders meet to review key business issues and challenges, discern what data is available, understand how the data is currently leveraged and identify how the results of current marketing initiatives are measured and reported. Coming out of this planning session, you will also be able to identify the data sets to be included in the audit.

The purpose of the audit process itself is to:

  • Evaluate and assess the integrity and completeness of data collected
  • Clean and dedupe data (as required)
  • Produce initial frequency reports
  • Provide insight into the available data fields
  • Identify key variables that might be useful in creating segments
  • Identify gaps in data (with recommendations for third-party overlay data if appropriate)
  • Identify and create variables that will be useful for the initial analysis
  • Identify other key insights

At the end of the data audit, specific analytical opportunities are identified and prioritized for the exploration stage. If appropriate, all or part of the analytical data model required for the initial exploratory analysis and segmentation are also created or identified.

Stage 2: Exploratory Analysis

There are several types of customer data analysis projects and reporting that are beneficial during the exploratory phase.

Decile Analysis

Decile analysis gives a preliminary insight into your customer base so you can begin to identify trends among your highly profitable customers. Identifying these customers will help with future analyses.

Decile analysis is a preliminary customer segmentation tool that looks into your customer base and splits customers into 10 equal parts. You can calculate a variety of variables, such as number of transactions, discounts, refunds, days since last order, and maximum, minimum or average expenditures.

Shown below is an example of a customer decile analysis. Here, the first decile represents the company’s most profitable group of customers, while the 10th decile represents the least profitable. Deciles one through three show a significant contribution to the company.

*Click to enlarge

You can dive into deeper explorations of the decile analysis to find additional insights. For instance, by examining transactions more closely, you can identify seasonality trends. These analyses help to better understand your customer base and provide direction on where the marketing department should focus its resources.

Loyalty Tracker

Loyal customers make more frequent purchases, spend more, and tend to buy a wider selection of products and services. The following five common loyalty indicators may be more or less present depending on the data discovery process:

  • Dollar amount of purchases
  • Time gap between visits
  • Visit consistency
  • Re-purchase rate
  • Duration of customer relationship

The loyalty tracker customer data analysis develops a loyalty score on these five attributes by applying weights to each indicator and calculating an overall score. The purpose of the tracker is to systematically quantify the loyalty of customers and monitor changes during specific tracking intervals. Identifying these changes can help to develop marketing campaigns to intervene when customers appear to be drifting away. The benefit of the loyalty tracker is that these scores can be incorporated into additional analyses, such as segmentation.

Migration/Change Analysis

Coming out of the decile analysis and loyalty tracker, you might next want to examine how customers’ behavior changes over time using a migration analysis. This will help identify which customers have the most stable purchasing patterns.

For instance, which customers have been growing their business with you over time?  Which customer’s sales are in decline, and which customers have defected? This information can be very helpful in developing more targeted and relevant marketing communications.

In the following example, 53 percent of a store’s Premiere members demonstrated a decline in spending in the first half of the current year versus the previous year. This indicates a need to develop programs aimed at protecting the value of these Premiere customers over time. In the same example, 24 percent of the store’s High Value members and 56 percent of Medium Value members are inactive. This shows that reactivation is another opportunity that should be pursued.

Cohort Analysis

Cohort analysis is used to look at distinct groups of customers and how they perform versus other groups of customers. Many of the preceding analyses may be repeated for specific cohorts.  For example, new customers often behave differently from existing customers. Understanding these differences can help guide new welcome and onboarding initiatives.

In the example below, new members represent 30 percent of the store’s active member base but only 17 perent of spend. Existing members spend more, buy more tickets and visit more often than new members. However, new members buy more tickets and make more visits than existing members. The significant number of “unknown” sex indicates a potential data collection issue. The results suggest the need for more concerted marketing efforts to accelerate sales growth among new members.

Basic Segmentation

Identifying customer segments and monitoring their behavior over time can lead to an increased understanding of a company’s customer base. Segmentation can occur on a multitude of levels. Based on information gleaned from the above analyses, you may be able to segment customers into distinct groups. The rule of thumb is to aim for six to eight segments. But this is highly dependent on the data.

First steps would include performing a basic analysis on the data given to form simple segmentation, such as recency/frequency/monetary (RFM), category and value (e.g., value placed by customers). Down the road, you could perform advanced segmentation concepts to fine tune segments and/or create new segments. You can then update your previous analyses to reflect these new groupings.

