Best AI Business Intelligence Software in 2026

Configuring backup and restore procedures for critical data in Data Cloud involves defining clear objectives (RPO/RTO), identifying and classifying critical data, selecting an appropriate backup method and frequency, implementing a robust storage strategy (like the 3-2-1 rule), and performing regular testing. You must also ensure data security through encryption and document your procedures to create a comprehensive policy.

The first step is to make calls to Mode API to periodically export report definitions, query contents and dashboards. A script can be scheduled (using cron or an orchestration platform such as Airflow) to visit the /reports, /queries and /spaces endpoints and store the JSON responses in a secure cloud storage bucket (AWS S3, GCP Storage or Azure Blob). This provides you with a version-backed up analytics asset. Keep each export time stamped, in case you need to roll back to a certain time in case some thing is deleted or corrupted.

In the case of underlying data, Mode does not store it, instead making requests to your connected warehouse (Snowflake, BigQuery, Redshift, etc.). In that way you support your actual data at the warehouse level. Don't forget to set snapshotting or time-travel backup option there as Mode only restores queries and reports and not the underlying tables.

Concerning restore: when a user deletes a report, they can rebuild the report out of the previous supported JSON using the API or move it in manually into the Mode UI. To ensure complete disaster recovery, you should document your process in a clear manner, including the backups that are available, the place they are and who is authorized to restore.

Finally, do not disregard access controls and encryption, encrypt all your backup files, encrypt them during transit, and only restore them to a position of an administrative role. In order to maintain it seamless, do a test restore once every three months - even basic dry run so that you do not get to find out that your API key is broken on the day you really need it.

The best way to reliably deploy these notifications and related features across engaged users in regions around the world is to create separate Metabase notification templates (for Slack or email alerts) for each supported language, one in English, one in Spanish, etc. and have each user routed to that version based on their resident language. You can keep track of that language in the identity provider (such as Okta, or Azure AD) you already have in use to manage the user context for your dashboard. You can maintain language preferences in your internal user DB, or even as group tags in Metabase if your teams are distributed in regions. From there, wherever possible just use your delivery option and send the appropriate localized message (SMTP, slack webhook, or API automation).

To bring Apache Superset to the WCAG 2.2 AA compliance, you will want to consider both automated accessibility evaluation and manual usability testing because data visualization tools are not exactly easy to accessibility-wise. Begin with automatic scans with axe-core, Lighthouse, or Pa11y to identify low-hanging bugs such as the absence of ARIA labels, color contrast violations, and inadequate heading structures. Dynamically rendered components are also common in Superset dashboards and filters, and thus you will need to run these tests on load and once interacted with (the application of filters or a chart type change).

Then, keyboard navigation tests are required - all controls (filters, dropdowns, chart selectors, date pickers) are to be accessible and controllable only using the keyboard, with visible focus indicators. WCAG 2.2 puts an extra emphasis on the look of focus and drag gestures and therefore they should test whether the user can accomplish all the interactions without having to drag or hover.

And you will have to check if screen readers are compatible with such a tool as NVDA (Windows) or VoiceOver (Mac). Ensure that charts include descriptions that can be accessed (through ARIA or in embedded SVGs), and that tabular data are exportable to a readable format to allow assistive technology users. To create color and contrast, ensure that your themes are at the 4.5:1 minimum ratio and that color is not the sole form of passing information e.g. use colored lines with shape or text labels in legends.

Lastly, perform a manual review among test users who depend on assistive technology and also test against WCAG 2.2 AA success criteria, including recent capabilities such as 2.4.11 ( Focus Not Obscured ) and 3.2.6 ( Consistent Help ). The ideal practice is that you can add accessibility tests to your CI pipeline - meaning that any new Superset customization or any new Superset plugin will be automatically tested. It is not a single audit but a long-term practice to make your analytics platform actually inclusive.

To track license utilization and right-size costs in Power BI, you can use the built-in Usage Metrics Reports, the Microsoft 365 Usage Analytics reports, and create a custom report from the Admin Monitoring workspace. These reports help identify which users are actively using Power BI and which licenses might be underutilized.