Data Analysis with Claude

Analyse data with Claude. Know when to trust the answer.

An AI assistant can now profile a spreadsheet, answer questions asked in plain English and build an interactive report in minutes. What has not changed is the job: ask the right question, check the answer, and present something people can act on. I teach both.

The tools are changing. The fundamentals are not.

Get the 20 prompt cards See four of them first

Twenty habits, one card each

I have taken the habits behind my live course and written each one on a single card: the idea in a few sentences, a prompt you can paste into Claude today, a check that tells you whether to trust what came back, and one thing to try on your own data. Here are four of the twenty. The full deck is the welcome gift when you join the list.

One idea. One prompt. One check.

Before you askCard 1 of 20

Ask for an inventory before you ask for an answer

The most useful first prompt for any new file is not a question. It is a request to describe what Claude can see. You learn the sheet names and the exact column headings, and you catch a wrong file before it costs you an afternoon.

The prompt

I have attached [file name]. For each sheet, list the sheet name and the column headings. Do not analyse anything yet.

Check it

Compare the reply with what you know is in the file. A missing sheet or an unexpected column means stop and look before you go on.

On your own data

Run it on the spreadsheet you use most. Note the column names exactly as Claude reports them. You will quote them in every prompt that follows.

Trust but verifyCard 5 of 20

A confident tone is not evidence

Claude sounds equally assured when it is right and when it is wrong. It can miscount when it answers from impression rather than calculation, and it can misread which column you meant. So verify one easy number by another route before you trust the hard ones.

The prompt

How many rows of data are in the [Sales] sheet? Give the number and state exactly how you counted it.

Check it

Open the file and look at the last row. If the easy number is right, confidence in the rest is earned rather than assumed. If it is wrong, nothing else in the chat is trusted yet.

On your own data

Pick the one figure in your file you can check in ten seconds and make it the first thing you ask for in every new chat.

Better questionsCard 12 of 20

Ask Claude to find the questions

Most prompts ask a question. This one asks Claude to find the questions. It is deliberately open, it works on any file, and it is one of the most useful prompts there is, provided you treat the answer as a list of leads rather than a list of facts.

The prompt

Explore this data and tell me three things worth knowing.

Check it

Pick the finding that surprises you most and ask one follow-up that would prove or disprove it. A finding you cannot check is a lead, not a result.

On your own data

Try the open version, then sharper ones: three risks, three things a new manager should know, three things that changed since last year.

Messy dataCard 17 of 20

Audit first, fix later

If you let Claude fix everything in one pass you cannot check anything. So the first prompt on a raw export asks for a list of problems and forbids any changes. Then you compare Claude's list with what your own eyes found in two minutes of scrolling.

The prompt

I have attached [export file]. Do not fix anything yet. Audit the [Sales Data] sheet and list every data quality problem you can find, with the number of rows affected by each. Check for duplicate rows, blank rows, numbers stored as text and inconsistent spellings. Present the findings as a table.

Check it

Ask about anything the audit does not mention, such as whether a stored total always equals its parts multiplied together. Between you and Claude, nothing goes unnoticed.

On your own data

Bring the export you already distrust. Scroll it first, then run the audit, then compare the two lists.

Get all twenty

Join the list

You get the 20-card deck straight away, as a PDF that reads well on a phone.

After that, one short email when something new goes up: a Thursday demonstration, where I analyse a real business dataset end to end with Claude and say what to trust and what to verify; a new card; and the dates of the next live course, which go to the list before they go anywhere else. Never more than one email a week, usually fewer.

The 20 prompt cards, today
Thursday demonstrations
New cards as they are written
Course dates, to the list first

What I teach

The tools of data analysis are being rewritten. The method is not. Whatever the tool, good analysis still means the same four things, and each one is a habit you can practise.

Clean: make the data trustworthy first
Structure: know the shape before you ask
Analyse: metrics, comparisons, trends, rankings
Communicate: results people can act on

Claude is the tool I demonstrate with, because it is the one that turns a plain-English question into a checked answer fastest. The habits transfer to whatever you use next.

Live course

Data Analysis with Claude

A live, small-group course in four sessions. It runs from attaching your first file to cleaning a messy export, reading customer reviews at scale and bringing in data from the web, and every section ends with an exercise on your own data, so the skills leave the course with you.

It is for mid-career professionals who work in Excel and choose their own tools: analysts, finance and operations managers, and anyone who is handed a spreadsheet and expected to have an answer.

Dates for the next course go to the list first.

Who is teaching

I have spent more than twenty years teaching business professionals to work with their data, and my Excel courses on Udemy have reached hundreds of thousands of learners. Everything I teach starts with why it matters at work, followed by clear steps, followed by practice on your own files. Analysing data with Claude is the same approach applied to a new tool, and it is the most useful thing I have taught in years.

- Ian Littlejohn

More about Ian

Three questions people ask

Do I need a paid Claude plan?

You can read the cards and try most of the prompts with a free Claude account. Attaching files and building charts and reports depends on your plan and your organisation's settings, so check both before you upload anything from work.

Is this for developers?

No. It is for people who work in Excel and are handed data to make sense of. There is no code to write. The skill is asking precisely and checking carefully.

How often will you email me?

At most once a week, usually less, and only when there is something worth sending: a demonstration, a new card, or course dates. Every email has a way to leave.

Twenty cards. Start with one.

Join the list and the deck is yours today. Pick one card, use its prompt on your own data, and do the check.

Get the cards