Reading the Structured Prediction Gadget
Structured Predictions generates a Predictive Cumulative Flow Diagram and provides specific metrics for each team.
The following sections explain each element of the Structured Predictions gadget.
1. Historical Cumulative Flow Diagram
This section of the graph shows the historical CFD of past workflow transitions and ticket inventory to the current day. The points on the graph show the date at which issues were in a specific workflow state:
- Blue graph area indicates the work that is still todo (in the backlog)
- Yellow indicates the work that is in progress
- Green indicates the work that is done
By plotting all of the issue transitions for the project, the graph shows a historical representation of what the project looked like at a given date in history. By assessing the slope of each workflow state, the rate at which work is being created and completed for a project can be easily seen.
2. Predicted Stories
The steep cliff in the "todo" area of the graph at the line labeled "Today" is the result of the Structured Predictions algorithm making a predictions as to the number of additional stories each team contributing to the delivery will create at some future point in the project. The predicted stories are added to the known project scope (the scope that is defined in Jira) at the current date.
3. Estimated Completion Dates
Each team's completion date is predicted by running the team's performance model (based on the training structure assigned in the configuration) against the remaining scope defined in Jira and the additional scope predicted by the algorithm for that team. Hovering over the team label pin shows some metrics for each team.
4. Today and Target Lines
The Today line, quite simply, is drawn at the current date. It represents the transition point of the graph from a traditional historical cumulative flow diagram to a predictive cumulative flow diagram. The portion of the graph to the right of the Today line is drawn from the simulated future for each team working on the delivery.
The Target date is provided to the gadget as part of the configuration. Teams that are predicted to finish after the target date are flagged in red in the metrics table to the right of the graph.
5. Team Metrics
The team metrics breaks down the key stats by team. There are six columns of data for each team.
- Team - is the name of the team as determined by the "Team Attribute" setting in the Predictive CFD gadget configuration screen.
- Throughput - is the average throughput for the team measured in items of scope per week. What issues represent scope is configurable. The default is stories and tasks. For example, “1.0” indicates that a team generally gets one story per week done. A value of “3.30” would indicate that every week a team gets a little over three items of scope done.
- Remaining - indicates how much scope remains out of the total scope, actual plus simulated, for each team.
- Predicted - reveals how much of that scope, column 3, is from the predictive algorithm. Predicted scope is scope that is not yet represented in Jira but is predicted to be created by the team at some point in the future. For example, a value of 48 in column 3 and 35 in column 4 tells you that while there are only 13 remaining issues defined in Jira (48 - 35 = 13), there are likely to be an additional 35 created in the future for a total of 48 remaining items of scope for the team.
- Trending - shows the predicted completion date for the team. This is when the team will likely complete all known and predicted scope.
- Delta - shows the standard deviation for each team's throughout. This is the standard deviation for the underlying data used to determine the throughout. Teams with low deviation have a more consistent focus on the project work, and correspondingly their predictions will be more trustworthy. Teams with higher deviations may require more attention to make sure they are focused when they need to be. They may be juggling multiple priority streams, or they may have a less predictable or mature process. Whatever the cause of the deviations, attention to the team is advisable.
Using Predictions to Set Target Dates
The Structured Predictions graph is a great way to start a conversation with the teams on a project about what dates they think they can hit, given the particulars of the project and the circumstances within the organization:
- Talk about the output from Predictions with the various teams.
- Let them know what project data you are using for the predictions.
- Ask them if the dates look good, are too aggressive, or if they think they can beat them. Ask them why they believe what they do.
- Record the key points from the dialog, along with the agreed date. These key points will provide context once the project gets further along and decisions need to be made as the project responds to change.
Including Predictions in Status Reports
Taking a screen shot of Structured Predictions is a quick way to get completion trends into a status report. A weekly report with a screen shot will reveal trends to stakeholders as well as validate the impact of decisions that were made and expected to impact project performance.