Examples
Bootstrap CUSUM applied to real data across healthcare, policy, and industrial settings. Each example asks the same honest question: did a structural change point appear — and if so, when and why? Several show what 15 years of flat line looks like, and what that means for the interventions applied.
Change point present: the date Bootstrap CUSUM returns, what changed in the system at that date, and whether the change was structural or a special cause event.
No change point: what that flat line means — stable at the wrong level — and which Joiner level the interventions applied were operating at.
Balancing measures: whether an apparent improvement in one metric was accompanied by deterioration in another.
Pre-committed predictions: where a prediction was made before the data arrived, the example tests it honestly.
NHS A&E — Why nothing has worked
184 monthly observations. Four structural stages of decline. Not one policy intervention visible as an improvement change point. The definitive case study in Level 1 and Level 2 responses applied to a Level 3 discharge constraint.
Corridor care 2029 — Prospective analysis
The government’s four initiatives assessed before the data arrives. Which will produce a Bootstrap CUSUM change point by 2029 — and which won’t, and why. The pre-committed prediction is made in 2026 so the data can test it honestly.
Corridor care — Bright Spots
Watford General eliminated corridor care. Bootstrap CUSUM, Joiner’s levels, and Deming’s “by what method?” applied to the candidate Bright Spot — and the honest questions the public picture does not yet answer.
Never events — Wrong route
Ten years of a never event that never stopped. Disaggregation reveals 16 of 20 events in 2023–24 were oral-to-IV — not the neuraxial pathway the NRFit mandate addressed. The dominant mechanism trap in practice.
Anticoagulation safety
A detectable change point in DOAC adverse reactions at 2016 corresponds to the rivaroxaban controversy. More patients, fewer bleeds — and the balancing measures that confirm whether improvement is real or redistributed.
Dementia — The 66% target
What Bootstrap CUSUM shows when a government target drives a metric upward — and whether the change point corresponds to genuine improvement in care or to diagnostic pressure on clinicians.
GP appointments
Are GP appointment numbers genuinely falling, or is the denominator (registered patients) growing faster than supply? Bootstrap CUSUM on the rate, not the count, gives the honest answer.
Sepsis Six
Does the public data show whether Sepsis Six worked? The national aggregate shows no change point — and why that null result is not evidence it failed, but evidence the aggregate is the wrong metric.
ULEZ — Air quality
A change point in London air quality data attributed to the ULEZ expansion — but the change point date precedes the policy by several weeks. Attribution errors in policy evaluation, and how Bootstrap CUSUM dates them precisely.
The Grid fixed itself
A structural improvement change point in UK grid carbon intensity — before any single policy intervention could explain it. What happens when a system’s incentive structure changes and the market responds faster than regulation.
UK GDP analysis
Bootstrap CUSUM on decades of UK GDP data finds the structural stages of growth and contraction — and dates them precisely enough to test political claims about which government caused which economic shift.
Hydrogen plant CUSUM
Bootstrap CUSUM applied to a continuous industrial process — hydrogen plant efficiency over time. What a genuine process improvement change point looks like in production data, and how to distinguish it from seasonal and feedstock variation.
Every example on this page was produced with the same free, browser-based tool. Upload a CSV of monthly or weekly data — no account required, no data uploaded to any server.