Zointly can analyze free-text, CSVs, exports from Intercom, Zendesk, Typeform, Notion databases, and more. The engine is flexible — but giving it clean, well-structured input will always produce sharper output. Here's the definitive guide to formatting your data for best results.
The Ideal Structure
The minimum viable input is a single column of text entries — one customer data point per row. If you have metadata (user ID, plan tier, date, NPS score), include it in adjacent columns. Zointly uses that information to weight signals and enable cohort filtering.
- Column 1 (required): The customer text — feedback, ticket body, survey response, review
- Column 2 (recommended): User or customer ID (for deduplication)
- Column 3 (recommended): Date (ISO format: YYYY-MM-DD)
- Column 4 (optional): Plan tier or segment label
- Column 5 (optional): Source channel (support, NPS, app store, etc.)
Common Mistakes to Avoid
Including Agent Responses
Zendesk and Intercom exports often include the full conversation thread — customer message, agent reply, customer follow-up. Only feed Zointly the customer-authored text. Agent language about workarounds and apologies will pollute the signal.
Keeping Auto-Responder Rows
Filter out any rows containing "Your ticket has been received," "Thanks for reaching out," or similar auto-responder language before exporting.
Encoding Issues
Save your CSV as UTF-8. Many Excel exports default to Windows-1252, which breaks emoji and non-ASCII characters. In Excel: File → Save As → select CSV UTF-8 (with BOM).
Tool-Specific Tips
- Intercom: Export from Reports > Conversations. Filter to 'resolved' status. Delete the 'Assignee' and 'Team' columns before uploading.
- Zendesk: Use the Ticket Export API for clean exports. The UI export sometimes truncates long descriptions at 1,024 characters.
- Typeform: Export response CSV directly from the Results tab. Keep the 'Submitted At' column for date filtering.
- Notion database: Export as CSV (Markdown & CSV). The first column is usually the page title — make sure the feedback text is in a dedicated property column.
If you're not sure your CSV is clean, paste 10–20 rows as plain text first to validate the analysis output before uploading the full dataset.
Optimal Dataset Size
Zointly is designed for datasets of 50–1,000 entries per run. Below 50, the statistical grounding for Demand Scores is weaker. Above 1,000, consider splitting by time period or user segment to get more targeted insights alongside a broader picture.