As big data takes precedence in curating a marketing strategy, it is important for enterprises to stay cautious while venturing into the journey of data and algorithm strategy development
We are living in an age where exceptionally high volumes of data — both structured and unstructured — are captured by technology. While large amounts of data by itself may not be valuable, data refined through proper algorithms can produce efficient results. Companies that are unable to keep pace with rapid changes in technology are at a huge risk of being taken over by those who have mastered leveraging data and algorithms to improve business outcomes. This means breaking away from the ‘traditional’ approach and moving towards machine learning. To illustrate this point, here are two examples to help differentiate between the ‘traditional’ approach and ‘sophisticated’ approach to big data analytics.
Intervening at the right time
Consider customer retention in the insurance industry. Traditionally, the analytics team within an insurance company would be tasked to build a policy lapse prediction model to help predict policies that are at a high risk of lapse. With this insight, the retention operations team can prioritise policies for intervention before they lapse. These models are typically linear regression models and are refreshed at periodic intervals. But, given the pace of change that we see across markets in terms of customer base, products and distribution mix, there is a strong need to make these algorithms robust, and self-learning using advanced machine learning techniques and artificial intelligence. Some innovative insurance companies such as AXA are actively replacing traditional linear models with algorithms that learn continuously, based on recent experiences to stay ahead of the curve.
Another example is fraud detection in a credit card portfolio. Traditionally, fraud is detected using a combination of algorithms (learning from historical experience) and a set of business rules defined by forensics experts in that domain. With big data processing capabilities and more advanced machine learning algorithms, we now have the ability to model what is ‘atypical’ behaviour at a particular customer level. This means, we have algorithms that model the ‘typical’ behaviour of each customer learnt through analysing historical transactions, compared against each new incoming transaction and flag the ones that are suspected as ‘abnormal’. American Express is a great example of a company harnessing the power of big data to continually solve various critical business problems including fraud.
The winning ways
While a few companies across industries have taken steps towards using data and algorithms to make rapid and relevant decisions, many companies are yet to follow. Analytics departments are increasingly elevated from being a support function to being an advisor to business strategy, marketing and risk areas to name a few. The question of how analytics, algorithms and artificial intelligence can be leveraged to win in the marketplace is now a question that has been asked by all management teams.
Consider an example of personalising product recommendations and offers to a food and grocery retail customer. A grocer would have thousands of products (aka SKUs) available at each of its stores, and a customer typically buys a few hundred unique SKUs across a set of categories and sub-categories. With advanced data processing capabilities, grocers can now model a customer’s propensity to buy a set of limited SKUs within a certain period and send this curated list along with offers to the customer to incent their visit to the store for purchase. A relevant set of personalised product offers would elicit a better response from the customer than a generic list. Taking this example further, let us imagine the possibilities of an FMCG brand leveraging this ‘in-market’ customer list for promotions of its brands. The UK-based food and grocery retailer Tesco had done phenomenal work to understand its shoppers, personalise its products and offer recommendations for years.
Here is another example of how data can be used to improve the customer experience. Telcos are installing passive probes across their networks to gather more granular data to understand the actual experience of their customers. This may cover many things including the quality of service and customer experience associated with call, data and video-streaming activities and specific actions to improve these. Now, let us imagine performing these actions in near real-time while customers are facing a quality issue. Middle East-based telcos such as Etisalat and Mobily have taken enormous steps in improving customer experience through collecting, processing and actioning on the granular customer-level behaviour data.
Data must be looked at as an asset. The strategy towards achieving this should include data that is captured in-house, through systems such as CRM, loyalty, web and apps, as well as external data.
With this rapid change, enterprises must also take caution while they venture into the journey of data and algorithm strategy development. Designing these analytics applications must be done keeping the privacy of the customer at the core. This includes taking the consent on the degree to which the data is captured and its purpose, in order to stay in-line or ahead of what regulation mandates. The potential of the algorithm economy is huge and the journey is
The author is VP — strategy and insights, Epsilon