Delivery and Estimation

The cone of uncertainty

Barry Boehm, Software Engineering Economics, 1981. Amazon

At the start of a project, estimates are off by up to 4x in either direction, and the uncertainty narrows as decisions get made. Boehm drew the funnel; McConnell named it. A single number at the start of a project is a fiction for structural reasons, not because the estimator is bad.

Used in Why I ask for ranges.

Software Estimation

Steve McConnell, Software Estimation: Demystifying the Black Art, 2006. Amazon

The cone narrows when the project resolves unknowns, not as time passes. Estimates and commitments are different things made at different times, and the most common estimation error is presenting the optimistic end of a range as the estimate.

Used in Why I ask for ranges, Build a schedule from estimate ranges.

Little's law

John D. C. Little, "A proof for the queuing formula L = λW," Operations Research 9(3), 1961. doi:10.1287/opre.9.3.383

Cycle time equals work in progress divided by throughput. Adding work in progress without adding throughput lengthens cycle time mechanically, before any context-switching cost. The argument for one initiative per engineer.

Used in Why I ask for ranges, Keep delivery predictable, Diagnose what's broken in an org.

Theory of constraints

Eliyahu Goldratt and Jeff Cox, The Goal, 1984. Amazon

Every system has one constraint that limits throughput, and improving anything else doesn't help. On an engineering team it's usually a person (the senior engineer who reviews everything) or a step (QA, release, security review). Starting more work queues it behind the constraint.

Used in Why I ask for ranges, Keep delivery predictable, Diagnose what's broken in an org.

The four key metrics

Nicole Forsgren, Jez Humble, and Gene Kim, Accelerate: The Science of Lean Software and DevOps, 2018. Amazon

DORA, State of DevOps reports, annual. dora.dev

Deployment frequency, lead time for changes, change failure rate, and time to restore predict delivery performance and resist gaming better than velocity. The research also found that Westrum's generative culture predicts delivery performance.

Used in Keep delivery predictable, Diagnose what's broken in an org.