If you experience slow-loading dashboards and charts, there are changes you can make within Analytics to address this. We recommend investigating and taking action on these key areas:

1. Chart queries

2. Dashboard features

3. Data source stats

4. Dashboard subscriptions

5. Organization's workspace settings

Follow our performance decision tree below to guide you through your performance optimization. We highly recommend going in the order we’ve set above.

Performance issues can seem daunting at first, so we also recommend bookmarking this page for easy reference when needed.

When does Analytics send queries to my data source?

Before we get to the weeds of performance optimization in Analytics, it’s important to know when Analytics sends new queries to your data sources. Sending new queries requires using your data source’s resources, and if your data source receives too many queries at one time, some or many of those queries may time out, leading to the performance issues you’re experiencing.

I. Optimize chart queries

If your dashboard has timeouts or slow-loading charts, start by checking how many queries each chart runs. When a dashboard loads, all chart queries attempt to run at the same time. Charts with many queries put extra load on the data warehouse running those queries, which can slow down the entire dashboard.

1. Does any chart in your dashboard run more than 3 queries?

   1. Yes, I have charts with 3 or more queries.

      1. Reducing the number of queries per chart is one of the highest-impact fixes. Continue to step 2.

   2. No, all charts have fewer than 3 queries.

      1. Move on to section 2 to check dashboard features.

2. Are those queries fetching related data that could be returned in a single query?

   1. はい。

      1. Try combining related queries into a single, optimized query using joins. If you have access to Rovo, you can use the following prompt in Rovo Chat:
         Combine the SQL queries to the Atlassian Data Lake into one query. Assume Databricks SQL dialect. Join the queries using appropriate INNER JOINs (or other joins when needed), and remove redundant logic where possible. Keep the output columns required for the chart. Query 1: <>, Query 2: <>
         After Rovo generates the combined query, make sure to validate that the chart output matches the original chart output. Replace the multiple queries in your chart with the optimized query.

   2. No, the queries are unrelated or serve separate purposes.

      1. Move on to section 2 to check dashboard features.

II. Check the dashboard features

If there’s a specific dashboard with slow-loading charts, we recommend checking its dashboard features as the next step in your performance optimization journey. Modifying these settings may help alleviate your performance issues.

1. Have you checked if your chart queries can be optimized?

   1. Yes.
      We recommend going back to section 1 first.

   2. No.
      Continue to step 2.

2. Does your dashboard require fresh data?

   1. Yes, I need my data to be up-to-date often.

      Update your If cache is expired, refresh dashboard setting to be When dashboard is open in the active tab and set an appropriate cache duration.

      Continue to step 2.

   2. いいえ。

      Consider using the When dashboard is first opened in a tab or Manual refresh method and/or increasing the cache duration.

      Remember—if you ever need fresher data before the next refresh, you can always manually update a dashboard or individual chart’s data.

3. Does your dashboard have any controls?

   1. いいえ。

      Go to step 5.

   2. Yes, I have controls.

      Continue to step 4.

4. Check if Automatically update dashboard when a variable changes is selected in the dashboard settings.

   1. It’s not.

      Continue to step 5.

   2. It is. What should I do?

      Deselect it.

      If you have any controls on your dashboard, you can reduce the number of queries to a single query by deselecting Automatically update dashboard when a variable changes in the dashboard’s settings. Otherwise, a query is sent upon every control change, which could impact your data source.

5. Do you have more than 10 charts?

   1. いいえ。

      Continue to step 6.

   2. はい。

      It’s best to have your most important charts at the top of the dashboard for quick and easy accessibility — we also prioritize loading those first. If you have any charts below the fold (i.e., you need to scroll down to see them), assess whether the charts are needed or if you can break the dashboard down, then link them together.

      This minimizes the number of queries a dashboard sends when refreshing, freeing up your data source resources. Overloading your data source leads to performance issues as its resources are taken up.

6. Do you have any dashboard subscriptions for this dashboard?

   1. いいえ。

      Skip to step 7.

   2. Yes, I’ve got dashboard subscriptions enabled for this dashboard.

      Continue to section 4 to check dashboard subscriptions.

7. After making the recommended changes we’ve discussed so far, is your dashboard still slow to load?

   1. Yes, it’s still slow.

      Continue to section 3 to check your data source stats.

   2. No, it’s resolved.

      Move on to section 5 for additional recommended changes you can make to your organization’s workspace settings to help mitigate problems for new dashboards.

III. Check your data source stats

1. Have you checked your dashboard features already?

   1. いいえ。

      We highly recommend you check your dashboard features first.

   2. はい。

      Every data source has a query log , which provides a breakdown of the queries run against the data source and their performance. You can filter the query log by your slow-loading dashboard to see the query durations for the dashboard’s queries. The query log can be sorted in ascending or descending order by the day the query was run or by the query duration.

