This content originally appeared on DEV Community and was authored by Jon Lauridsen
I see and hear of so many teams stuck on estimating tasks, but estimates almost always have a hopelessly poor signal-to-noise ratio. There are better ways of working!
Here's the thing: If you can demonstrate your estimations predictably and reliably map to clear buckets of cycle times then, okay, your estimates provide a signal (but there's still an easier way of working). But I have effectively never seen that happen. Inevitably estimates do not reliably map to actual time spent, and actually almost no-one even does that analysis so they just go on blind faith that estimates provide a signal. Most of us got into estimates just because it was part of the scrum ceremonies that snuck in during the 2000s, and it's beyond time we question it!
Let's be clear: If estimates don't map to actual time then they're not worth doing. End of story.
So what to do about it?
Here's my pitch: Use past performance to forecast likely futures instead. It's very easy: Use the start- and end-dates of past stories to forecast future outcomes (e.g. Troy Magennis has a free tool for this or ActionableAgile for a paid product). That's a probabilistic forecasting model that shows likely end-dates for the next set of stories that make up a feature. Because the team is probably slicing stories the same way now as they did the last couple weeks this provides a great indicator of probable end-dates. And every story that completes makes that forecast more precise, like a Google Maps journey that ticks down the ETA: The closer you get the more precise it becomes.
I assert this is already as or more useful than estimations, because it's at least as accurate and it's just plain mechanically easier: No more planning poker, no more arbitrary fibonacci numbers, just focus on the actual work and let the model give you estimates. It's a simple tool that really connects the day-to-day work with a speedometer of general progress.
It's also just the beginning of an incredible journey where the team is now free to experiment with ways of squeezing out variability, to increase the accuracy and stability of the forecast:
- Right-size stories so they generally take the same amount of time (anchored around each story delivering customer value)
- Eliminate bugs/rework, because they wreck havoc on predictability
- Identify queues and other sources of delays (e.g. via value stream mapping), and challenge each step to minimize waste (that's Lean principle #2)
There are hundreds of exciting experiments to run for your team to find the optimal way to work. Simplify your work-process today and drop those estimates.
This content originally appeared on DEV Community and was authored by Jon Lauridsen
Jon Lauridsen | Sciencx (2022-02-08T20:13:30+00:00) Estimates don’t work, but there’s a simpler way. Retrieved from https://www.scien.cx/2022/02/08/estimates-dont-work-but-theres-a-simpler-way/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.