Before you begin

To use Action Intelligence, you must:

• Medallia Alchemy Experience Reporting must be enabled for reporting. For more information, see Medallia Alchemy Experience reporting.

• Your company must be using a supported language for surveys, as listed in Text Analytics supported languages.

• While Action Intelligence can provide value with a minimum of 10,000 records, a minimum of 50,000 records is recommended.

Attention

Action Intelligence analyzes every record and for each makes an assessment on whether the respondent is in need of immediate attention, either for negative reasons (such as an urgent unresolved issue) or positive reasons (such as a possible promoter conversion or a revenue opportunity).

Attention score and flag

Attention score provides a more precise way to identify respondents in need of attention than by using numeric feedback scores alone. When calculating whether a respondent needs attention, Action Intelligence considers the Main Score Field indicated on the Company settings page (typically Likelihood to recommend or Overall satisfaction), and one or more comment fields you select on the Reporting > Text Processing > Attention screen in Medallia Setup. The field you select as the main score field must be a numeric field with one of these ranges: 0-5, 1-5, 0-10, or 1-10.

After processing, Action Intelligence records a score in Attention Score (a_attention_score) to provide a ranked order of importance for records in need of attention. Action Intelligence also populates Attention Flag (a_attention_flag) with a Yes or No value.

Attention types

In addition to identifying respondents that need any type of attention, Action Intelligence identifies the specific type of attention needed. Depending on the phrases in the feedback, a record can be flagged for multiple attention types:

• Churn intent — The respondent indicated intent to reduce business with your company, or stop doing business completely. This attention type is flagged by the Attention label churn intent flag (a_attention_label_churn_intent_flag). Example phrases include:

  • “Guess I’ll go back to your competitor. I will not be back.”
  • “I am switching service providers.”
• Eroding faith — The respondent has not explicitly stated they will reduce or stop business with your company, but has indicated that their trust or satisfaction with your company has lessened. The respondent no longer recommends the company and might actively recommend others away from your company.  The respondent might state that they are actively considering alternatives. This attention type is flagged by the Attention label eroding faith flag (a_attention_label_eroding_faith_flag). Example phrases include:

  • “So that's why I'm not going to recommend your company.”.
  • “The service has really gone downhill.”.
  • “We have been your customers for many years and have never had such a bad experience!”.
• Legal action — The respondent indicated an intent to sue your company or contact government regulatory agencies. This attention type is flagged by the Attention label legal action flag (a_attention_label_legal_action_flag). Example phrases include:

  • “I’m going to sue you for false advertising.”.
  • “The FTC needs to know about this.”.
  • “If you don’t give me my money back you can expect to hear from my lawyer.”.
• Opportunity — The respondent might be a promoter of your company who has expressed an interest to do additional business. The respondent might be passive or a detractor, but has offered an opportunity to improve the company’s reputation with him or her, resulting in additional business opportunities. This attention type is flagged by the Attention label positive opportunity flag (a_attention_label_positive_opportunity_flag). Example phrases include:

  • “I want to learn more about your membership program!”.
  • “I’ll gladly recommend you to my friends, if you just fix my issue.”.
  • “What other smart devices do you have?”.
• Unresolved issues — The respondent asked for help or described problems that require follow-up from your company. This attention type is flagged by the Attention label unresolved issues flag (a_attention_label_unresolved_issues_flag). Example phrases include:

  • “Please call me back as soon as possible.”.
  • “It’s been over a week and the issue is still not resolved.”.
  • “I am still getting charged for services I didn’t ask for.”.

Use Attention fields

In Responses Form and Responses Feed, use the Medallia AI Tags to Display property to show whether a response needs any type of attention or specific types. If you select Needs Attention, a record displays an Attention Needed Flag for records with an Attention Flag of Yes. If you select Attention Types, a record displays a different flag for each attention type when its Attention Label field has a value of Yes. For example, this image shows the heading of a response in the Responses Feed module, with a response that is flagged for attention because of possible legal action from the respondent. In this example, both the Attention Flag and Attention Label Legal Action Flag fields have a value of Yes:

In Responses List reports, consider adding as columns the Attention Flag field to indicate when a response needs any type of attention, the Attention Types field for a concatenated list of all of the attention types flagged for a record, and one or more Attention Label fields if you want to list each type of attention separately.

Consider using Attention fields as part of a field-based conditions in alerts to notify the right people in your organization as soon as Action Intelligence identifies a respondent in need of attention. For more information, see Alerts.

