Claude for data
Use Claude forreal data work.
Write SQL faster, check your results, and explain what you found, while you stay the analyst who decides what the data means.
From a question to a trusted answer.Six lessons for everyday data work.
- 1
Give Claude the context it needs
Describe your tables, grain, and business rules so answers fit your data instead of a generic example.
- 2
Write and explain SQL
Draft joins, window functions, and CTEs, then have Claude walk through each step in plain language.
- 3
Check the numbers
Ask for row counts, edge cases, and sanity checks before a result reaches a dashboard or a stakeholder.
- 4
Clean messy data
Find duplicates, inconsistent labels, and missing values, and get the transformation to fix them.
- 5
Build calculations in your tool
Translate a business question into a Tableau calculated field, Sigma formula, or ThoughtSpot search.
- 6
Explain findings to people
Turn an analysis into a short summary that says what changed, why it matters, and what to do next.
Ask in plain language.Get SQL you can read and check.
You ask
Monthly revenue by region for 2026, but only for customers whose first order was this year.
With context
orders (order_id, customer_id, region, amount, ordered_at)
Claude drafts
with first_orders as (
select customer_id, min(ordered_at) as first_ordered_at
from orders
group by customer_id
)
select
date_trunc('month', o.ordered_at) as month,
o.region,
sum(o.amount) as revenue
from orders o
join first_orders f using (customer_id)
where f.first_ordered_at >= '2026-01-01'
and o.ordered_at >= '2026-01-01'
group by 1, 2
order by 1, 2;Before you trust it: compare the customer count in first_orders with your CRM’s new-customer total for 2026.
AI makes you faster.Good habits keep you right.
- You stay the analyst
- Claude drafts and explains. You decide what the data means and whether the answer is right.
- Verify before you share
- Every lesson ends with the checks that catch a confident answer built on the wrong join.
- Protect sensitive data
- Learn what to share, what to describe instead, and how to follow your company’s AI policy.
- Learn faster, not less
- Use Claude to understand why a query works, so you can write the next one yourself.