Best practices for working with PowerBuddy for Data Analysis

Be specific

If you are trying to retrieve the data for a specific school year, mention the entire school year such as ‘2022-2023’ in the question, instead of using just 23 or 2023.

Mention the keyword ‘school’ in the natural language question or prompt instead of using abbreviated school names.
For example, you can ask PowerBuddy for Data Analysis to retrieve the count of incidents created for students in each grade during the 2022-2023 school year at Redwood High School.

If you are asking a question about a user by name, specify the type of user, such as the teacher named John Doe or the student named John Doe.

Providing specific information, helps PowerBuddy for Data Analysis to understand exactly what is needed, reducing ambiguity and improving the accuracy of results.

There can be multiple ways to write queries. If you are expecting the generated query to use table A and table B, but if it uses table A, table C, and table D, specify your preference of using tables A and B for the query. Although the generated query might be right, specifying your preferred table will generate a query in the way you need it.

Provide specific conditions or exact timeframes

When formulating questions about specific dates, use clear and complete date formats to ensure accurate responses.

For example, instead of asking How many students were present on 4/29? which may cause ambiguity regarding the year, rephrase the question as How many students were present on April 29, 2024?

Instead of asking:

Provide list of students in 12th grade who are absent for more than 5 days in the last academic year

Rephrase the question as:

Provide list of students in 12th grade who are absent for more than 5 days in the academic year 2022-2023

In Incident Management queries, differentiate behaviors and actions to ensure accurate interpretations of the data.

For example, How many detention incidents have occurred this year? may not effectively recognize "detention" as a specific disciplinary action. Rephrase query as How many incidents have occurred this year in which a student received an action of detention? to provide clear context. This explicitly defines “detention” as an action, making it more clear and avoiding ambiguity.

Including specific keywords or context such as grade levels or specific demographics narrows down the search scope and ensures the most relevant data is retrieved.

Use natural language

If the response contains duplicates, explicitly ask a follow up question in the chat:

Remove duplicates from the data.

If you're unsure about specific information, engaging in a chat or conversation can be helpful. This allows you to clarify details, and refine your query for more accurate results.

For example, if you want to retrieve the distribution of historical grades on a particular course, but you are uncertain about the exact course number or course name, you can ask PowerBuddy for Data Analysis:

Get a list of available courses with course numbers.

Once you get the exact course number and course name, you can refine the query accordingly and ask PowerBuddy for Data Analysis:

List the distribution of historical grades for 9th-grade students in English 9 with course number ENG100.

Asking follow-up questions

You can rephrase a question if the generated SQL query does not provide the expected results. Chat with PowerBuddy for Data Analysis to make any modifications to the prompt or question.

For example, you ask PowerBuddy for Data Analysis to show me all the students who received current grades of D or F.

After the results are retrieved, if you want the school names to be included in the results, you can enter a follow-up prompt like Can you also show the school names associated with these students?

To narrow down the list of students to specific grade levels and schools, you can ask another follow-up question as filter this data to display only students in grades 10,11,12 at Redwood High School.

To obtain the list of students who meet the above criteria and have been absent for 3 or more days, you can further ask, show me the students who meet this criteria and have been absent for 3 or more days.

Follow-up questions work best when you try to maintain the context from previous question.

Data retrieval vs Analysis prompts

It is recommended to separate data retrieval prompts from analytical prompts. First retrieve the data, then create a pivot table or chart.

Please note that analytical or summarization requests will not work on the Data dashboard.