Suggested actions

Action Intelligence parses comments in feedback to discover suggested actions based on those comments, and then scores those suggestions based on actionability. This feature helps you understand issues, and prioritize business decisions and actions that have the highest impact on respondents happiness and loyalty.

In reports, users can sort feedback by date or actionability. For example, this image shows a Responses Feed module with a response expanded to show the full comment. The module is filtered to highlight any suggested actions discovered by Action Intelligence.

Actionability score

To determine whether a response is actionable, Action Intelligence begins by using phrase classifier that predicts whether each phrase contains a suggestion. Some phrases appear to be suggestions, but are too generic to be actionable, too specific to a particular situation to be generally applicable, or too obvious to require attention by actual users. The phrase-level suggestion score ranks sentences by likelihood that a person will judge the suggestion useful enough to take action or to share with someone else. This table shows examples of phrases that would receive low, moderate, and high suggestion scores:

For calculation purposes, Action Intelligence assigns each phrase in the response a score between 0 and 1, with scores closer to 1 being more actionable. Action Intelligence discards phrase scores below the system threshold of 0.5, since those phrases are not good suggestions, and should not factor in determining the actionability of the entire record. Action Intelligence then calculates the record's actionability score by dividing the number of phrases with suggestion scores above the threshold by the total number of phrases in the record.

After Action Intelligence processes comments for suggested actions, it populates Actionability Flag field (a_actionability_flag) with a value of either Yes or No to indicate whether the record contains a suggested action. Action Intelligence also populates Actionability Score (a_actionability) with a numeric value between 0-100 to help you sort suggested actions by actionability in reports. For example, the module in the image above is sorting responses by actionability.

Show actionability in Responses Feed

You can configure a Responses Feed module to show only responses with suggested actions. With the Responses Feed module opened in Medallia Admin Suite, turn on the Enable Suggested Actions Highlighting property, and create a Data Filter for records with an Actionability Flag value of Yes (a_actionability_flag = 1). This image shows that filter in the Filter editor:

Effort

Effort measures the amount of work a respondent exerted to achieve his or her goals while interacting with your company. High effort is a leading indicator of low NPS, and can potentially lead to churn.

Action Intelligence identifies high-effort phrases and low-effort phrases across all verbatim comments.  Based on the relative density of high-effort and low-effort phrases, Action Intelligence computes an internal effort score (ranging from -10 to 10). If there are many more high-effort phrases than low-effort phrases, that score is low.The score is then used to map feedback to one of these categories: Very Hard, Hard, Neutral, Easy, Very Easy.

High-effort phrases are typically neutral or negative experiences, and usually describe instances where the respondent had to try something multiple times before succeeding, waited a long time, or experienced frustration from difficulty in communication. For example:

• "We had to call 3 to 4 times to get a policy issued.".
• "Sitting on hold for 20-40 minutes is too long.".
• "I had to reschedule 2 times for a total of 3 visits, which wasted a good chunk of my week and was a large inconvenience.".

Low-effort phrases are typically positive experiences that describe easy transactions, a pleasant respondent service experience, or an intuitive workflow (such as a website transaction). For example:

• "Quick & easy process.".
• "I found what I needed on the website.".
• "Mark really knew his stuff and answered all of my questions immediately.".
• "No problems with the returns.".

Effort phrases are highlighted in reports if you have selected Customer effort for the Default Highlighting property in Responses Form reports, or if you have turned on the Enable Customer Effort Highlighting property in Responses Feed and Comment Stream modules. Phrases are highlighted to indicate hard and easy effort. Neutral phrases are not highlighted. For example:

Based on the relative density of hard-effort and easy-effort phrases in a record, Action Intelligence generates an overall effort score for that record, with a value ranging from -10 (for Very Hard effort) to 10 (for Very Easy effort), and stores that score in the the Customer Effort Score (a_customer_effort_score). Action Intelligence compares that score to the Category Thresholds defined on the Reporting > Text Processing > Customer Effort screen to determine the effort category for that record, which is then populated in the Customer Effort Bucket Name (a_customer_effort_bucket). Effort labels include Very Hard, Hard, Neutral, Easy, and Very Easy. Reports use this value to display a flag for the record, as shown in this image:

If Action Intelligence identifies effort at the record-level, it also changes the value of the appropriate effort flag field to Yes. For example, if Hard effort is identified, the High Effort Flag (a_customer_effort_high_flag) changes from No to Yes. Use these fields as part of alert conditions. For more information, see Alerts.

