A New Copy!

We are back!  Today we begin Part 3 of Daniel S. Vacanti’s Actionable Agile Metrics for Predictability: An Introduction (buy a copy today) with Chapter 10 which  is titled, Introduction to Cycle Time Scatterplots. Scatterplots take us beyond the analysis of average cycle time (or even approximate average cycle time).  Scatterplots provide a visual representation of the data so we can begin to use the data to predict the future.

Vacanti opens Chapter 10 by suggesting that we can do better than analysis based on averages.  Scatterplots are the tool he suggests.  Scatterplots are not the same as statistical process control diagrams (SPC).  Action Agile Metrics does not address SPC and Vacanti makes a substantial argument that SPC is not useful for software development and enhancement work.

Scatter plots are a relatively simple graphing technique that plots the progression of time on the x-axis.  For example, in cycle time scatterplots for software development and enhancements the x-axis usually is denominated in calendar days or weeks. The y-axis represents cycle time in days.  Scatter plots are useful for analyzing any number of metrics; however, this chapter only focuses on cycle time.

Once constructed Vacanti adds percentiles lines on the chart.  When drawing the 50 percentile line half of the observations would be above the line and half below. Knowing the 50% line allows to you predict how long an average item will take to complete.  Drawing the line at 80 or 85% allows for an even more confident prediction that any piece of work will complete in less the number of days indicated by the line you draw.

For example:

Using the following data

 Date Cycle Time Story 1 1/1/2018 1 Story 2 1/2/2018 4 Story 3 1/3/2018 7 Story 4 1/4/2018 3 Story 5 1/5/2018 7 Story 6 1/6/2018 9 Story 7 1/7/2018 0 Story 8 1/8/2018 11 Story 9 1/9/2018 40 Story 10 1/10/2018 1 Story 11 1/11/2018 6 Story 12 1/12/2018 8 Story 13 1/13/2018 1 Story 14 1/14/2018 11 Story 15 1/15/2018 15 Story 16 1/16/2018 6 Story 17 1/17/2018 5 Story 18 1/18/2018 11 Story 19 1/19/2018 12 Story 20 1/20/2018 13

Would yield the following cycle time scatterplot.

There is a natural tendency to establish the mean and calculate the standard deviation of the observations. From there it is a simple step to create upper and lower control limits (SPC).  The problem is that the data do not follow a normal distribution.  Normal distributions assume that observations are equally distributed around the mean (central limit theorem). The assumption that observations follow a normal distribution is really extremely rare in software development. (Note: Tools like EXCEL make assuming normality and drawing all sorts of trendline very easy).  Vacanti uses scatter plots (XY Plots in Excel) because they don’t make the assumption of a normal distribution, rather they make no assumptions as to the distribution of the data. Another of the strengths that Vacanti points out is that the percentile lines are easily calculated. Another strength is that outliers do not have a significant statistical impact on scatter plots.  The discussion of outliers assumes that there is a standard distribution of observations leading us back to a discussion of why are assuming a normal distribution when we don’t have to!

Returning to our example:

We could add both a 50% and 85% line to our scatterplot

The data shows that 50% of the stories complete in under 7 days and 50% above.  The team could indicate that a story could be done in 7 days — a 50/50 possibility.  Or if they needed a higher level of certainty (somethings who you are talking to requires something more than 50/50 confidence) the 85% line tells us that 85% of the stories complete in 13 days or less.  The use of a cycle time scatterplot is straightforward.

Chapter 10a (next) discusses Cycle Time Histograms.

FYI – I heard from a colleague this week that said the re-read influenced him to buy and read the book. It is heartening to hear that our hard work is helping our readers to change how they work!

Previous Installments

Introduction and Game Plan

Week 2: Flow, Flow Metrics, and Predictability

Week 3: The Basics of Flow Metrics

Week 4: An Introduction to Little’s Law

Week 5: Introduction to CFDs

Week 6: Workflow Metrics and CFDs

Week 7: Flow Metrics and CFSs

Week 8: Conservation of Flow, Part I

Week 9: Conservation of Flow, Part II

Week 10: Flow Debt

Support the author (and the blog), buy a copy of Actionable Agile Metrics for Predictability: An Introduction by Daniel S. Vacanti

Book

Kindle