Getting recommendations from Fern starts with connecting your data sources and running an initial scan. Here's how to set it up.

What you'll need

Before turning on recommendations, connect at least one support integration - Crisp, Intercom, or Zendesk. You can also connect GitHub or Linear to track product changes, though that's optional.

If you haven't connected integrations yet, the setup flow will prompt you to do it.

Starting the setup

Head to your Inbox or the Signals dashboard in your sidebar. If recommendations aren't enabled yet, you'll see a prompt to set up monitoring.

Click through to open the setup dialog.

Connect integrations

If you don't have any integrations connected, the setup dialog will show you links to:

  • Support tools (Crisp, Intercom, Zendesk)

  • Codebase tools (GitHub, Linear)

Connect at least one support tool. You can always add more later - the goal is to get something connected so Fern has data to work with.

Connecting both support and codebase tools gives Fern a fuller picture, but start with support if you're only using one.

Choose a lookback period

Pick how far back Fern should analyze for your first scan:

  • 7 days - Recent activity only

  • 14 days - Good starting point for most teams

  • 31 days - More comprehensive history

  • 90 days - Quarterly view

  • 365 days - Full year (takes longer)

Starting with 14 or 31 days usually gives you enough data to see patterns without being overwhelming.

Run the scan

Hit "Start monitoring" and Fern will begin analyzing your data. The scan imports tickets and product changes, identifies gaps, and creates prioritized recommendations.

Expect the scan to take 5-30 minutes depending on data volume. You can leave the page - Fern keeps working in the background.

Scans handle up to about 3,000 tickets. If you have more, Fern focuses on the most recent and relevant items.

Monitoring progress

The Signals dashboard shows scan status while it runs - setup, in progress, complete, or error. When it finishes, you'll see:

  • Recommendations populating your Inbox

  • Analytics in the Signals dashboard

  • A completion status on the monitoring card

What happens next

After your first scan completes, Fern automatically runs daily scans in the background. This means new documentation gaps get surfaced as they emerge, without you needing to trigger scans manually.

Your Inbox will stay updated with fresh recommendations based on recent support patterns and product changes.

Running manual scans

Even with daily automatic scans, you can trigger manual ones whenever you want:

  • After connecting a new integration

  • Following a major product launch

  • To analyze a specific historical period

Just go to the Signals dashboard, click "Check for new tickets," pick your lookback period, and start the scan.

Daily automatic monitoring keeps you ahead of documentation debt without constant manual checking.

Troubleshooting

No recommendations after scanning

If the scan completes but your Inbox is empty:

  • Check that your integrations have data during the lookback period you chose

  • Make sure you have some baseline documentation already - Fern compares what users are asking against what exists

  • Try a different or longer lookback period

Scan fails

If you see an error:

  • Verify your integrations are still authenticated

  • Check that the integrated tools are accessible (no API outages)

  • Try a shorter lookback period if you have high data volume

  • Use the retry button to try again

Can't find the setup button

If you don't see a setup prompt, you may have already enabled recommendations. Check your Inbox for existing recommendations or look at the Signals dashboard to see your monitoring status.

Next steps

Once recommendations are flowing:

  • Review what's in your Inbox and assign high-priority items to Fern

  • Explore the Signals dashboard to understand the data behind recommendations

  • Connect additional integrations if you want more comprehensive coverage

The more data sources you connect, the better Fern gets at spotting documentation gaps.

Was this helpful?