Below is a segmentation example showing how percentiles, scoring and aspects from previous analyses can help enhance your understanding of how each segment performs.

*Click to enlarge

Campaign/Contact Management Analysis

Depending on the availability of data, the exploratory analysis step may also include some initial analysis of past marketing campaigns and how they have impacted customer behavior. Ideally, this would also reveal insights about how the different customer value segments respond to marketing initiatives.

Stage 3: Action

Once you have basic segmentations in place and understand how different customer segments perform, you can create measureable goals. Explanation of these goals and the evaluation of their progress have more meaning at this stage, since the results from the previous analyses help frame and point the direction to take. Following are three additional analyses that you can perform during this stage to help facilitate an actionable approach.

Next-Best-Sell Model

The purpose of a next-best-sell model is to gain insight into the types of products that existing customers are most likely to acquire in the future. The basic premise of this customer data model is that customers who have purchased certain products will also have an affinity to purchase other types of products.

Next-best-sell modeling seeks to determine the next most logical product to promote to a customer using information you already know about the products and services the customer already has. Depending on the range of products available and the penetration of products into your existing customer base, rankings can be formulated in order from highest to lowest next-best product to offer a customer.

Shown below is an example of a next-best-sell model. As can be seen, customers in Segment C had the greatest likelihood to make their next purchase from the plumbing category.

Market Basket Analysis

The market basket analysis is similar to next-best-sell in that it uses logic and enhanced reasoning to decode and organize patterns of customer purchase behavior. This analysis broadens the customer data analysis performed on next-best-sell and attempts to add additional rules or patterns to the way customers buy.

For example, if customers in Segment C buy plumbing, then they also may tend to buy appliances. This analysis will look at the strength of the rule and the confidence that it would happen. The benefits of these two analyses working together are immense, as they will help dictate how to best communicate to customers and can be used to help facilitate actionable outcomes for goals set by management.

Reactivation Model

The purpose of a reactivation customer data model is to find lapsed, or inactive, customers who could be converted back into active customers. A reactivation model is typically applied to existing customers who have been inactive (not made any purchases) for a specified minimum time period.

Customer data analysis in this area will leverage insights gained from previous analyses. However, the scope of the project will focus on identifying not only customers who are inactive, but also customers who have a good chance of being reactivated into profitable groups.

Stage 4: Prediction

With the customer data analytics foundation in place and evaluations possible, opportunities to look down the road are promising. Understanding how customers will respond can help you stay one step ahead of your competitors and meet the needs of your customer base. The following customer data analysis examples show what can be done during this stage and will be dependent on the data available.

Attrition Model

The purpose of an attrition model is to pinpoint those customers who are most likely to discontinue transacting any business with your company. Attrition occurs when a customer has decided that they no longer need your service, has defected to a competitor or has become inactive for a long period of time with no indication of returning.

Attrition models look for key transactional indicators that are potential precursors to a customer actually drifting away. These can include downward trends in purchase activity, dissatisfaction with a particular product or service, and other signals that indicate impending fall off in customer interest.

Best Customer “Look-Alike” Modeling

The purpose of a best customer “look-alike” model is to identify existing customers with the characteristics of your best customers, but who have not exhibited transactional behaviors that qualify them to be classified as a best customer. Characteristics typically used in the model include demographics, lifestyle, life-stage or other descriptive data elements that can be collected from customers or appended to a customer database by a third-party data provider.

The theory behind a best customer look-alike model is that certain customer attributes can help to explain why a particular customer is one of your best customers. This customer data model requires that you have developed a quantifiable definition of a best customer based on past customer purchase behaviors.

Customer Acquisition

The purpose of a customer acquisition model is to identify the most attractive new customer prospects from a larger universe of prospect names. Most customer acquisition models begin with an overall profile of your customers compared to a profile from a pool of prospects. The prospects are typically available from a third-party list source.

The names from your customer database are matched up against names from the pool of prospects. Data is appended to the names that are common to both files. And a profile is then created to highlight the characteristics that are most descriptive of your current customers.

The model takes into consideration this information to score prospects on the basis of how closely their characteristics match the characteristics of existing customers. The best matching names are considered to be the best prospects to target.

Your Partner for Customer Data Analytics and Analysis

CCG’s team of data analysts and retail marketing experts can guide you through the four stages of customer data analytics, cutting through big data clutter to bring you quantifiable, actionable insights and real-life recommendations.