Spectral chart
The Spectral Chart shows the time distribution of tasks and allows you to predict how long it will take you to complete a new task.
This chart is also known as the Lead Time Distribution Chart (LTDC).
How to read and understand the Spectral chart
On the X-axis you can see days.
On the Y axis – the number of cards that were completed in the corresponding number of days.
For example, the graph above shows that 1 card was done in 1 day, 1 card in 2 days, 4 cards in 3 days, etc.
Have you noticed the two vertical lines on the graph?
The blue line shows the average time for all tasks.
The red line marks the percentile line (in this case the graph has 85 percentile). It marks where 85% of the observed items (cards) are in the interval.
For example, the graph above shows that the average task time is 6.35 days. But there is an 85% probability that a new task will be completed in 10.5 days.
If you click on a bar on the graph, a list of the cards included will open.
You can use this graph to predict when the tasks will be completed:
Filter the cards to be accounted for on the graph. To get more accurate information, include tasks of the same category in the report (tasks are performed by the same team, same type of tasks, tasks are on a track with the same priority or urgency).
If you include ASAP tasks or, on the contrary, tasks that are delayed for some reason, the schedule will not be accurate.
Build your forecast based on percentile, not just based on average time.
If you see "long tails" on the graph, i.e. tasks that have been executed significantly longer than the others - you should pay attention to them.
When predicting the execution time of a new task, correlate which tasks are more similar in their essence? If it is to the left of the red line - then you can consider that the new task is most likely to be completed in the time indicated by the percentile. If it's to the right of the red line, it's more likely to take longer to complete.
If you have a normalized process, and you have collected information over a long period, you can predict times based on this information.
Explaining with an example
The graph above shows that a total of 18 cards passed through the system during the selected period.
The maximum solution time for one of the tasks was 21 days, and most of the tasks - 14 cards - are completed in 1 day.
Consequently, we can say that a new task can be processed from 1 to 21 days. Pretty vague and imprecise, right?
If you want a more accurate prediction, it makes sense to look at the red percentile line. It tells us that there is an 85% probability that we will be within 9.5 days.
We can also look at our new task and compare it with the tasks in this graph. Which tasks does it look more like? The ones that were completed in 1 day? Or the one that has been in the works for 21 days?
We should also pay attention to the task that has been running for 21 days - find out the reason it took so long to complete.
It turns out that with a spectral chart we can give predictions of task completion with a certain accuracy based on statistical data.