Debts and emotional intelligence
Debt collection is usually an emotionally charged situation, both on the part of the debtor and on the part of the collector. Debt is an intimate and delicate matter. Debtors are often afraid and ashamed of owing money, and talking to a debt collector can be very uncomfortable for them. Their emotions need to be taken into account when setting up any debt collection process.
The fundamental challenge of introducing our voicebot technology into Kooperativa’s processes was to test how people would respond to a voicebot when it comes to collecting debts related to insurance premiums.
Kooperativa defines two groups of customers:
those contacted before the termination of the contract with a follow-up reminder
those whose contract is no longer valid, but a debt is registered, with a follow-up early debt collection call
The aim of the calls, whether reminder calls or early collection calls, is to save people money so they are not turned over to collection agencies for recovery.
GDPR-compliant voicebots
From a conversation design perspective, a large number of different types of objections had to be managed in order to minimize the risk of misunderstanding. It was also important that the voicebot spoke to the debtor in an appropriately polite manner and that the conversation ended with a mutual agreement. In addition, all processes had to comply with the strict data security and privacy requirements that insurance companies must follow.
The fundamental challenge of introducing our voicebot technology into Kooperativa’s processes was to test how people would respond to a voicebot when it comes to collecting debts related to insurance premiums.
“The results of the pilot project were very positive in terms of the high answer rate and high debt collection success rate. We were pleasantly surprised by the cheerful and proactive communication style of Eva, the virtual assistant.”
Phase I – Pilot
For a meaningful comparison of voicebot and live agent results, contracts with the same parameters were analyzed.
The monitoring period included in the comparison was November–December 2019 for live agents and November–December 2020 for Eva. To compare the debt collection success rate between live agents and Eva, the number of contracts collected was used.
The phenomenon of abandoned calls
In the pilot phase, there were calls where the customer hung up right at the beginning without confirming their identity to Eva. This type of hanging up at the very beginning did not occur during the entire pilot phase with human agents. Based on this finding, the call script was modified to emphasize the introductory part of the call. This change resulted in an improvement: the percentage of hung-up calls decreased from 14.7% to 12% (of all calls). In the overall results, we differentiate between two main categories: "answered" (including hung-up calls) and "contacted" (excluding hung-up calls).
Phase II – Live operation
Based on the successful results of the pilot phase, Eva was put into production and began calling even more customers. Eva's main achievements include:
her overall results were comparable to a live agent
the results from 'contacted' completed calls were even better than the live agents' calls
customers accepted communication with Eva
there is great potential for Eva to improve even further in answer rate and debt collection success
“On the part of VOCALLS, we appreciated their willingness to make changes and the speed of the implementation. Last but not least, we enjoyed working with both Designer, their easy-to-use and user-friendly tool for building conversational experiences, and their call management application.”
Eva's main tasks in the live operation phase reflected lessons learned from the pilot phase: reduce hang ups, fine tune the communication style, and improve the reporting and call management systems.
Further adjustments to the call scripts have reduced the number of customer hang-ups, and even resulted in higher response rates and high collection success rates for more demanding and diverse contract types. Our voicebot flows have shown that it is more effective to continually communicate with the customer (e.g. ask them yes/no questions) rather than collect information from them with open-ended questions. At the same time, customers frequently voiced objections and protests during the call. For this reason, endless fine tuning of the flow and each of its nodes is not effective. It is better to focus on the "happy path" by maximizing the optimal route through the flow, and when analyzing each objection, consider whether it is worth eliminating.
Phase I – Pilot
November 2020 – December 2020
21.2x
60.6%
1 036
47s
2.83Kč
Phase II – Live operation
July 2021 – December 2021 (live operation still ongoing)
