Every project generates numbers: task durations, sprint velocity, ticket cycle times, budget actuals, defect counts. Most of us look at them through a dashboard that has already boiled them down to a single figure.
Descriptive statistics are the tools that do that boiling down. Knowing how they work tells you when the summary on your screen can be trusted and when it's hiding something.
What are Descriptive Statistics?
Descriptive statistics summarize a dataset: where its center is, how spread out it is, and what shape it takes. They describe the data you have. They don't predict anything or test a hypothesis, which makes them the natural first step in any analysis. (If you've read Exploratory Data Analysis is an Art, this is the toolkit that post starts with.)
The Tools, with Project Examples
Central tendency: where's the middle?
The mean is the average. The median is the middle value when the data is sorted. The mode is the most frequent value. On a project, these answer questions like "how long does a typical ticket take?" But the three can give very different answers, as the example below shows.
Dispersion: how much does it vary?
The range is the gap between the smallest and largest values. The standard deviation measures how far values typically sit from the mean. Two teams can have the same average velocity, but the team whose velocity swings wildly from sprint to sprint is much harder to plan around.
Percentiles and quartiles: what share falls below a given value?
The median is the 50th percentile. Project people often find the 85th or 90th percentile more useful, because it lets you say "85% of our tickets finish within X days." That's a far better basis for a commitment than an average.
Skewness: which way does it lean?
Skewness describes asymmetry. Project data is usually skewed to the right: most tasks finish in a reasonable time, and a few drag on for a very long time. That shape is the reason the mean and median so often disagree.
Kurtosis: how heavy are the tails?
Kurtosis describes how peaked or flat a distribution is, and how prone it is to extreme values. You'll rarely calculate it by hand, but it's worth knowing the term for when a data team mentions it.
A Worked example: Nine Tickets
Here are the cycle times, in days, for nine tickets: 2, 3, 3, 4, 4, 5, 5, 6, and 41.
- Mean: about 8.1 days
- Median: 4 days
- Range: 39 days
- Standard deviation: roughly 12 days
A dashboard showing "average cycle time: 8 days" would suggest a typical ticket takes over a week. In reality, eight of the nine tickets finished in 6 days or less. One ticket, probably blocked or forgotten, pulled the average up. The median tells you what a typical ticket looks like. The standard deviation, larger than the mean itself, tells you something unusual is in the data.
Neither number is wrong. They answer different questions, and you need both to understand what's going on. The 41-day ticket is also worth investigating on its own, because it's probably a process problem rather than a statistical curiosity.
Why this Matters for Project Work
- Summarizing: A project with thousands of data points becomes a handful of numbers you can discuss in a status meeting.
- Comparing: Set this sprint's median and spread against the last five, or compare one team to another, and real trends become visible.
- Checking data quality: Impossible values, extreme outliers, and suspicious clusters show up quickly in summary statistics, before they distort a report.
- Making commitments: Percentiles turn "it usually takes about a week" into "85% of the time it takes six days or less."
- Generating questions: A surprising median or a huge spread rarely answers anything. It tells you where to look next.
Best practices
- Know your data first. Understand where the numbers came from and what they measure. A "cycle time" that includes weekends means something different from one that doesn't.
- Never report a mean alone. Pair it with the median and a measure of spread. If they disagree, that disagreement is the finding.
- Draw the picture. A histogram or box plot shows in seconds what a table of statistics takes minutes to convey.
- Read numbers in context. A standard deviation of 12 days is alarming for ticket cycle times and unremarkable for a portfolio of multi-year programs.
- Remember what they can't do. Descriptive statistics summarize what happened. They don't explain why, and they can't establish cause and effect.
- Write down your assumptions. Note how you handled outliers and missing values, so the next person can reproduce and trust your numbers.
In Conclusion
Descriptive statistics are the first questions you ask of any dataset. Where's the center? How much does it vary? What shape is it? Answering them takes a few minutes and prevents a lot of bad decisions built on a single misleading average. The next time a dashboard hands you one number, ask what the median, the spread, and the outliers look like behind it.
Good Luck
