When creating a forecast first ask yourself whether you are forecasting One Thing or Multiple Things. It’s not always clear which of these situations you are in but the approach you take to creating the forecast will differ significantly. This post will help you to figure out which approach to take.
When forecasting Multiple Things the best approach to take is to use a Monte Carlo forecast. This approach creates a statistically significant probability distribution of delivery on various dates based on thousands of simulations using historical throughput data. To turn this into a forecast you will choose a confidence level (I recommend starting with 85%) and find either the date by which 85% of those simulations were finished with all of the work or the number of items finished at least 85% of the time by that date.
The Multiple Things approach is often used for planning purposes, for example:
- How many things can we get into the next release?
- How many things should we plan for the next sprint?
- How long until all of the cards for this project are done?
Note that all of these questions are focused on forecasting a group of items over some time period.
When forecasting One Thing the best approach to take is to use a Cycle Time Scatterplot. This approach uses a percentile to determine by what date the item you are forecasting will be done (once again, we recommend 85%). The percentile will be focused on how many days (or weeks, hours, months, etc.) it has taken to get work items done in the past. The forecast is as simple as saying that “85% of the time a work item completes in X days or less”.
This One Thing approach is used when you want to forecast a single thing. For example, a single story, a project, an epic, etc. Note that projects and epics are single things made up of multiple things. The approach you take can depend on how well understood the items that make up the single thing are. If you have a good understanding of how many items make up your project then you can theoretically use either approach and should come up with a similar answer. The Multiple Things approach described above will often be less susceptible to variability in the size of items so when possible I would recommend using that approach.
If you want to learn more about these approaches and when to use each one we offer a 2 day training course on Applying Metrics for Predictability! Check out upcoming courses at our training page
Originally posted at JEMSoftware.co