Conversation Analyzer - customer sentiment

The Conversation Analyzer - customer sentiment dashboard presents conversational insights about customer and agent sentiment during calls filtered by key elements such as agents, categories, subcategories, duration, and disposition codes. Information includes insights for inbound and outbound calls. This dashboard is designed to help a business answer questions such as “Are we giving customers a positive experience?” or “Who or what is causing positive or negative sentiment?”.

Most historical analytics are designed to show numbers and reports, while Conversation Analyzer analytics tells you what happened in the call and how the customer felt. Having as much context about the call is important; for example, initial call direction and disposition code.

In this dashboard you can select various data points such as sentiment ranges, agents, and subcategories, and the rest of the dashboard will refresh to reflect that you only want data about that specific context.

At the top of the dashboard, you can set global filters for this dashboard:

• Date — the period of time you want to analyze data for. You can choose different preset ranges or you can specify your own custom data range. By default, Date is set to Last 7 days.
• Profile — the profile (set of categories, subcategories, and rules) you want to analyze data for. You can choose different profiles to narrow down the interactions in the dashboard. By default, the Profile filter contains all profiles.
• Category — the category you want to analyze data for. You can choose different categories to narrow down what was discussed in the dashboard. By default, the Category filter contains all categories.
• Subcategory — the subcategories you want to analyze data for. You can choose different subcategories to narrow down what was discussed in the dashboard. By default, the Subcategory filter contains all subcategories.
• Disposition code — the outcomes of the interactions you want to analyze data for. Choose from different disposition codes available in your account. By default, the Disposition code filter contains all disposition codes.
• Initial direction — the direction of the interaction when it started. Initial direction can be either Inbound (initiated by an external party), Outbound (initiated by an agent to an external party), or Internal (initiated by an agent to another agent). By default, the Initial direction filter contains all directions.
• Agent — the agents (identified by name) you want to analyze data for. Choose from agents available in your account. By default, the Agent filter contains all agents.
• Group — the group (identified by name) you want to analyze data for. You can choose different rules to narrow down the agents (assigned to the selected group or groups) in the dashboard. By default, the Group filter contains all groups and displays data for all agents.
• Skill — the skill (identified by name) you want to analyze data for. You can choose different rules to narrow down the agents (assigned to the selected skill or skills) in the dashboard. By default, the Skill filter contains all skills and displays data for all agents.
• First queue — the first queue (identified by name) that you want to analyze data for. The first queue is where the call was initially routed to. By default, the First queue filter contains all queues.
• Subcategory matches — the number of times the subcategory was matched during the conversation. By default, the Subcategory matches filter is set to allow for greater than or equal to zero matches. There may have been a transcription that generated sentiment, but no subcategories were found.
• Interaction transferred — whether the interactions you want to analyze data for have been transferred. Interaction transferred can be Yes or No. By default, the Interaction transferred filter includes all interactions whether they have been transferred or not.
• Recording time — the total time of the call recording. This value is affected by the Duration unit. Recording time includes the time the interaction started to be recorded until it was ended. This may include IVR and queuing time depending on when recording is configured to start. By default, the Recording time filter allows for all time.
• Duration unit — duration unit you want to set on the Recording time that you want to analyze data for. You can choose duration units so you can set recording duration by milliseconds, seconds, minutes, hours or days. By default, the Duration unit filter is set to Minutes.
• Customer hold time (ms) — the average time an agent put the call on hold in milliseconds. By default, the Customer hold time (ms) filter allows for all time.
• Overall sentiment score — the overall sentiment scores you want to analyze data for. By default, the Overall sentiment score filter allows for all scores.

Average Sentiment Score tile

The Average sentiment score tile displays the average of all the calls' average sentiment scores. This gives an insight into the high-level sentiment that customers are expressing towards the business. This will help answer the question “How do customers feel about engaging with the business during their interactions?”.

The Average Sentiment Score tile contains the following data:

• Average sentiment score

Average sentiment score trends tile

The Average sentiment score trends tile displays how the average score across all of the call’s average sentiment scores has changed over time. This gives an insight into the high-level sentiment that customers are expressing changes daily. This will help answer the question “Has there been a change in how customers are feeling about the business?”.

The Average sentiment score trends tile contains the following data:

• Average sentiment score
• The date the interactions occurred

Calls by sentiment type tile

The Calls by sentiment type tile displays the number of calls that were, overall, either very positive, positive, neutral, negative, or very negative. This gives an insight into how many calls fall into each of these sentiment areas. For example, if the overall sentiment is neutral, it may be because all the positive sentiment is countered by all the negative sentiment rather than most calls being neutral. This will help answer the question “Which calls are creating a very negative sentiment from the customer?”.

The Calls by sentiment type tile contains the following data:

• Sentiment type

Calls by sentiment type trends tile

The Calls by sentiment type trends tile displays the number of calls by day that were, overall, either very positive, positive, neutral, negative, or very negative. This gives an insight into how many calls fall into each of these sentiment areas and has it changed. For example, if the overall sentiment is less positive it could be because there are fewer positive calls or more negative calls. This will help answer the question “I can see there is a change in the customer's sentiment, but do I have more unhappy customers or fewer happy customers?”.

