Planning and Scoping Your Capstone Project

Choosing a Dataset and a Question

Start from a question you're genuinely curious about, not a dataset you happen to have. A dataset without a driving question tends to produce exploratory noise — a dozen half-finished charts and no real conclusion — while a specific question naturally tells you which parts of the data matter and which you can safely ignore. If you're choosing between one of the course-provided datasets and your own data from work, pick your own only if you can share the resulting analysis freely; confidentiality concerns discovered midway through a project are a common source of last-minute scope changes.

A good capstone question is specific enough to be answerable with the data you have, matters to a plausible real audience (even a hypothetical one), and has a clear point at which you'd know you'd answered it. "How has revenue changed" is too broad to resolve into a finding. "Which customer segment's declining retention is driving this quarter's revenue slowdown, and what does their usage pattern look like right before they churn" is specific enough to actually investigate and answer.

Scoping the Project So It's Finishable

The most common way capstone projects fail isn't a lack of skill — it's scope. An ambitious question that would require weeks of clean, complete data, five different dashboards, and a dozen segments explored in depth is not a realistic scope for a project with a deadline. Deliberately narrow the question until you can picture the finished deliverable: this specific dashboard, answering this specific question, supporting this specific recommendation.

A useful scoping exercise: write down what you'd cut first, second, and third if you ran short on time. If you can't answer that quickly, the project probably isn't scoped tightly enough yet. Better to fully answer a narrow question than to partially explore a broad one and end up with no clean conclusion at all.

Setting a Realistic Timeline

Break the project into the same phases as the course itself — cleaning, exploration, dashboard building, and narrative — and estimate time for each honestly, based on how long each phase actually took you in earlier modules. Cleaning and exploration together will very likely take longer than dashboard building and narrative combined; treating them as a quick preliminary step before "the real work" of building the dashboard is one of the most common planning mistakes.

Build in a review pass near the end, separate from the work itself — time to look at your dashboard and narrative with fresh eyes, check the numbers still make sense together, and catch anything that reads clearly to you but wouldn't to someone unfamiliar with the project. Leaving zero buffer between "finished building" and "submitted" means the first person to spot a problem is whoever's grading it.

Defining What "Done" Looks Like

Before you start building, write a short, concrete description of what a finished capstone looks like: the specific question answered, the dashboard views included, the narrative's core recommendation. This becomes your finish line — without it, it's easy to keep polishing indefinitely or to stop short of a real conclusion because there was never a clear target to reach.

"Done" doesn't mean exhaustive. It means the question you set out to answer has a clear, supported answer, the dashboard lets someone explore that answer for themselves, and the narrative explains what to do about it. Anything beyond that is a nice-to-have, not a requirement.

Practical Review Checklist

Before moving on, confirm that you can:

• Explain why starting from a question is better than starting from a dataset
• Write a capstone question that is specific, answerable, and has a clear resolution point
• Describe the "what would I cut first" scoping exercise
• Estimate time across cleaning, exploration, dashboard building, and narrative honestly
• Write a concrete definition of "done" for your own capstone before starting

Conclusion

A capstone succeeds or fails on planning more than on any single technical skill. Choose a specific question, scope it to fit your timeline, and define what finished looks like before you touch the data.