Estimates and edges

With the cheer feature merged, Dana asks the next question at planning, and it is the question engineers dread most in exactly these words: how long would projected finish times take?

What Dana is asking for is an estimateConcept · lights on your mapestimateAn informed guess at how long work will take, famously unreliable because most tasks contain work nobody has discovered yet. Good teams estimate in ranges and re-estimate as they learn. It is not a promise, though it will be remembered as one., and the room goes quiet because everyone knows the two truths of estimates at once. They are necessary, and they are guesses. Most software tasks contain work nobody has found yet, which is why Traversal estimates in ranges, and why the phrase it will be remembered as a promise gets said at least once per planning.

The planning words that mean things

Around the estimate, the table talks in a compact vocabulary worth having. A blockerConcept · lights on your mapblockerAnything currently preventing work from proceeding: a missing decision, a broken environment, another team’s unfinished API. The highest-priority sentence in any standup is “I am blocked.” is whatever is stopping work right now. A riskConcept · lights on your mapriskA possible future problem worth naming now. Naming risks is how teams buy insurance while it is cheap, whatever it sounds like. is a problem that has not happened yet but deserves a name early, and the model needs historical data we have never touched is the one Finn names out loud. A milestoneConcept · lights on your mapmilestoneA named checkpoint in a larger plan: visible progress, and the unit executives actually track. is a checkpoint worth an email: live at Maple Grove is one, live at Harborview is the one.

Sizing happens in story pointsRecognition — just know it existsstory pointsAbstract effort units many teams estimate in instead of hours, sizing work relative to other work. The source of much ritual; recognise and nod. at Traversal, abstract units that dodge false precision, a ritual you should recognise and not worship.

Buying an answer with two days

Nobody can estimate the projected-finish feature honestly, and the team does the honest thing about that instead: it buys knowledge. A spikeRecognition — just know it existsspikeA time-boxed investigation ticket: spend two days finding out how hard the thing is, producing knowledge instead of features. How a team buys down a risk before estimating for real. goes on the board, two days, Finn’s name on it, one question inside it: can we predict a finish time from splits and ten years of race history, at all?

Two days later Finn comes back with a maybe worth its own act, a folder of charts, and one sentence for Dana: the data exists, and it lives somewhere we have not visited yet. That somewhere is four scenes away, and it is where Case lives too.