Today, I want to dive into something that might seem a bit technical at first but is super useful for anyone looking to make reliable delivery forecasts: percentiles.
Trust me, understanding this concept will save you tons of time and help you make more informed decisions.
What Are Percentiles?
Percentiles are statistical measures that show where a particular value stands within a dataset. Imagine slicing your data into 100 equal parts, each part representing 1% of your data. That’s basically what percentiles do.
For instance, the 50th percentile (aka the median) is the middle value of your data – half of your data falls below it. The 98th percentile? That’s the value below which 98% of your data falls.
Let’s break this down with an example. Say we have a dataset of completed items per sprint over several sprints:
Here is a very simple, step-by-step approach to calculating your percentiles:
- Sort the Data: 22, 20, 18, 16, 14, 14, 12, 10, 8, 5
- 50th Percentile (Median): The middle value is between the 5th and 6th items, so we’ll pick the 6th one: 14 work items
- 70th Percentile: 70% of 10 items is 7. The 7th item in the sorted list is 12 work items.
- 85th Percentile: 85% of 10 items is 8.5. The 9th item in the sorted list is 8 work items.
- 95th Percentile: 95% of 10 items is 9.5. The 10th item in the sorted list is 5 work items.
So, what does this tell us?
- 50th Percentile (14 items): Half of the sprints completed 14 or more items.
- 70th Percentile (12 items): 70% of the sprints completed 12 or more items.
- 85th Percentile (8 items): 85% of the sprints completed 8 or more items.
- 95th Percentile (5 items): 95% of the sprints completed 5 or more items.
Why Do We Need Percentiles?
Percentiles help us understand the distribution and probability of certain outcomes in a dataset. They’re incredibly useful for setting reliable delivery commitments.
Using the percentiles, you can gauge your team’s capacity:
- Minimum Commitment: Confidently commit to completing at least 5 items (95th percentile) in the next sprint
- Lower Confidence: There’s an 85% chance of completing more than 8 items and a 70% chance of completing at least 12 items in the next sprint
- Risky Goal: If you’re aiming to get to a higher number of items, the chance of meeting that goal goes down. There’s a 50% chance of completing at least 14 items, though this comes with higher risk (it will either happen or not!)
By understanding and applying percentiles on your historical performance data, you can better predict your team’s performance and set realistic goals for your next sprint.
Where Will You Find Them and How to Read Them
On Nave, you can find percentiles in various charts depending on the type of forecast you need to make.
- Cycle Time Scatterplot: Look at the horizontal dotted lines representing different percentiles. For instance, the 50th percentile line might show an 8-day completion time, meaning half the tasks are completed in 8 days or less.
- Throughput Histogram: Vertical percentile lines indicate the probability of achieving certain throughput levels in the future. A 50% line might show 3 items/sprint, meaning there’s a 50% chance the team will complete 3 items or more in their next sprint.
- Monte Carlo Simulation: Dotted vertical lines show the probability of completing your project. The 85th percentile line might intersect the x-axis at 23 Sep, meaning there’s an 85% chance of completing the project before 23 Sep.
By understanding and using percentiles, you can better manage expectations and plan more effectively, ensuring that your team can meet its commitments and deliver value consistently.
They give you a clear picture of your past performance, helping you make accurate predictions about your future work and set realistic goals.
Alright, my friend, that’s a wrap! If you found this useful, please share it with someone who might benefit from it. Thanks for tuning in, and I’ll see you next week same time and place for more managerial insights. Have a wonderful day!
Hey, the example, based on data order from highest to lowest (table), is a little bit confusing.
“The 98th percentile? That’s the value below which 98% of your data falls.”
Let’s take this one:
“85th Percentile: 85% of 10 items is 8.5. The 9th item in the sorted list is 8 work items.” – based on the table: 8 (work items) is one before last, it should be in 15th percentile, not 85th.
If we change the data order from the lowest to the highest – it makes sense then, there are 85% of other work items below 8 work items.
Wju, we are forecasting how much work we’ll be able to finish in our next iteration. The calculations are optimized to give you that answer. You can turn the percentiles around, but then you’re giving an answer to a different question.