Outcome Measurement
Establishing a baseline before a change and measuring the intended effect afterwards, so that investment decisions are informed by evidence rather than by narrative.
Most technical investment is never measured. The work is delivered, the team moves on, and whether it achieved anything is settled by whoever tells the story most convincingly.
The requirements are unglamorous and rarely met:
A baseline captured before starting. This is the step that is skipped, and without it every subsequent claim is unfalsifiable. It costs a day and it is the difference between evidence and assertion.
A stated hypothesis with a target and a date — "reducing checkout latency from 800 ms to 300 ms will raise completion rate by at least two percentage points, measured over four weeks after rollout".
A measurement plan agreed in advance, including what would count as failure. Agreeing the success criterion afterwards guarantees success.
Controls for confounders — seasonality, concurrent changes, and traffic mix. A/B testing where possible; comparison against the same period last year where not.
Publication regardless of result. A programme that only reports successes teaches everyone to discount its numbers.
The compounding benefit: an organisation that measures outcomes builds a body of evidence about what kinds of technical investment actually pay, which makes the next set of decisions better. One that does not repeats the same arguments indefinitely.