How to Turn Salesforce Cases and Opportunity Notes into Product Insights

Salesforce feedback sits in five places across service and sales. Cases carry the volume; opportunities carry the dollars.

Pick your first slice by where growth is stuck. For churn, pull Cases from accounts renewing in the next two quarters. For win rate, pull the last two quarters of closed-lost and closed-won deals and look for what winners had that losers lacked.

Salesforce counts labeled records well, and its AI labels or summarizes one record at a time. Short of a custom build, nothing native groups free text into themes you haven't defined.

• Reports: Count Cases by Case Reason and sum closed-lost Amount by loss reason. Fast, but only for feedback someone already labeled.
• Case Classification and work summaries: Classification predicts values "like priority, reason, or type" from a case's text, and work summaries draft "a summary, issue, and resolution" from a chat (Trailhead). Neither creates a new category.
• Conversation Intelligence: Flags competitor, product, and custom keywords you set on sales calls, with reports such as Opportunities with Competitor Keywords (Trailhead).
• Data 360 and Agentforce: Data 360 indexes case notes and transcripts so agents can ground answers in them (Trailhead). A theme report on top is yours to build.

Split your catch-all Case Reason. A value like "Feature Request" becomes a junk drawer for every idea customers send, which is exactly the list product needs broken apart. Rule of thumb: once one value holds more than about 10% of cases, read 50 of them and split it with a feedback taxonomy.

Use reports for fields, and Data Loader or the Salesforce CLI for related objects. Comments, emails, and Lightning notes don't come out in a report export.

1. Cases: Report on Case ID, Subject, Description, Case Reason, Created Date, Account ID, and Contact Email, minus password resets and spam. Add CaseComment and EmailMessage bodies only if most Descriptions are thin.
2. Opportunities: Report on Opportunity ID, Account ID, Stage, Amount, Close Date, loss-reason fields, and Description. Include closed-won deals for comparison.
3. Notes: Query ContentDocumentLink by LinkedEntityId with your opportunity IDs, since that object must be filtered by an ID (Salesforce Developers). Then pull the matching ContentNote records and decode the body.
4. Clean the file: One row per piece of feedback, with text, date, source, contact email, Account ID, and Opportunity ID. Drop signatures, auto-replies, and logistics like "sent the order form."

One quarter of Cases and two quarters of opportunities is enough for a first pass. Budget half a day with an admin, most of it on notes.

Give every request an Account ID and, when it came from a deal, an Opportunity ID. Then roll revenue up per request, counting each account once. Two Salesforce patterns make this repeatable.

• A Product Gap object: A custom object with lookups to Opportunity and Account and a single-select Product Area picklist. Reps log one record per gap; a report sums Amount by area. Skip multi-select picklists, which reports show as one semicolon-separated value and can't cleanly group by.
• Account roll-up fields: A roll-up summary field on the Account can sum its Opportunities, as Savio's guide shows (Savio). Build Open Pipeline or Lost to Product Gap, and any tool that reads Account fields can use them.

Keep three numbers apart; each answers a different question.

Never add them. ARR is a fact, pipeline is a forecast, and closed-lost revenue is a story about the past. Don't total a column across themes either: a deal citing three gaps counts in full under each.

Verify every closed-lost reason. "Missing feature" is the least awkward reason a rep can pick. Before a lost deal counts, find the buyer saying it in a note, email, or call.

A worked example (illustrative numbers). Two themes top a team's first pass.

By the picklist, the report builder looks five times bigger. The notes show only 3 of those 9 buyers asked for it. Audit log export is the stronger bet: three times the paying accounts, two renewals inside 90 days, and evidence you can name.

To compare tools that total revenue per request, see the best tools to prioritize feature requests by revenue. For why raw request counts mislead, see why feature voting fails.

Deck uses Salesforce as account context, not as a feedback source. Deck turns support tickets, sales calls, interviews, surveys, and Slack into Themes and a ranked Opportunity Backlog; Salesforce tells it which account said what, and what that account is worth.

Choose by who does the work. If reps will log requests in Salesforce, pick a tool with a widget on the record. If requests are buried in text nobody tags, pick one that reads the text.

Productboard, Savio, and Canny are only as good as your reps' logging. Enterpret reads Salesforce text directly, at enterprise contract sizes. Deck fits a small team whose evidence lives in support, calls, and Slack; start on the Free plan and move to Business when you connect Salesforce.