      For every data source queried by your dashboard, check if charts from your slow dashboard are displaying error messages or long query durations listed in that data source’s query log.

2. If a chart on your dashboard is slow-running or has errors, review the chart’s performance. Does it have any performance-related errors (for example, timeout errors) or long-running query durations?

   1. いいえ。

      Try to resolve the non-performance errors (for example, syntax errors, relation errors, etc.). Once you’ve done that, check your organization’s workspace settings in section 5.

   2. はい。

      There are a few routes you can take. First, refer back to section 1 and try to optimize the queries for the chart. Second, try to reduce the row limit of your query to have fewer rows returned. Third, consider moving some of the aggregations/manipulations outside of the query and into other Visual SQL steps. These transformations performed outside of the query are post-SQL manipulation, so new queries aren’t sent to the underlying data sources when you add more transformation steps, as long as the initial query doesn’t change.

      Once you’ve done that, check your organization’s workspace settings in section 5.

3. For 3rd-party data sources only: Does the chart have any frequently used queries?

   1. いいえ。

      Skip to step 5.

   2. はい。

      Continue to step 4.

4. Is the query (or queries) pulling data from a data warehouse or performant database?

   1. いいえ。

      Continue to step 5.

   2. Yes, I’m using my own data warehouse/performant database.

      You may need to consider modeling your data or creating materialized views.

      After data modeling, continue to step 5.

5. Is your dashboard still slow to load?

   1. No.
      Move on to section 5 for additional recommended changes you can make to your organization’s workspace settings to help mitigate problems for new dashboards.

   2. はい。

      1. Discuss with your data or engineering team about adjusting your data source’s Workload Manager or maximum user connections. After adjusting the data source settings on your end, continue to section 4 to check your dashboard subscriptions.

IV. Check your dashboard subscriptions

A dashboard subscription is a scheduled email containing a PDF or CSV of your dashboard. Dashboard subscriptions have schedule settings, which you set to tell Analytics when to trigger them. When the dashboard subscription is triggered, new queries are sent to your data source. If all your dashboard subscriptions trigger at the same time, this could overwhelm your data source, leading to timeout errors.

1. Are you using dashboard subscriptions?

   1. いいえ。

      Continue to section 5 to check your organization’s workspace settings.

   2. はい。

      Continue to step 2.

2. Is your dashboard subscription failing (in other words, does your PDF or CSV of the dashboard have charts with timeout errors)?

   1. いいえ。

      Let’s go over some best practices for dashboard subscriptions so you don’t run into performance issues later on.

      If you have any dashboard subscriptions, try to:

      • Stagger them so they don’t send queries at the same time.

      • Offset them to non-peak hours so they don’t send queries during high traffic times, such as business hours.

      Doing the above will help to ensure your data source has enough resources available when your dashboard subscriptions are triggered. In other words, your dashboard subscriptions are less likely to fail.
      After doing this, skip to step 4.

   2. Yes!

      For each data source queried by your dashboard subscription, go to the data source’s query log and look at the Start column to see how many queries are running at once, then continue to step 3.

3. Is your dashboard subscription triggered during peak hours?

   1. No, but it’s still failing.

      Continue to step 4.

   2. はい。

      If you’re triggering a dashboard subscription during peak hours, it’s competing with other potential dashboard subscriptions in addition to the other reasons why Analytics triggers queries, which could explain why your dashboard subscription is failing.

      If you have any dashboard subscriptions, try to:

      • Stagger them so they don’t send queries at the same time.

      • Offset them to non-peak hours so they don’t send queries during high traffic times, such as business hours.

      Doing the above will help to ensure your data source has enough resources available when your dashboard subscriptions are triggered. In other words, your dashboard subscriptions are less likely to fail.

      After doing this, continue to step 4.

4. Is your dashboard subscription still failing?

   1. いいえ。

      Perfect! To prevent future dashboards from impacting your data sources, continue to section 5 to check your organization’s workspace settings.

   2. はい。

      If you’ve already checked your dashboard features and data source stats, contact support for further help.

V. Check your organization’s workspace settings

1. If you’re an organization or Analytics admin, you can take some preemptive actions by changing your organization’s workspace settings. Changing these settings won’t affect existing dashboards, but they will affect any new ones.

2. Is the cache duration low (e.g., one hour or less)?

   1. いいえ。

      You’re all set. Enjoy your faster data exploration.

   2. はい。

      Consider increasing the cache duration. Lower cache durations mean new queries are sent to your data source more frequently. If you don’t expect your data to change often, it’d be a good idea to increase the cache duration to a more reasonable time. Plus, you can always lower the cache duration for individual dashboards, if needed.

As we mentioned earlier, updating your organization settings won’t affect your existing dashboards, so you still need to update the cache duration and refresh method for any existing dashboards experiencing slow performance.

We hope this guide helped you resolve your performance issues. If you have any questions or are still experiencing performance issues, feel free to contact support for further help.