Outlier Analytics – Learning From Those on the Fringe

While most companies focus their business intelligence efforts on the masses, those few examining outliers (consumers who don’t exhibit expected behavior) are finding hidden gems of information they are using to develop new offerings...

What?

Unexpected events, extreme values, or disturbances are common occurrences in the world of data analysis, occurrences which are labeled as outliers by those practicing in the field. Ask any statistician or a data miner what he or she would do with outliers (e.g. a customer makes 100 calls to a contact center in a week, or, a small business makes thousands of bank transfers in a month), you’ll usually get the same answer – “I’d remove them from my analysis.”

Although this practice of removing outliers is, in fact, the prescribed action when it comes to conducting a traditional analysis aimed at understanding a given situation or predicting future behavior, such an action prevents one from gaining insights through these random outliers. It’s in these out of the norm behaviors and relevant transactions where one can gather unique and interesting insight, allowing for innovation to blossom.

We recommend companies immediately begin analyzing this data set in a separate effort from their traditional ones. Such data can yield valuable findings which can result in an expansion of offerings, development of alternative treatment methods, or even help see the future in terms of how customers may behave one day.

Why?

Traditionalists in the business intelligence realm have naturally examined mass consumer behavior, hoping to identify certain facts that could help their companies acquire more, sell more, or retain more. The natural tendency has been to ignore outliers, with the theory being that they hold little value in terms of trying to understand them, sell to them, or cater to them. But it’s really these unique outliers that can pave the way for the future – the unique way in which they interact with or use the products and services of a company can help that company identify how to push their boundaries in terms of innovation.

Take 3M for instance; renowned for its uniqueness, 3M is a company that has constantly strived to cater to those people we normally deem as on the edge – from astronauts to warzone doctors, Hollywood make-up artists to mountain climbers,

this innovative company is defined by products that originated thanks to these outlier consumer segments. Today, world-over, individuals across all segments use such 3M products as high-strength bonding tape or scouring pads, products inspired by those that can be deemed as on the fringe.

Customer analytics efforts that focus on the behavior of outliers can help companies identify how to excel in such a manner, resulting in service offerings catering to such outliers’ needs. Take Bharti Airtel, for example, a mobile operator from India that identified a segment of farmers with low voice traffic that were frequently using certain value added services such as weather forecast. In cooperation with farmers associations, the company launched an innovative service package for farmers. The package, which includes critical information on seeds, fertilizers, market prices, and weather (temperature, rainfall, wind conditions), also allows farmers to communicate with agriculture experts about crop problems. Farmers can take photos of particular crop problems and send these photos to agriculture experts by using MMS service. Agriculture experts with local market knowledge discuss and help to resolve the problems.

How?

We recommend a four-step approach for identifying outlier customers, understanding their nature, and making the most out of the knowledge gained.

1. Identify Outlier Customers: In order to benefit from the behavior patterns of outliers, they first need to be identified. There are three main sub-categories of outlier behaviors:

Exceptional Behavior: Extreme usages of a product or a service are considered in this category, behavior which is not expected from a traditional consumer. Examples of this include a customer who logs onto an internet banking account more than 30 times a day, or a small business that purchases hundreds of plane tickets in a week.

Unexpected behavior: Cross analysis of certain product or service usage with other descriptive dimensions may reveal such behaviors. If a customer who is not considered as a prospect for a certain product or a service (by looking at similar customers’ behaviors in a customer base) converts into a consumer, he/she can be considered as an unexpected behavior outlier. A very low value customer who signs onto a very expensive annually contracted product or service bundle can be an example of such a case in the telecommunications industry.

Untriggered Behavior: An unlikely behavior that emerges without any efforts of the company can be considered as an untriggered behavior outlier. A once upon a time subscriber who comes back years later to make a purchase can be considered as an example of this.

Companies need to examine their own situation and marketplace facts to hypothesize about which scenarios to examine relative to the above behavior categories. Each company has its own set of scenarios to consider and devise, as the varying market and sector factors dictate a company-specific examination be conducted.

2. Analyze Outlier Customer’s Comprehensive Behaviors: After outliers are identified under three sub-categories by looking from several perspectives, a customer analysis base needs be established. Analysis on this base will reveal possible correlated behaviors of the outliers in other dimensions and also will be used to understand how customers deviate from the regular customer base in the remaining behavior dimensions. Comparisons should be made for the periods before and after the outlier behaviors are observed.

3. Understand the Root Cause: Since outliers deviate significantly from the regular customer base

in terms of how they behave, information that is derived from the observations will not be enough to answer why they deviated. In order to understand the certain needs and motives that drive them, companies should consider conducting various methods of market research to identify the root causes of the behavior. Understanding why certain customers interact with the company’s channels, products, or services in a given way can only be truly understood through talking directly with said customers.

4. Create New Offerings: As a final step, new offerings would need to be developed to capitalize on the findings. Whether it’s the launching of a new pricing scheme, a new product, or a new channel of communication, benefits can only be obtained through creating a unique offering meant to satisfy the needs of the unique outliers. As with all new offerings, they should be tested through pilots with the outliers to ensure acceptance by the outlier sub- groups.

What Next?

Companies need to make the examination of outliers a common-part of their business intelligence practices. Business intelligence units need to work hand-in-hand with marketing, sales and product development units regarding the possible benefits of outlier analytics. Regular get-togethers and workshops should be planned to ensure continuity of such practices.

The effects of innovation derived from the insights from the outlier analytics should be measured and monitored as well in order to create momentum around the practice. Once the benefits are realized of such efforts, the more likely it will be that such types of analytics become a part of day-to-day business. 


About Utku Sarioz

Partner at Forte Consultancy Group
Share this post:

Leave a comment

You must register to leave a comment