Understanding how to retain a customer is a key challenge for many organisations. But what if you can retain a customer, while reducing the cost of handling issues and delivering better outcomes? The business case for retaining customers who report a problem is. To predict outcomes, you need some key ingredients…
- Classification of the issue(s)
- Customer emotion
- Outcome offered
- Key indicators
- Satisfaction with outcome
Capturing predictive insight
To predict an outcome, the first stage is to understand the issue or issues by type. This can be done by automatic classification, classification by the consumer or classification by the organisation employee.
Classification by the employee gas been shown to be subject to maximum level of variability, as here is no incentive to ensure accuracy of the data.
As an alternative, a consumer chooses their service, issue and issue type. For example:
This self-classification approach means around 80% of consumers classify their issue correctly, and the “other” category is sued to fine tune the issue options provided. This approach is simple a way to identify key issues, however, investigation has shown that complaints are often multiple smaller issues. The more complex the complaint, often the more issues are involved and the harder it is to resolve.
A simple classification approach no longer delivers sufficient granularity. Therefore, it becomes important to capture key factors in an issue so the nuance of an issue raised is understood. To deliver this it becomes essential to understand primary and secondary issues. For this approach a considerable additional amount of data is required.
To classify a case there needs to be a corpus of key words. These words have been classified into the three groups.
- Complaint corpus: these are the key words used by consumers when complaining. These words are common across all service sectors.
- Service corpus: these are the key words used by consumers when raising an issue. The words are specific to the service type for the issue being raised.
- Danger corpus: these are words used by consumers that should be given a priority, as the severity/impact associated with these words are high.
Establishing what are key complaint words does not help with priortisation to help determine what are the key issues and the key concern. To determine these an assessment of the emotion expressed in the communication is undertaken. This is an assessment of the case’s emotion around the key words, principally the sentences in which they are mentioned. Once the key words are priortised these are then associated with the communications with a primary and then secondary words. The primary word been the main issue and secondary words are supporting issues that the consumer has experienced.
The next stage is understanding what has been offered to the consumer. This requires automatic reading of the communication from the organisation to the consumer to understand the offer being made. Once a case has been agreed as closed by the customer the responses are analysed and the offer extracted. This requires all the unnecessary information to be removed including introductions, sign off and supporting information.
Key indicators are factors relating to the consumer raising the case, the case type, the customer’s emotion at time of opening and at each touch point, and the number of touch points. These provide supporting data on the case that are used as a third assessment factors, after the key and secondary indicators, where there is sufficient data.
Satisfaction with outcome
The final element to take capture the consumers satisfaction with the outcome. Resolver captures this from consumers at each touch point and at the case close. This data is captured and cross referenced with the offer. This then provides a full set of data capable of predicting outcomes and in understanding the satisfaction with the complaint journey process.
Make a Prediction
All this information is captured and recorded. Then the final element is to join all key data points to allow prediction of outcomes. The size of the data set then becomes a key factor. The larger the data set the better the prediction capability. Prediction consists of the best response to the customer and what is the likely best outcome to be offered.
The development of prediction is the fundamental leap that will allow business to deliver personalised customer service automatically. This means a complaint; the unsolved loyalty touch point can become predicted. This will deliver better resolution for consumers, improve the customer experience and importantly, reduce the cost of case handling and customer retention.