Subscribe to DSC Newsletter

Increasing customer loyalty using data analytics

Loyalty Marketing has become a key strategy for most companies in today's competitive marketplace. The practice is based on a very simple premise - as you develop stronger relationships with your best customers, they will stay with you longer; the longer they stay, the more profitable they become.

It costs less to retain a customer than to acquire a new one. To retain a customer involves many factors - personal relationships, product quality, customer service, price, and other brand values.

Loyalty is relationship-based, focused on the overall customer experience and is predictive of how a customer will behave in the future. The most reliable predictor of how your customer will behave is the strength of that relationship and the secret to measuring it lies in customers’ past behavior.

Fractal has developed new, proprietary mathematical models to measure loyalty with greater reliability than traditional methods, provided that past transaction data (Invoices, orders, or other customer interaction data) is available.

Fractal’s methodology examines the entire available history of customers’ purchase behavior and interaction with the firm. It enables one to rank customers by loyalty score, and segment them into groups based on their behavior.

These rankings reveal trends, uncover purchase patterns, and help predict customer’s future purchases, and hence help forecast company’s revenue. Loyalty Builders’ reports will identify the customers whose loyalty is slipping and customers whose loyalty is growing - the knowledge that can help suggest the most effective way to serve the customer who shows up in your showroom, at your website, or your call center.

Traditionally, the approach has been to segment a population into separate ‘buckets’ based on demographics and attitudes. Fractal’s methodology, however, segments customers by behavior and generates quantitative measures of loyalty based on that behavior.

This behavioral data already exists in the order entry or accounting systems. Every new transaction adds to the data set and hence the learning can be continuous. Most important, the parameters of loyalty in the Fractal model create a multi-dimensional picture of the customer and produce segmentation that illuminates behavior patterns, which may otherwise go unnoticed.

In the realm of customer behavior, the kinds of information most often requested are account-level predictions:

Which customers are likely to buy again in the near future?
What products or services are they likely to buy?
Which customers are potential defectors?
Which customers are good prospects to make an incremental purchase above their usual purchase rate?
Which customers are good candidates to buy a type of product that they have not previously purchased?
What is the lifetime value of a particular customer?

Customer satisfaction surveys, which are backwards-looking measures, can’t deliver this kind of information. Satisfaction is quick to grow and just as quick to disappear. It’s tied most closely to a customer’s last transaction or interaction. In contrast, customer loyalty is slow to grow and slow to fade away. Because loyalty analysis predicts future behavior, it can answer the questions above.

What does one like to know about customers’ future behavior? How valuable would this kind of information be to a company?

Depending on one’s unique marketplace situation, one may have a variety of Loyalty Marketing objectives. These include:

· Frequency – driving customer purchase frequency
· Retention – improving customer loyalty strategies, creating barriers to exit, preventing defection of valuable customers
· Relationship Building – developing two-way communication and enhanced customer learning, and improving customer satisfaction
· Advocacy – creating a loyal customer who promotes your brand and refers new customers

Whether one is trying to retain customers, motivate them to increase their purchase activity, trying to establish loyalty relationships, or all of the above, the basic principles of Loyalty Marketing rely on four key components:

1. Dialogue Marketing – the right messages delivered the right way to the right people at the right time

2. Customer Rewards & Benefit Programs – an effective platform of earning and reward offers with a broad selection to match individual customer needs and preferences

3. Customer Behavior Tracking – a systematic approach to tracking and storing customer spending and response behaviors and integrating the right mix of communications and rewards

4. Measurement – a plan to track and measure key performance objectives and customer retention analysis data against loyalty program objectives

Views: 439

Tags: churn, customer retention, fractals, lifetime value, loyatly

Comment

You need to be a member of AnalyticBridge to add comments!

Join AnalyticBridge

Comment by Vladimir Lukin on May 14, 2008 at 11:32pm
I wonder why I completely agree with Mark and Satindra's comments...

My 2 cents are (maybe from slightly different angle):
I wish that data guys and marketers talked to each other more often.
I believe that better, having more precise statistical models require more elaborate, fine tuned marketing strategies. Too often requests from marketers come as too generic - give us churn models, response models, etc., so our campaigns would work It is true that data may (can) reveal specific marketing segments or groups of customers for which campaigns work better without even knowing details of those campaigns, but, more often, it is just easier simply to ask and read that mail piece or look at a web page. Sometimes, it is apparent that the campaign itself is useless, and models that find significant lifts in those campaigns are bound to be of Type III error, or "...giving the right answer to the wrong problem" .
http://en.wikipedia.org/wiki/Type_III_error#Various_proposals_for_further_extension (see Kimball)

In short, if there is a chance talking to someone who is designing a marketing campaign - I would not waste it.
Comment by Satindra Chakravorty on May 13, 2008 at 10:42am
I would like to emphasize the bit about differentiatied retention offers/messages. Often, a client will proclaim the existence of a churn problem within a customer portfolio and either independantly or with an assist from some analytic group, will prescribe a churner detection tool, as a remedy.

We know that loyalty exists in grades and a churn score signals the future loyalty grade of a particular customer. However, we also know that not all customers (loyal or otherwise) are the same. Some will respond to a message about the business/product/service as a price leader; others need to be reminded about the importance of their affiliation with the business, to the business and yet others might want to be informed about product innovation (while all will probably appreciate a x% off offer on their next visit!). In the absence of a set of different, yet relevant (to subgroups) retention treatments, it is difficult to build sustainable loyalty; it is easy enough to accelerate the next purchase.

Such marketing communications should be periodic; not just before a business recognizes a customer is signaling churn. In some portfolios, it might not be too early to think about churn (and therefore customer loyalty) at the moment of customer acquisition (just as we know that among societal portfolios, life expectancy is affected by factors related to birth and the birthing environment).

As I see it, building customer loyalty is a process of maximizing the number of opportunities to derive revenue from a customer. A customer is not equally motivated to cooperate along the progression through his/her customer lifecyle. Without this knowledge (gained through the analysis of customer data), a marketer risks being ineffective (at best) and possibly counter-productive by communicating the wrong message to the wrong customer at the wrong time!
Comment by Mark Richards on April 26, 2008 at 12:49pm
The last 'competition' on churn modeling I saw was won by Salford (who win quite a few of these challenges). http://www.fuqua.duke.edu/admin/extaff/news/teradata_churn_2004.htm

Anything more recent than this one in 2004?

In my experience, predicting churn is fairly easy (particularly when one has access to customer behavior &/or transaction data). The really difficult thing is to come up with a strategy (and appropriate offers) for having a positive effect on churn behavior. This is a LOT harder than people seem to realize or at least admit. Some offers / interventions don't do anything (a free pen!), others don't make sense from an overall portfolio perspective (strong offers are accepted by your false positives' while your true positives still defect at an alarming rate - the ROI isn't there).

IMO, the appropriate approach is to model to incremental churn specific to an offer or set of offers. Other approaches I've seen work well are cross-selling (multiple relationships increase stickiness) and peer acquisitions (i.e. rewarding your existing portfolio for acquiring new accounts - the acquisitions are better from a # of different metrics, including loyalty and it has been shown to increase the loyalty of the current portfolio who participate).

On Data Science Central

© 2019   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service