Try a piece of the work · 6 min read

Is data analysis right for me? Try this before buying a course

Try a complete beginner data-analysis project: clean a small dataset, compare attendance, explain uncertainty and decide what to investigate next.

Written by Nivaan, founder of OffLadder · Last reviewed 10 October 2026 · Sources listed at the foot of this page

The work starts before the chart

A polished dashboard is the visible result. The less visible questions come first. What does this column mean? Is that record duplicated? Does a blank mean zero, or that nobody entered the number? Are two groups genuinely comparable?

O*NET's business-intelligence analyst profile includes preparing reports, working with information and supporting decisions. It provides one example of analytical work, not a definition of every data job. Different roles may require statistics, programming, domain knowledge or engineering that the exercise below does not cover.

For now, use a spreadsheet or paper. Allow roughly an hour, longer if the tools are unfamiliar. Keep notes on the work that draws you in and the work you keep trying to skip.

Your brief: should a community centre add a Saturday session? The centre runs the same introductory workshop on Tuesday evenings and Saturday mornings. It has funding for one additional session next month. The manager wants to know which time to try.

These are fictional practice records, not observations from a real organisation. Each session has a capacity of fourteen. “Booked” means registered participants; “attended” means people recorded as present. A missing entry is shown as blank.

Your output should contain a clean copy of the data, one comparison and a short recommendation. You do not need an elaborate presentation.

First, preserve the mess

Keep the original table unchanged. Make a working copy and record each alteration. Standardise the capitalisation so that “Tue Evening” and “Tue evening” are not treated as different categories.

Flag the repeated S08 record. For this exercise, treat it as an accidental duplicate because the identifier and every value match, and state that assumption. In real work, you would confirm what the identifier represents before deleting records.

Do not turn the missing S05 attendance into zero. Zero would mean nobody attended. Blank means the attendance is unknown. Those statements can lead to very different recommendations.

Check that attendance does not exceed bookings and bookings do not exceed the stated capacity. Here, they do not. In another dataset, an apparent contradiction could be a data error or a legitimate walk-in policy. Finding the exception is only the beginning of understanding it.

Choose a fair comparison

Calculate two things for the sessions with known attendance: average attendees per session and attendance as a proportion of bookings. The second measure should use total attendance divided by total bookings for those same complete sessions.

After removing the duplicate, the five Saturday sessions contain 61 bookings and 56 attendees. That is an average of 11.2 attendees and an attendance rate of about 91.8%.

The four Tuesday sessions with known attendance contain 39 bookings and 27 attendees. That is an average of 6.75 attendees and an attendance rate of about 69.2%. A fifth Tuesday session has 11 bookings but no attendance record. Keep it visible as missing rather than silently excluding it from the story.

Make a simple chart of average recorded attendance, or present the two values clearly in a table. Label the number of complete sessions. A decorative chart cannot repair a comparison whose missing data has been hidden.

Write a recommendation that does not overclaim

A reasonable first memo would read:

In these practice records, Saturday sessions have higher average recorded attendance: 11.2 people across five sessions, compared with 6.75 across four complete Tuesday sessions. I would provisionally test the additional session on Saturday. Before booking it, confirm the missing Tuesday attendance and check whether differences in instructor, dates or promotion explain the pattern. One duplicate was removed under a stated assumption. These records do not show that Saturday timing caused the difference or that the pattern will continue.

That memo contains a decision, evidence, qualifications and a next check. It is more useful than “Saturday performs 66% better,” which invites the reader to treat a small, incomplete comparison as a settled result.

You could reasonably recommend collecting the missing information before choosing. The quality of the answer depends on how you handle uncertainty, not on agreeing with the sample recommendation.

Now make the exercise slightly less comfortable

Imagine the manager replies: “Tuesday is the only time some participants can attend.” Does your recommendation change?

Attendance is not the only possible objective. The decision might also consider access, unmet demand, costs or who is excluded. The current table does not answer those questions. Write down what additional information you would request and how it might affect the choice.

This is a useful part of the trial. Did the complication make the work more interesting, or did you wish the question would remain purely numerical? Neither reaction decides your career, but both are worth recording.

What to notice about your own response

Did cleaning the categories feel satisfying or merely tolerable? Did you want to know why the Tuesday record was missing? Were you interested in checking the calculation? Did you enjoy explaining the caveats without making the recommendation disappear?

Difficulty with a spreadsheet formula is not the same as disliking analysis. Try a short explanation of the function and repeat the calculation. On the other hand, liking the finished chart does not tell you whether you enjoy the preparation and checking that support it.

Keep the working file, a brief cleaning log and your memo. Together they show the reasoning better than a screenshot of the final chart alone.

What this test cannot establish

Eleven rows cannot tell you whether you would enjoy a full-time analytical role. You have not tested large datasets, coding, statistical inference, stakeholder conflict, recurring reporting or the specific domain of an employer.

A useful next step is to repeat the process with a larger public dataset in a subject you care about, then ask a practitioner what their normal work adds. Search current local vacancies and note which tools and methods repeatedly appear. Learn those in context rather than enrolling in every course labelled “data.”

Do I need advanced mathematics? It depends on the role. This exercise uses arithmetic and careful definitions. Other analytical jobs require much more. Investigate the occupation you are considering, not a universal claim about all data careers.

Can I use AI? It can help explain a formula or question a draft, but do the first reasoning yourself. Check its calculations and keep private data out. A tool-generated answer tells you little about whether you enjoy the work you are testing.

Keep the reaction, not just the spreadsheet

OffLadder helps you explore career directions by trying small pieces of work and using what happens to shape the next step. Tell Laddie whether you enjoyed cleaning, interpreting or explaining the data, and what you avoided. Keep the file and notes in You, then investigate another task through Explore. Start your first reading. You now have something more informative than “I think I am good with numbers.”