Free stuff
Metrics in general
- Different levels of the organization should be looking at very different metrics. Understand who should be looking at what.
- Goodhart’s Law says that “when a measure becomes a target, it ceases to be a good measure.” Many people struggle with this one.
Measuring flow metrics
- Understanding basic flow metrics of Cycletime, Throughput, Work In Progress (WIP), and Work Item Age
- Understanding flow efficiency, a very powerful metric that’s rarely used because of the difficulty in measuring it.
- Visualizing your data: Cumulative flow diagrams and visualizing WIP limits
- In order to measure any of the basic flow metrics, we need a clear definition of the start and end points
- Common problems with flow metrics
- We often start collecting data before we even know what we want to do with it.
- We often collect the wrong data
- We often misinterpret what’s there
- We focus on keeping people busy instead of optimizing for any of effectiveness, efficiency, or predictability.
- We don’t slice the work well and this skews the data. See how to slice stories or slice epics
Gaming metrics
- If people know that a thing is being measured, they’ll change their behaviour to ensure that measurement makes them look good. Sometimes the result of that behaviour change is so wildly different from the desired outcome, that it causes other problems. That’s a perverse incentive.
Metrics from Jira
You should really use an existing tool to extract metrics from Jira, like my own JiraMetrics, but if you really want to do it yourself, I’ve documented the API’s I use.
When will we be done?
When trying to answer the question “when will we be done?”, we often try to estimate how big the individual work items are, and this is complete waste. What we want to do instead, is to create a probabilistic forecast. In a nutshell, an estimate is a made up number based on how we feel about complexity, and a probabilistic forecast is a far more accurate mathematical calculation based on historical data. It’s time for estimates to go away.
Part of having an accurate forecast is ensuring that the data itself is inherantly predictable and there are specific things we can do, to improve that predictability.
Once we have that forecast, we need to understand the cognitive biases that make us trust it less, even though it’s more accurate. If we want to make better decisions, we need to understand our own biases.