The stated reason for calling in the customer’s own words. Not in ‘internal language’.

Operative Intelligence powers insights by classifying your customer’s stated reason for making contact in the customer’s own words. This can be applied over a large volume of interactions using language algorithms to help organizations understand their customer needs and operations.

Interactions are generally classified into three layers of classification:

The underlying reason for making contact with the organization, not necessarily the underlying issue. Think of this as the antithesis of the interaction: if this didn’t happen, the customer would not have needed to contact you.

The main topic of the interaction, specific to the root cause.

(Note: If your organization is based in Australia, New Zealand or the UK, this will be spelled 'Enquiry')

Specific to the inquiry. The next layer of detail about the topic (if applicable).

(Note: If your organization is based in Australia, New Zealand or the UK, this will be spelled 'Sub Enquiry')

Example

A customer calls and says “Hi, I placed an order online about ten minutes ago, I didn’t receive any confirmation so I’m calling to find out if the order has gone through”.

This interaction could be classified as follows:

Based on the remainder of the call transcript we can also predict for satisfaction score and resolution of the interaction.

A score between 1 (low satisfaction) and 5 (high satisfaction).

Predicted Satisfaction score is developed by looking at the language used by customers in an interaction and scoring using trained data sets. By using language analysis we can predict a customer satisfaction score for all interactions, not just those customers who respond to surveys. Typically less than 10% of customers who are offered a survey will respond. By having the ability to analyze the language used by customers, this can provide insight at scale whilst controlling for survey offer and response biases.

A score of 1 (resolved) or 0 (not resolved) averaged to provide a percentage of interactions resolved for a specific classification.

Resolution can be predicted for voice, messaging and web chat interactions. To predict for resolution we look at the ending of the interaction, if the interaction ends with a customer stating they’ll need to take further action (for example: call back, message again later, supply a document or visit a store or branch) the interaction will be marked as not resolved. Otherwise the interaction will be considered resolved.

Note: Unresolved interactions do not necessarily indicate that an agent has not followed the correct process or has done something wrong. The issue blocking resolution could be down to a business process or regulatory requirement.

A lower score is better for this metric.

Agent Effectiveness Score predicts for interactions with efficient resolution and satisfaction. This score is calculated as:

[Handle time / (Predicted Satisfaction + Predicted Resolution)].

When extrapolated across a large sample of calls for a given call type, this can help identify differences in agent performance or which interaction types need focus for process improvement by looking at which interaction types have a large span of Agent Effectiveness scores.

Example:

Let’s compare two calls which both have the same handle time of 600 seconds (10 minutes).

Call 1

Handle time: 600 seconds

Predicted Satisfaction: 5 (high satisfaction)

Predicted Resolution: 1 (resolved)

Agent Effectiveness Score: [600 / (5 + 1)] = 100

Call 2

Handle time: 600 seconds

Predicted Satisfaction: 1 (low satisfaction)

Predicted Resolution: 0 (not resolved)

Agent Effectiveness Score: [600 / (1 + 0)] = 600

In this example the two calls each have the same handle time, however the predicted satisfaction and resolution mean Call 1 has a lower (better) Agent Effectiveness score meaning it has been resolved with higher predicted satisfaction than Call 2.

Using the customer side of the interaction, analysis is conducted on the text to identify if the overall sentiment of the interaction is positive, neutral or negative.

Scoring is from -1 to +1, the scores can be interpreted as follows:

Opportunity types are inquiries further classified into quadrants using the lenses of 'Is this inquiry valuable to the business?" and "Is this inquiry valuable to the customer?".

Once understood these can be leveraged to identify opportunities for improvement and measure success of any initiatives or projects that impact the inquiry or quadrant in question.

Common Opportunity Types Identified:

These interactions are valuable to customers, but not necessarily valuable for the organization to service in contact center channels. Organizations should aim to enable and migrate as many of these interactions as possible to self service or automated channels.

These interactions are valuable to the business or organization, but not valuable to customers to spend their time doing. Organizations should aim to make these types of interactions as easy as possible for customers.

These interactions are not valuable to the business or to the caller, generally organizations would benefit from eliminating as many of these interactions as possible.

These interactions are valuable to the business and are examples of where a contact center channel is providing value or generating revenue.