Databricks beginner workflow: bakery data to an AI/BI dashboard

Start with the notebook and inspect the data

Thanks to Databricks for sponsoring this video.

The tutorial uses simulated bakery data in Databricks Free Edition. Treat the notebook as a sequence of Python and SQL cells, with an inspectable output after each step.

1. Import the notebook through the route shown at 1:32.
2. Run the data checks before the ranking query: inspect the locations, sales values and available dates.
3. Rank locations by recorded sales value and create the bar chart. Read the labels and units back against the underlying rows.

Test sentiment on one review first

The review sample contains 12 selected customer reviews. AI sentiment labels need checking against the original text.

1. Select the review sample and compare its date range with the sales data.
2. Run sentiment analysis on one review and inspect the label before processing the rest.
3. Label the remaining sample, review questionable cases and summarise the counts. Keep the original review text available for inspection.

Build and verify the dashboard

The saved sales and sentiment results are brought together in an AI/BI dashboard. Their appearance on the same page does not establish a causal relationship.

1. Prepare the saved query results and import the dashboard layout shown at 14:14.
2. Check the chart values against the notebook outputs.
3. Test refresh, publishing and sharing with the intended audience. Recheck access rather than assuming a published dashboard is visible to everyone.

The takeaway

Use the dashboard to explore the sample. Different sales and review periods, a small review sample and AI labels cannot explain what caused a location’s sales.