Recognition

Recognition is a mention of an individual or team in the company, acknowledging their contribution (positive, negative, or neutral) to the experience being described. Individuals might be mentioned by name (such as "Randy") or by role (such as "the tech guy" or "the front desk manager"). Examples of team mentions include "The entire wait staff" or "The Medallia family".

Managers might filter feedback to display only records containing recognition, then review them to identify areas of strength and improvement, and to find phrases appropriate for official performance reviews or promotions.

Recognition phrases are highlighted in reports if you have selected Recognition for the Phrase Level Highlighting property in Responses Form reports, or if you have turned on the Enable Recognition Highlighting property in Responses Feed and Comment Stream modules.

If a record contains recognition phrases, Action Intelligence changes Recognition Flag (a_recognition_flag) to a value of Yes, which makes the Recognition flag available to be displayed for that record in Responses reports, Responses Feed modules, and Comment Stream modules.

Depending on the sentiment of the recognition, one of these fields is also set to Yes: Positive Recognition Flag (a_positive_recognition_flag), Negative Recognition Flag (a_negative_recognition_flag), or Neutral Recognition Flag (a_neutral_recognition_flag).

Override Action Intelligence processing

While most Action Intelligence processing is accurate, some feedback records or specific phrases were flagged or highlighted incorrectly, or are missing specific flags or highlights. Administrators with permission to do so can flag or un-flag records, and highlight or un-highlight specific phrases in Responses Form reports and Responses Feed modules.

To change a record-level flag, click Edit at the top of the report, select or unselect the flag, and then click Apply. For example:

To see highlighting for specific phrases with suggested actions, scroll to the verbatim feedback of the record and then select Suggestions for the Sentence highlight. For example, this image shows comments with three phrases highlighted with suggested actions:

To open that feedback for editing, click Edit and then set the Edit Mode to Suggestions. To add or remove suggested action highlighting for a phrase while in edit mode, click Suggestion at the beginning of the phrase, select the correct highlighting option, and then click Update. For example, in this image the first phrase highlighted as a suggested action should not have been highlighted. By setting the highlighting option to None, the highlighting is removed.

If overrides are not displayed immediately, refresh your browser.

Alerts generated based on Action Intelligence flags are not changed by overrides. For example, if an alert was generated because a record was flagged as needing attention, and you later override that flag for the record, the alert is not resolved automatically. Likewise, using overrides to add flags does not cause records to be reprocessed for alerts. Records are processed for alerts automatically only when they first enter Experience Cloud. To reprocess records for alerts, you must manually set the Survey status (e_status) field to 0.

For more information, see Action Intelligence alerts.

Action Intelligence filters

Action Intelligence provides these filters, which you can add to your reports to help users find specific responses processed by Action Intelligence:

• Attention Flag — Shows responses that were flagged as needing attention (Yes), not needing attention (No), or all responses (All).

• Attention Type — Shows responses needing one or more specific types of attention.

• Customer Effort — Shows responses with one or more levels of effort.

• Recognition — Shows responses that were flagged for recognition (Yes), not flagged for recognition (No), or all responses (All).

• Suggestion — Shows responses that were flagged for including a suggestion (Yes), not flagged for a suggestion (No), or all responses (All)

These filters appear in the control panel, as shown in this image:

Action Intelligence alerts

Create Alerts to notify users when Action Intelligence processing shows that follow-up is needed. For example, you might want managers to be alerted when a respondent calls attention to their work. You might want the Legal team to be alerted when respondents indicate they will take legal action.

Action Intelligence fields are available in alert conditions. For the type of field, select the Action Intelligence field, and then select the appropriate field and value for your alert condition. For example, the following image shows an Action Intelligence Insight condition component:

Configure Action Intelligence

Complete these steps to enable and configure Action Intelligence:

1. On the Reporting > Text Processing > Global Text Analytics Settings screen, complete these steps and then click Save:

   1. Select the Marking Surveys Enabled property.

   2. Select the Automatic Processing Enabled property.

   3. For each listed language, select one of these values to configure how Action Intelligence processes that language, and then click Save:

      • Do not process — Action Intelligence does not process the language.

      • Native — Action Intelligence processes feedback in its original language, before any translation. After selecting this option and clicking Save, the Use Medallia AI Sentiment Model option becomes available (for languages that have sentiment processing available). If your company uses Medallia Text Analytics, turn on that property to enable the sentiment model for that language as part of processing. For more information, see Configuring sentiment.

