Artifact: A #NoEstimates Alternative for a Chronically-Wrong Backlog

ex-35 · exercises co-19 · slicing smaller and counting throughput, instead of estimating.

Problem: the last six sprints' estimates have missed actual delivery by 40-90% every time, regardless of who estimated or how carefully.

#NoEstimates-style proposal (Woody Zuill's critique -- questions estimate VALUE, not a blanket ban on ever sizing work):

1. Slice every backlog item until it is small enough to ship in 1-2 days (no story gets an
   estimate -- it either fits the size band or it gets sliced again until it does).
2. Track ACTUAL throughput: how many sliced items shipped last week, the week before, etc.
3. Forecast a release date from throughput (e.g. "we ship ~6 items/week, 40 items left, so ~7
   weeks"), the same way a manufacturing line forecasts from a measured cycle time -- not from
   summed estimates.

Verify: the proposal removes the per-item estimate step entirely (step 1 explicitly has "no story gets an estimate") while still producing a forecast (step 3, from measured throughput) -- satisfying co-19's rule that the alternative keeps the ability to forecast without keeping the broken estimate step.

Key takeaway: forecasting from measured throughput of small, uniformly-sized items sidesteps the question "how long will THIS take" entirely, answering "how long will everything left take" instead -- which is usually the question that actually mattered.

Why It Matters: #NoEstimates is not "never plan" -- it is a specific bet that, for teams whose estimates are chronically and unpredictably wrong, a measured historical rate forecasts the future better than another round of guessing does, especially once items are small and uniform enough that individual guess-error mostly cancels out across many of them.