Overview
Ask-ai, Datadrew’s conversational analytics copilot, is designed to dramatically speed up analysis, reporting, and decision-making.
It helps you explore data with natural language, draft SQL instantly.
But just like every LLM-powered system, Ask-ai can occasionally produce incomplete, outdated, or partially fabricated responses (the infamous “AI hallucination”).
This guide tells how to avoid these inconsistencies using our DREW framework.
Prompt Like a Pro Using the D.R.E.W. Framework
(Data Drew's branded method for getting clean, predictable responses)
Letter | Principle | Ask Yourself | Quick Check |
D | Define | “Did I specify the exact metrics or KPIs?” | ROAS, AOV, CPA, net profit, LTV |
R | Restrict | “Did I set a tight scope or date range?” | Q4 2024, last 14 days, top 10 campaigns |
E | Ensure Availability | “Does Ask-ai have access to this data today?” | have i connected the right connectors/integrations |
W | Yielding | “Will this answer help me take an action?” | ‘Top 5 drivers’, ‘worst 10 SKUs’, ‘% change’ |
If your prompt satisfies D.R.E.W. you would idealy start getting the right answers.
Reporting Bugs or Inaccuracies
If you feel Ask-ai returned an incorrect number, or contradictory logic:
Try checking if you followed Drew Framework, or give it a try again
Just share your feedback within the answer section, we will thoroughly check it & come back to you with an improved version. The feedback section looks like:
