Case Study: The Annual Review

It's January, and the founder is presenting to investors. They've got ten minutes, and they want one page which sums up the whole year. All of your work comes down to this. What really matters?

• Ten minutes with investors
• One page for the whole year
• What really matters?

The headline numbers

Keep it to a few numbers. Revenue, profit, and margin. The best cafe and the lowest. The top product. And the weekend share. Run it. About three and a half million in revenue, two point two million in profit, and a margin of sixty five percent. Osaka leads, and Cairo is at the bottom. Latte is the star. And weekends bring in a third of revenue, from just two days in seven. Every line has a story behind it, and you've already told most of those.

import pandas as pd

s = pd.read_csv("sales.csv", parse_dates=["date"])
stores = pd.read_csv("stores.csv")
y = s[s.date.dt.year == 2025]
print(f"Revenue ${y.revenue.sum():,.0f}, profit ${y.profit.sum():,.0f}, margin {y.profit.sum() / y.revenue.sum():.1%}")
top = y.groupby("city").profit.sum().sort_values()
print("Best:", top.index[-1], "Lowest:", top.index[0])
print("Top product:", y.groupby("product").profit.sum().idxmax())
print("Weekend share of revenue:", round(y[y.date.dt.dayofweek >= 5].revenue.sum() / y.revenue.sum(), 3))

Revenue $3,454,040, profit $2,243,562, margin 65.0%
Best: Osaka Lowest: Cairo
Top product: Latte
Weekend share of revenue: 0.332

The story of the year

Numbers aren't a story. So here's one. Harbor Coffee grew by about fifteen percent, with steady margins. The summer dip is seasonal, and it isn't a warning. Two large cafes need a closer look, because they're using a lot of floor space for their sales. The new app wins customers, though baskets are smaller. Next year's priorities, keep the middle group of customers, fix sandwich margins, and find out why growth is slowing.

What you just did

Look back at these case studies. Each one started with a confident assumption, and the data changed it. The panic was a season. The winning test wasn't a clear win. The best seller wasn't the best earner. That's what data science is. It isn't charts or models. It's replacing a confident guess with something that's closer to the truth.

Summary