Measure Databricks Assistant impact
This article provides information on measuring the impact of Databricks Assistant in terms of adoption, egagement, and reported productivity gains.
Requirements
In order to measure the impact of Databricks Assistant, you need:
Account admin privileges required to enable system tables. See Enable system tables.
An internal survey to get subjective feedback on Assistant from your team.
Data vs. metrics
You get raw data from the system tables and a survey. To understand Assistant impact, you must analyze and report data as metrics. Metrics are calculated values you use to measure specific aspects or activities related to Assistant impact. This article also refers to metrics as measures.
Tips for measuring Assistant impact
To understand how your organization is using Databricks Assistant, start by measuring its adoption and engagement with Databricks Assistant. This data can be calculated from system tables.
Review your data routinely and make it easily available in a shared dashboard. For a dashboard example and template, see Databricks Assistant system table reference and example.
Assistant impact metrics
The following are recommended measures of Assistant impact, both from system tables and from user feedback. For examples of metrics calculations, download the Assistant events dashboard file from GitHub and read the calculations in JSON. To learn how to import a dashboard file, see Import a dashboard file
Surveying your organization also helps you understand the effectiveness of its engagement with Assistant. See Recommended survey questions.
Measure |
Definition |
Stage |
Data source |
---|---|---|---|
Top users overall |
Users in a given period who interact most frequently with Assistant |
Adoption |
Calculate from Databricks Assistant system table data. |
Submissions data: per workspace and total |
Number of requests submitted to Assistant per workspace and per account |
Adoption |
Calculate from Databricks Assistant system table data. |
Active users by day and month |
Unique users who have received and accepted 1+ suggestions or participated in 1+ chats on a given day. |
Engagement |
Calculate from Databricks Assistant system table data. |
Active users per workspace |
Unique users in a given workspace who interact with Assistant |
Engagement |
Calculate from Databricks Assistant system table data. |
Top job roles using Databricks Assistant |
Numbers of people from each role identified in your org who use Assistant |
Engagement |
|
Top tasks for Assistant |
Most common tasks Assistant helps with |
Engagement |
|
How often do you use Assistant |
Self-reported frequency of Assistant use per user |
Engagement |
|
Top areas of Assistant use |
Self-reported usage areas: SQL editor, notebooks, or both |
Engagement |
|
User satisfaction with Assistant help |
Self-reported satisfaction with Assistant’s answers on a scale of 1-5 |
User satisfaction |
|
Gains in user productivity |
Self-reported increase in productivity gained using Assistant on a scale of 1-5 |
User satisfaction |
|
Amount of time users save using Assistant |
Self-reported percentage of time saved by using Assistant to complete tasks |
User satisfaction |
Recommended survey questions
System tables won’t capture all information about how your organization uses Databricks Assistant. Use survey questions, such as the ones below, to understand how your organization uses Assistant.
What is your role? List roles in your organization
How often do you use Assistant?
Hourly
Daily
Weekly
Monthly
Where do you use the AI Assistant?
SQL editor
Notebooks
Both
What are your main use cases for Assistant?
Fixing errors / troubleshooting
Help writing code
Code refactoring
Writing SQL queries
Writing unit tests
Looking up documentation
How satisfied are you with Assistant’s answers? Scale of 1-5, Not satisfied to Extremely satisfied
Since using the AI Assistant, would you say you’re more productive? Scale of 1-5, Much less productive to Much more productive
How much time do you believe you save on an average task using the Assistant?
0-10%
10-30%
30-50%
50-70%
More than 70%