Data analysis

01The problem

Kolanut's data is clean. Now the managing director asks: "Did we do better in the first half of 2026 than the first half of 2025, and where did the money come from?"

Answering needs only a few calculations, used carefully. Most business analysis is built from the same small toolkit.

02The concept

Most business analysis is built from a small toolkit of calculations. The skill isn't the arithmetic; it's choosing the right one and knowing its traps.

Revenue at Kolanut is calculated per order line as:

revenue = quantity × unit_price × (1 − discount_pct ÷ 100)

The toolkit

The sections below take them in turn, each with Kolanut's real figures.

Totals, counts and distinct counts

A count and a distinct count answer different questions: 4,266 order lines came from 90 customers. Mixing them up is a common source of wrong numbers in reports.

Mean, median and mode

Three ways to describe a "typical" value:

Why do the mean and median differ? Most lines are small or medium, but a long tail of big lines (405 of the 411 lines worth ₦400,000 or more are wholesale) pulls the mean up. The median only cares about the middle, so the tail doesn't move it.

Which to use?

• When values are roughly symmetric, mean and median are close; use the mean.
• When a few values are much larger than the rest (salaries, order sizes, house prices), the median is the fairer "typical" figure. Report both when they differ a lot.
• The mean is still the right one when you need totals: mean × count = total. You can't do that with a median.

Minimum, maximum and range

The extremes are a quick sanity check as well as a fact. Kolanut's order lines run from ₦3,420 (one pack of bottled water with 5% off) to ₦713,400 (29 packs of body lotion). Quantities run from 1 to 30 packs.

Always look at the minimum and maximum of every number column before analysing. A negative quantity, a price of zero, or a date in 2099 tells you there's cleaning to do.

Share of total

Share = part ÷ total × 100. It answers "how much of the whole does this part make up?"

Shares let you compare parts of very different sizes, and they always add up to 100%, which makes a good check.

Growth rate

Growth = (new − old) ÷ old × 100. Divide by the old value: the starting point.

Growth can be negative (a fall). Report a fall as a fall: "fell 46.6%", not "grew −46.6%".

Rates: comparing groups of different sizes

Lagos brings in far more revenue than South South, but it also has far more customers: 34 against 11. To compare fairly, divide by the size of the group: revenue per customer, orders per customer, sales per rep. A rate turns "bigger" into "better or worse".

Averages of groups: the trap

Average line value by channel:

The overall average is ₦194,689. But if you average the three channel averages, you get ₦159,413, which is wrong. That simple average treats Kiosk's 777 lines as if they counted as much as Wholesale's 2,181. To combine averages, go back to the totals: total revenue ÷ total lines.

Compare like with like

Compare January with January, the first half of one year with the first half of the next. Comparing December with January mostly shows the festive season, not performance. And if the data stops at June 2026, compare January to June 2026 with January to June 2025, not with all of 2025.

Rounding

Round for display, not for calculation. If you round each region to one decimal place first and then calculate from the rounded numbers, the last digit of your answer can change. Calculate with the full figures, then round the result.

03Example

Revenue by region, first half (January–June) of each year, in millions of naira:

Growth. (290.7 − 244.2) ÷ 244.2 × 100 = 19.0%. The business grew. (From the unrounded totals, ₦244,163,070 and ₦290,730,455, it's 19.1%: an example of rounding changing the last digit.)

Share. Lagos brought in 152.8 ÷ 290.7 × 100 = 52.6% of H1 2026 revenue (52.5% from the unrounded figures): more than half the business comes from one region. A year earlier it was 48.4%. Kolanut is depending more on Lagos, not less.

The exception. North West went the other way, from 31.1 to 16.6: a fall of (31.1 − 16.6) ÷ 31.1 × 100 = 46.6%. A total that grew 19% hides a region that nearly halved. That's why analysts always break totals down.

A rate explains it. North West had 10 ordering customers in H1 2025 and 11 in H1 2026, so it didn't lose customers. But its order lines fell from 176 to 91. Lines per customer fell from 17.6 to 8.3: the same shops ordered less than half as often.

04Walkthrough

Mean versus median on real data. Across all 4,266 of Kolanut's order lines:

• The mean order line is worth ₦194,689.
• The median order line is worth about ₦166,680.

The mean is higher because wholesalers place many very large orders (up to ₦713,400 on a single line), which pull the average up. If a manager asks "what does a typical order line look like?", the median is the more honest answer.

Averages by group often tell the real story. Average quantity per order line:

One overall average (13.8 packs) would describe none of these customers well.

Try it yourself. In a spreadsheet with orders.csv:

1. Add a revenue column with the formula above.
2. Calculate the total, count, mean and median of revenue. Check you get ₦830,541,245, 4,266, ₦194,689 and ₦166,680.
3. Find the minimum and maximum. Do they look sensible?
4. Work out what share of all revenue came from the single largest line (713,400 ÷ 830,541,245 × 100). It's tiny, under 0.1%, so no one line distorts the total, even though big lines do lift the mean.

Summary

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