Nimble is raking in mountains of customer, agent and collection data and uncovering patterns in this data using artificial intelligence and prediction analytics. This boosts the business decision-making process by providing valuable insights and identifying opportunities for the future.
Predictive analytics has taken the data world by storm. Today many people are comfortable with the idea of a predictive analytics engine matching them with a partner.
As agents are the heart of a collections business, Nimble is looking to significantly redefine their recruitment practices by using analytics as a key driver in the overall hiring process.
The Agent Early Performance Predictor model identifies what specific data is predictive of an agent’s performance by spotting patterns and distilling themes. When the model is applied at 30 days, 60 days and 90 days after employment, it is able to accurately predict the agent’s performance at month 6. This enables Nimble to identify rising stars and drive agent performance improvements with early intervention programmes. By highlighting specific performance components, the model helps identify why top performers exceed their peers so that coaching can be built around these successful behaviours.
Nimble uses speech recognition to convert the unstructured data, hidden in the relatively unexplored audio files, into structured data, so that the calls can be analysed. The output of the speech recognition model i.e. decoded text, is analysed in order to identify the defining features of calls with voice mail messages as well as those answered by humans e.g. a long string of words such as “The person you have called is not available. Please leave a message…” as opposed to a live caller saying something similar to “Hello” followed by a post-greeting silence. The 40% to 60% of outbound calls identified as voicemail calls are then screened out allowing the agent to primarily interact with live calls and maximise talk time.
What about more abstract concepts like a commitment or promise?
Once the agent has made contact with the customer, a successful collections call concludes with a commitment to pay an agreed amount of money on a set date (PTP). There is however no guarantee that the PTP will lead to a payment. The Promise to Pay Strength predictive model is another one of Nimble’s bespoke algorithms that analyses collections data to determine the features that are predictive of the strength of a PTP. It then uses the absence or presence of these features to score a commitment in terms of the likelihood of being honoured.
The features contributing to the score are as important as the score itself. Features are classified as those that are under the agents control and those that are not. Actively using these insights enables Nimble to increase the PTP honoured ratio by delivering personalised agent coaching and by performing effective and appropriate customer segmentation.
These examples highlight some of the exciting ways in which Nimble is leveraging its structured and unstructured data.
Written by Ingrid de Leeuw – Executive