Learn about machine learning and its potential impact on customer loyalty.
Machine learning has been a hot topic in the loyalty industry in recent years, but now seems to be building more momentum amid advances in Big Data technologies. Expectations Are High for Machine Learning and Its Potential Impact on Customer Loyalty.
Loyalty360 talked to Manoj Das, VP of Products, Stellar Loyalty, to find out more about machine learning and its potential impact on customer loyalty.
Expectations Are High for Machine Learning and Its Potential Impact on Customer Loyalty
JIM TIERNEY, LOYALTY360, DECEMBER 20, 2017
For those still trying to wrap their arms around this topic, can you explain what Machine Learning is and why expectations are high for its potential impact on customer loyalty / CX?
Das: At its core, Machine Learning means the ability for computers to learn without being explicitly programmed. In practice, as applied to Data Science, Machine Learning means building models using statistical and related techniques based on data for which a conclusion is already available to make predictions on new data sets. For example, if you have large data set available on customer attributes and historical book purchases, you can use it to predict which book they are likely to buy next.
Machine Learning is a broad category of algorithms that include Random Forest, Support Vector Machines, and Neural Networks. Deep Learning is a commonly used term for the Neural Network family of algorithms. In popular discourse, Machine Learning is also referred to as Predictive Analytics.
While Machine Learning techniques have been around for some time, they have recently become significantly more useful because of the emergence of Big Data technologies that make storing and working with extremely large data sets viable as well as the availability of processing power well-suited for this purpose. Combined with these trends is a related trend that these algorithms have become a lot more accessible with availability of packages such as Tensor Flow and H2O.
Machine Learning will have impact across functions in an organization. The most significant impact may be in customer-facing functions, particularly loyalty marketing and customer support. It can lead to:
Better targeted campaigns,
Improved lead generation and qualification,
Relevant up-sell and cross-sell recommendations,
Reduced customer churn,
Enhanced customer loyalty and profitability, and
Proactive fraud and anomaly detection.
What advice do you have for a marketer/brand in terms of where and how to get started?
Das: As with anything, I recommend marketers adopt a crawl, walk, run strategy for embracing Machine Learning.
Crawl: The primary goal of the crawl phase is to understand the data you have, where it comes from, what are the gaps, and what it is useful for. Marketers can use this phase to get some basic insight about the data and prepare for the next phases.
Walk: With the data science effort on firmer ground, marketers can focus on the analysis, which in this phase is more exploratory and deeper in nature. The goal of this analysis is to answer business questions as well as to hunt for further avenues to explore. In addition to analysis, this phase should start exploring simple, easily-accessible Machine Learning techniques such as forecasting, causal impact analysis, and anomaly detection.
Run: As the data science team matures, Machine Learning can be expanded beyond forecasting, impact analysis, and data analysis and into initiatives spanning Propensity Scoring, Recommendations, and Segmentation. With widely available packages and frameworks, even modest data science teams can expect to leverage the power of Machine Learning. Actual returns may vary, but combined with the results from the previous phases, the team can justify the investment and recommend improvements.
What type of skills and investments are needed to get started?
Das: A cross-functional team is the best recipe for success. There are three types of people that need to be assembled into the data science team:
The Business Drivers: I recommend starting with business persons who not only know the right questions to ask, but are also inquisitive and intellectually agile, and can keep asking the next level of questions at each stage of analysis.
The Data Dorks: Next, you need the data dorks. These are the people who understand the data from a business perspective and can shape and reshape it, as needed, for the different analysis as well as prediction algorithms. They also can look at the analysis and have an intuition on what it is saying and whether it makes sense. The ideal people to evolve to this role are your business analysts.
The Mathematical Programmers: Finally, you do need some members in the team who can actually program the analysis and the models using technologies such as R and Python. It is ideal if you can hire some data science trained engineers. If you find it difficult to compete for that resource pool, pick some of your analytic engineers, or those with SQL type skills, and have them pick up the skills needed.
Data will always play a key role in assessing customer insights, etc. How are brands using Big Data well now, in relation to customer loyalty efforts, and where do the challenges lie?
Das: Leading brands have started using data to understand consumer behavior and predict what would motivate a desirable change in behavior. For example, instead of sending everyone the same offer, leading brands now leverage known customer propensities to do 1:1 targeting. Also, leading loyalty programs are moving away from the same points and rewards for every customer to more targeting opportunities.
The challenges in working with Big Data include:
Privacy and Security: Brands need to be extremely careful with protecting personally identifiable information as well as ensuring that data does not leak out of protected walls.
Framing or Structuring: Any data science effort depends on how the available data is framed, that is what is included vs. what is not, and how what is included is shaped. For example, should a customer’s yearly visit or should a time series of daily visits be used. In Data Science lingo, this is called Feature Engineering.
Regarding the loyalty industry, what do you foresee for Big Data and Machine Learning in the near future?
Das: I think Big Data and Machine Learning will make loyalty programs significantly more engaging and rewarding. From a consumer perspective, instead of being a discount program, the loyalty program will start becoming more of a personal concierge that will help you steer through the brand’s product and service offerings.