The Calls by sentiment type trends tile contains the following data:

• Sentiment type
• The date the interaction occurred

Calls by sentiment type by agent tile

The Calls by sentiment type by agent tile displays the number of calls that were, overall, either very positive, positive, neutral, negative, or very negative. This gives an insight into which agent had calls that fall into each of these sentiment areas. For example, if the sentiment for an agent is neutral it may be because all the positive sentiment is countered by all the negative sentiment rather than most calls being neutral. This will help answer the question “Which agents are creating a very negative sentiment from the customer?”.

The Calls by sentiment type by agent tile contains the following data:

• Sentiment type
• Agent name

Sentiment distribution by agent tile

The Sentiment distribution by agent tile breaks down the interactions to show, on average, the percentage of call that is very positive, positive, neutral, negative, or very negative. This gives an insight into how consistent an agent is during a call to manage the customers sentiment. For example, how much of a call is either very positive or negative. This will help answer the question “Which agents are having interactions with customers that are mostly very negative?”. This is different from the Calls by sentiment type by agent tile as a call may have an average sentiment of positive but maybe 20% of the call was very negative, 10% was neutral, and 70% was very positive.

The Sentiment distribution by agent tile contains the following data:

• Sentiment type
• Agent name

Average sentiment score by agent trends tile

The Average sentiment score by agent trends tile displays the average sentiment of all calls by agent. For example, you can see which agents are consistently providing customers with a negative experience. This will help answer the question “Which agents are providing the best customer experience and should be rewarded, and which are giving the worst customer experience?”.

The Average sentiment score by agent trends tile contains the following data:

• Average sentiment score
• The date the interaction occurred

Average sentiment trends (by category) tile

The Average sentiment trends tile displays how the average sentiment score changes over time by category. This gives an insight into category mentions associated with sentiment scores for that interaction. This does not mean that a specific category has a specific sentiment. This will help answer the question “Which categories are mentioned in calls that generate a poor sentiment?”.

The Average sentiment trends tile contains the following data:

• Category names
• Average sentiment score
• The date the interaction occurred

Average sentiment trends (by sub category) tile

The Average sentiment trends (by sub category) tile displays how the average sentiment score changes over time by subcategory. This gives an insight into subcategory mentions associated with sentiment scores for that interaction. This does not mean that a specific subcategory has a specific sentiment. This will help answer the question “What is being talked about and generates a poor sentiment?” or “Does a particular competitor or product name cause negative sentiment?”.

The Average sentiment trends (by sub category) tile contains the following data:

• Subcategory names
• Average sentiment score
• The date the interaction occurred

Declining sentiment by agent trends tile

The Declining sentiment by agent trends tile displays the number of times a call started (first 30% of the call) poorly and declined towards the end (last 30% of the call). This gives an insight into which interactions could have been managed better as the customer felt worse about the experience as it progressed. For example, the customer called with the intent to renew, but realized there was a price increase. This will help answer the question “Which agents are creating unhappy customers?”.

The Declining sentiment by agent trends tile contains the following data:

• Category names
• Stacked subcategory names
• Number of times the subcategory was found

Improving sentiment by agent trends tile

The Improving sentiment by agent trends tile displays the number of times a call started (first 30% of the call) poorly and improved towards the end (last 30% of the call). This gives an insight into which interactions were managed very well as the customer felt better about the experience as it progressed e.g. the customer called with a complaint and threatened to churn and the agent recovered the customer and they renewed. This will help answer the question “Which agents are able to recover unhappy customers?”.

The Improving sentiment by agent trends tile contains the following data:

• Category names
• Stacked subcategory names
• Number of times the subcategory was found

Interaction details tile

The Interaction Detail tile displays basic details about the interaction, such as start time, initial direction, and subcategories found. The main purpose of this tile is to enable you to open the analyzed interactions and then review them in more detail.

The Interaction details tile contains the following columns:

• Interaction ID — the unique identifier for the interaction. Click the ellipses to see the Interaction details dashboard or the Conversation Analyzer player for that interaction.
• Start time — the date and time on which the interaction started.
• Name — the agent’s name.
• Overall sentiment — a verbal representation of the overall sentiment score.

• Overall sentiment score — an average of sentiment scores throughout the call.

  Sentiment	Sentiment score range	Color
  Very negative	–1 <= score < –0.6	Dark red
  Negative	–0.6 <= score < –0.2	Red
  Neutral	–0.2 <= score < 0.2	Yellow
  Positive	0.2 <= score < 0.6	Green
  Very positive	0.6 <= score < 1	Dark green

• Sentiment delta — the average sentiment from the first 30% of the call minus the average sentiment from the last 70% of the call. If the difference is less than zero then Sentiment delta is Declined, if the difference is zero then Sentiment delta is No Change, and if the difference is greater than zero then Sentiment delta is Improved.
• Handle time — the time an agent or agents have spent working with an interaction. In hh:mm:ss format.
• Talk time — the time the customer and agent spent in the connected channel state. The time excludes any time the customer was on hold.
• Recording time — the total of the recorded time of the interaction. This can be longer than the talk and handle time, but possibly less than the connected time.
• Customer hold time — the average time the agent put the call on hold.
• Very negative sentiment — percentage of the call that is considered very negative.
• Negative sentiment — percentage of the call that is considered negative.
• Neutral sentiment — percentage of the call that is considered neutral.
• Positive sentiment — percentage of the call that is considered positive.
• Very positive sentiment — percentage of the call that is considered very positive.
• Sentiment score for 0-10%, 11%-20%, etc of call — average sentiment for the interaction divided into ten 10% sections.