For many companies, using predictive models to uncover and forecast customer intent is critical to managing and increasing customer satisfaction. Businesses have begun applying predictive analytics to unstructured, text-based data sources, such as customer calls and web chats, in order to gain a better understanding of the customer and to catch major issues before they become critical.
The challenge? Predictive models have traditionally been created with only structured, quantitative data like scores and ratings. It can be immensely difficult to augment these models with the unstructured, qualitative data sources (like open-ended survey responses, contact center tickets, and social media) where the most valuable insights and indications of customer intent live.
One of Luminoso’s clients, a large government-run public health organization, faced this exact problem. While they already had sophisticated models to estimate the spread of disease, the data they used was limited to quantitative data reported after-the-fact by doctors’ offices, hospital emergency rooms, and urgent care centers. They knew that in order to be truly predictive and thus more effective, they would need to incorporate real-time data from unstructured sources like Twitter.
The public health organization partnered with Luminoso to use text-based data to improve their predictive analytics models. Using Luminoso Compass, the organization optimized their models and made them more responsive to changes like sudden flu outbreaks.
In addition, they were able to quickly detect and track key conversations happening about the Ebola outbreak – notably misinformation and conspiracy theories – to more rapidly address public concerns and provide accurate information to the American populace.