Analyze a Public Dataset & Write a Metrics Readout
Part of the Product Management Roadmap
In short
Take a real dataset, find the story in it and write it up like a PM would for their team. It's one of the best portfolio pieces you can make.
Practice the full loop — pull data, find a real pattern, and communicate it in a short, decision-oriented readout, the way you'd report findings to a team.
Think of it like this
The match report
A sports reporter gets a scorecard full of numbers. The story isn't the numbers; it's "a late collapse in the last five overs cost the match". Your readout does the same: find the one story the data tells.
Watch it happen
From raw file to readout
1/5Pick a public dataset, like an e-commerce orders dataset on Kaggle. Read its column descriptions first.
The big ideas
What you need to know
Lead with the question
A clear question keeps you from drowning in columns. Everything else serves it.
Headline first
Start the readout with the main finding in one sentence. Details come after.
State the caveats
Mention missing data and what you can't conclude. Honesty makes the rest believable.
End with a next step
Recommend one action or one follow-up analysis. A readout should help someone decide.
See it in code
Walk through the code
1import pandas as pd2 3df = pd.read_csv("orders.csv", parse_dates=["order_date"])4df["month"] = df["order_date"].dt.to_period("M")5 6monthly = df.groupby("month").agg(7 orders=("order_id", "count"),8 customers=("customer_id", "nunique"),9)10monthly["orders_per_customer"] = monthly["orders"] / monthly["customers"]11print(monthly.tail(6))1/3Load the CSV with real dates, then label each order with its month.
A starter pandas script: monthly orders, unique customers and orders per customer. Rename the columns to match your dataset.
Save yourself the pain
Common mistakes
Oops
Dumping twenty charts into the readout.
Do this instead
Pick the two or three charts that support your headline. Put the rest in an appendix.
Oops
Claiming one thing caused another from a correlation.
Do this instead
Say "is linked with" unless you ran an experiment. Suggest a test to check cause.
Your mission
First milestone: question and first chart
Quick check
What should the first line of a metrics readout usually be?
Go deeper
- 1Find a public dataset (Kaggle, government open data, a company's public metrics).
- 2Use SQL or a spreadsheet to find one genuinely interesting pattern.
- 3Write a half-page readout: what you found, why it matters, what you'd do next.
Stuck on this topic? Ask an Insider
Get 1:1 guidance from people who've walked this exact path — on the InsideEdge app.
