They are a proxy for time and nothing else. An intentionally imprecise and noisy proxy, and worth keeping that way. Their purpose is to be a notional size that

They are a proxy for time and nothing else. An intentionally imprecise and noisy proxy, and worth keeping that way. Their purpose is to be a notional size that can be compared with the rough size of a “sprint”. A sprint might be denominated as points or as time, but since they’re typically time, the point is forced to be roughly time too. Otherwise the unit of measurement doesn’t make sense: what else would you sum up and compare to a duration except durations? Also: effort-time, not elapsed time. You can’t usefully predict blockers, and when you’re blocked you can usually work on something else, maintaining effort on overall progress. “Effort” and “complexity” only translate into “time spent doing something useful” and “it’s not all figured out, need time to figure out and probably will take more work than something known simple. If a point was “effort”, you can work half as hard and take twice as long, or twice as hard and take half as long. Implying a wide range of throttle on your participants. Estimating is always done with a particular throttle setting in mind. “Complexity” on its own doesn’t make sense. Entering the contents of the Harry Potter series into a Markdown file is pretty simple therefore 1 point but may take a sprint. I do believe in the value of estimating in points but also in getting granular to identify all the time drivers. Consistency there and constant review of accomplished work to estimated work in a story makes future estimating much more reliable as well as makes progress in a sprint much more transparent.