Government exam aspirant to data analyst: explaining the gap years

A letter to Arjun, three attempts in

Arjun, you wrote that you have spent four years preparing, that you cleared prelims-type stages twice without a final selection, and that you are now 27 and worried the years were wasted. Let me be direct about what I think, because vague encouragement will not help you.

The years were not wasted, and they were not work experience. Both are true. An interviewer for an analyst role will not count exam preparation as industry experience, and you should not ask them to. But you did build things that are genuinely useful, and some of them are already sitting in your notebooks.

What your preparation actually produced

Look at what you kept, not what you studied:

• A mock-test log — dates, test series, score, attempts, correct and wrong answers, often by section. That is a time-series dataset about your own performance.
• An error notebook — questions you got wrong, grouped by topic and reason (concept gap, calculation slip, misread, time pressure). That is categorical data with a clear question attached.
• Topic-wise accuracy sheets — you probably already calculated accuracy per topic by hand.
• Data-interpretation practice — hundreds of tables, bar charts and pie charts read under time pressure. You can compute a percentage change faster than most analysts.

What preparation did not give you: SQL, a spreadsheet beyond basic formulas, a dashboard tool, working with messy data someone else produced, and explaining a finding to a manager who disagrees. That is the gap, and it is learnable.

Your first project: analyse your own mock-test log

Do not start with a stranger's dataset. Start with the one you understand best. If you prefer not to use your real scores, generate a synthetic version with the same structure. Here is the shape of the analysis:

import pandas as pd
pd.set_option("display.width", 120)
pd.set_option("display.max_columns", None)

# Synthetic mock-test log: one row per section per test
log = pd.DataFrame({
    "test_no":   [1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6],
    "section":   ["quant", "reasoning", "english"] * 6,
    "attempted": [22, 28, 25, 24, 30, 26, 25, 29, 27, 27, 31, 26, 26, 32, 28, 28, 33, 27],
    "correct":   [14, 23, 18, 17, 25, 19, 16, 24, 21, 21, 27, 20, 19, 28, 22, 23, 29, 21],
    "minutes":   [20, 18, 15, 20, 17, 15, 20, 17, 14, 20, 16, 15, 20, 16, 14, 20, 15, 15],
})
log["accuracy_pct"] = (log["correct"] / log["attempted"] * 100).round(1)
log["correct_per_min"] = (log["correct"] / log["minutes"]).round(2)

summary = log.groupby("section").agg(
    tests=("test_no", "nunique"),
    mean_accuracy=("accuracy_pct", "mean"),
    first_three=("accuracy_pct", lambda s: s.iloc[:3].mean()),
    last_three=("accuracy_pct", lambda s: s.iloc[-3:].mean()),
    correct_per_min=("correct_per_min", "mean"),
).round(1)
summary["change"] = summary["last_three"] - summary["first_three"]
print(summary.sort_values("change", ascending=False))

On this synthetic log, quant starts as the weakest section and improves the most between the first three and last three tests, while reasoning is strongest throughout. The analyst's job is to say what that does and does not show. Six tests is a small sample. Tests from different series may differ in difficulty, so a rise in accuracy might be an easier series rather than improvement. And accuracy rose while attempts also rose, which is a different claim from "I got better at the same questions".

Write those caveats into the project. Then extend it: join the error notebook as a second table and ask which error reason fell over time. Rebuild the same thing in SQL and in a one-page Power BI report. That is a complete, honest first portfolio piece, and it is about something you can discuss for twenty minutes without notes.

After that, do one project on a business-shaped dataset — synthetic sales, orders or support tickets — so your portfolio is not only about exams.

Junior MIS or data analyst: where your mock-test habits fit

Arjun, I would aim for junior data analyst, MIS analyst or reporting analyst roles, and be open to data operations roles where accuracy under volume matters. These value the precision and arithmetic you already have. Use the free skill check on your resume once the first project is done; it will show which analyst skills are still not evidenced.

Answering "What were you doing for four years?"

Do not apologise and do not over-explain. Something like this:

"After graduating I spent four years preparing for government recruitment exams. I reached the later stages twice but was not selected, and last year I decided to move into data analytics instead. Preparation gave me fast arithmetic and a lot of practice reading tables, and it also left me with a detailed log of my own mock tests. I learned SQL and Power BI by analysing that log — including why my improvement in one section might partly be an easier test series. Since then I've built a second project on synthetic sales data, which I'd be glad to walk you through."

That answer is two sentences of history and the rest is evidence. Keep it that way.

If a structured route would help, the Data Analyst programme is organised around weekly deliverables with mentor review, which suits someone used to a disciplined timetable. And if you are still deciding between analytics and modelling work, read data analytics vs data science first.

Good luck, Arjun. Decide once, then build.