      • Translation — Action Intelligence processes feedback after it has been translated.

2. If you are using Action Intelligence Attention scoring, complete these steps on the Reporting > Text Processing > Attention screen, and then click Save:

   1. Select the Enabled property.

   2. In the Comment fields property, select the comment fields to process.

   3. In the Units property, select the units to process.

   4. Optionally, use the Condition property to restrict the scope of records considered for processing. For example, you might want to process records only for specific feedback programs. For more information, see Conditional expressions.

3. If If you are using Action Intelligence suggested actions functionality, complete these steps on the Reporting > Text Processing > Suggested Actions screen, and then click Save:

   1. Select the Enabled property.

   2. In the Comment fields property, select the comment fields to process.

   3. In the Units property, select the units to process.

   4. Optionally, adjust the value of the Suggested Action threshold property to change the sensitivity for determining which phrases are marked as suggestions.

      Actionability scores above the value are marked as suggestions.

   5. Optionally, use the Condition property to restrict the scope of records considered for processing. For example, you might want to process records only for specific feedback programs. For more information, see Conditional expressions.

4. If If you are using Action Intelligence effort scoring functionality, complete these steps on the Reporting > Text Processing > Customer Effort screen, and then click Save:

   1. Select the Enabled property.

   2. In the Comment fields property, select the comment fields to process.

   3. In the Units property, select the units to process.

   4. If needed, use the Set thresholds property to change the effort scoring thresholds. Enter a comma-separated list of numbers between 10 and -10, starting with the highest threshold. For example, a Set thresholds value of 6,2,-2,-6,-10 sets these thresholds:

      • Very Easy, score >= 6
      • Easy, 2 <= score < 6
      • Neutral, -2 <= score < 2
      • Hard, -6 <= score < -2
      • Very Hard, -10 <= score < 6

   5. Optionally, use the Condition property to restrict the scope of records considered for processing. For example, you might want to process records only for specific feedback programs. For more information, see Conditional expressions.

5. If If you are using Action Intelligence recognition scoring functionality, complete these steps on the Reporting > Text Processing > Recognitions screen, and then click Save:

   1. Select the Enabled property.

   2. In the Comment fields property, select the comment fields to process.

   3. In the Units property, select the units to process.

   4. Optionally, use the Condition property to restrict the scope of records considered for processing. For example, you might want to process records only for specific feedback programs. For more information, see Conditional expressions.

6. On the Install Apps screen, verify that the Action Intelligence (Medallia AI): Generic tile appears in the Installed section. This app contains Action Intelligence reports and their supporting configuration entities. For more information, see Action Intelligence Generic app.

   • If the app is available, but not installed (the tile appears in the Available Apps section of the Install Apps screen), install the app as described in Installed apps screen.

   • If the tile does not appear on the Install Apps screen at all, contact Medallia Support.

7. Configure the reports in the Action Intelligence Generic app per the needs of your company.

8. Optionally, create or update other reports to include information processed by Action Intelligence:

   • Update see Responses Feed, Comment Stream, and Responses Form to include needed Action Intelligence flags and comment highlighting.

   • For Attention, use one or more of the Attention Score, Attention Flag, Attention Types, and Attention Label fields in Responses List reports to call out records in need of attention.

   • For Suggested actions, use the Actionability Flag and Actionability Score fields.

   • For Recognition, add Recognition Flag fields to the Responses List report to call out records that include recognition phrases in comments.

   • Add Action Intelligence filters to reports.

9. Process historical records, as described in Process historical records, below.

10. Perform an analysis of Action Intelligence data processing, as described in Data assessment.

11. Create new Roles and assign users to them, as needed. If you create new roles, add those roles to the reports needed by those roles.

12. Assign one or more of the Text Analytics and Action Intelligence reporting permissions to roles, as needed.

13. Optionally, create one or more Action Intelligence alerts to notify users when a response requires follow-up.

Process historical records

Enabling Action Intelligence affects only the feedback received after you enable Action Intelligence. By default, historical records are not processed. To process historical records:

1. In Setup, open the Reporting > Text Processing > Global Text Analytics Settings > Processings screen.

2. In the Processing option property, select Custom process, and then click Save. The screen updates to show additional, custom processing options.

3. Select one or more of the Custom Processing Options:

   

   Note: If you select Tag Pools as a custom processing option, all feedback records associated with the tag pools you select are processed. If Action Intelligence features are associated with those records, Action Intelligence options are also processed.

4. Click Save.

5. Select the Confirm property, and then click Custom process.