Customer Analytics: How to Derive Valuable Insights from Big Data
In my last blog, I talked about the importance of understanding customers at a deeper level, and how this understanding can translate into stronger loyalty and increased demand for products. In this blog, I start to talk about the "how."
Customer analytics, and really all analytics for that matter, start with data. Most companies will start by analyzing their structured transactional data, which typically includes information such as demographics, purchase history, acquisition, and retention information. Statistical algorithms such as linear regression, clustering and RFM (recency, frequency, monetary value) can help companies create meaningful segments, and get insights on buying patterns. These insights and propensities can be encapsulated in models for future predictions.
In their quest to make these models more accurate, companies are starting to integrate other sources of data with their demographic and transactional data. These sources include call center records, email communications and as well as usage pattern on company websites. Few companies mine this "gold mine" of information. Most of this data is unstructured, and requires text analytics and NLP (Natural Language Processing) algorithms to extract sentiments and concepts for further analysis. In addition to text analytics, social networking algorithms can quantify the relationships and influence, which is also important. One telco company reduced its customer churn from 19% to 2% by integrating call center records with its transactional data.
The other source of data is attitudinal data, which comes from market research studies as well as mining social conversations. Such data provides deeper insights into why people are doing what they're doing.
Models based on the statistical analysis of this integrated data set can help companies attract more profitable customers, increase their wallet share, reduce customer churn, reduce fraud and mitigate risks.
We are at a very interesting time in history where the velocity of change continues to accelerate. A lot of this is driven by trends such as mobility, social media and e-Commerce. Analytical algorithms continue to evolve to deal with the changing landscape.
Here are some examples of such ongoing innovation at IBM:
Algorithms are being developed to predict which tweets are likely to go viral. Propensity models are becoming more dynamic to deal the geo-spatial and temporal dimensions, acknowledging the fact that location and time events impact people's propensities. Temporal-causal algorithms are being developed to associate the revenue lift that can be attributed to sentiment change in social conversations. Finally, algorithms are making systems more cognitive so they can learn and get smarter over time, as we saw with the Watson computer beating the Jeopardy! champions.
As companies deploy customer analytics, they must keep in mind that in this new world, trust matters more than ever before. The need to be transparent and authentic has never been greater. There is a fine line between using customer analytics to create value by serving customers with optimized precision or to destroy value by surprising customers with actions that erode trust. And this fine line is hard to see because it exists in customers minds. Privacy policies and consistent execution across the enterprise are essential. Done properly, Customer Analytics can not only help drive near term value, but also longer term value as it can provide the mechanism to understand where the fine line is.