Your first analytics project

01The problem

Kolanut's HR manager is worried: "It feels like we keep losing people, especially in customer service. Is that true, and what should we do?"

This lesson walks through a small but complete analysis of that question. The course's final project then asks you to take it further on your own.

02The concept

A complete analysis, even a small one, has five parts. You'll produce them for the final project:

1. The question, in measurable form.
2. The data you used and anything you cleaned or excluded.
3. The analysis: the calculations and a chart or table.
4. The finding, in one or two sentences.
5. The recommendation and its limits: what the data can't tell you.

What each part looks like

A note on small numbers

Kolanut has 80 employees. When you split them by department, some groups have only 6 or 8 people. One resignation in a group of 8 moves the rate by 12.5 percentage points. Report the counts next to the rates ("3 of 8", not just "37.5%"), and be careful about strong conclusions from small groups.

Look for a hidden third factor

Leavers earned less: a median of ₦355,000 a month, against ₦525,000 for staff who stayed. It's tempting to conclude "low pay makes people leave". But every leaver was Junior or Mid level, and junior staff earn less because they're junior. Job level could explain both the lower pay and the resignations.

A factor that is linked to both things you're comparing is called a confounder. Before claiming that A causes B, ask: is there a C that drives both? Here, the fair comparison is pay within the same job level, and with only 11 leavers, the numbers are too small to settle it. Say so.

Checks before you share

1. Do the totals add up? Department staff should sum to 80; resigned to 11.
2. Does one number match a second method? A pivot count and a COUNTIFS should agree.
3. Are the counts shown next to the percentages?
4. Is every number in the summary also in the analysis? No figure should appear from nowhere.
5. Could someone else repeat it? Your data and method notes should be enough.
6. Is the limit stated? What the data can't tell you, in one sentence.

03Example

Question. Of everyone Kolanut has employed since 2018, what share has resigned, and does it differ by department?

Data. employees.csv: one row per employee, with department, job_level and status (Active or Resigned).

Analysis.

Finding. Customer Service has the highest share of resignations (3 of 8 people, 37.5%), about three times the company-wide rate. By job level, all 11 people who left were Junior or Mid level; no Senior staff or Managers resigned.

Recommendation. Hold short exit and "stay" conversations with Customer Service staff to understand why people leave, and review junior pay and workload there. Limit: these are small numbers, and the data has no reasons for leaving, so treat this as a signal to investigate, not proof.

04Walkthrough

How to produce the table above in a spreadsheet:

1. Open employees.csv in Google Sheets or Excel.
2. Insert a pivot table (Sheets: Insert → Pivot table; Excel: Insert → PivotTable).
3. Put department in Rows.
4. Put employee_id in Values, summarised by COUNTA (Sheets) or Count (Excel). That's the Staff column.
5. Put status in Columns. You now have Active and Resigned counts per department.
6. Next to the pivot, calculate Resigned ÷ Staff × 100 for each department.
7. Sort by the percentage, largest first.

05Practice

06More practice

Optional drills. They don't count towards the certificate, but they're the fastest way to make this lesson stick. Several use a different dataset from the lesson on purpose: if you can do the same thing on unfamiliar data, you've really learned it.

07Check your understanding

When you've finished this lesson, take the final assessment, then open the final project from the course page.