Is performance improving, getting worse, or simply varying from one result to the next?
A practical resource to help people plot data, recognise possible changes, ask better questions and learn from improvement work. It is not a substitute for professional judgement or specialist analysis when important decisions depend on the result.
The improvement guide explores right problem, right intervention and right measures. The separate StepChange Analyzer below helps make patterns visible with run charts and step-change analysis. Choose a method that suits your data: a chart can suggest a change, but cannot explain its cause or prove an intervention worked.
Intended use: the StepChange Analyzer supports process improvement and the retrospective analysis of historical data, by making changes in business, operational and service-level data visible. It is not a medical device: it is not designed, validated or intended for diagnosis, treatment or any other decision about an individual patient. Read the full intended-use statement →
Open the StepChange Analyzer ↓Explore the improvement guide →
Plot performance over time. Investigate possible changes. Evaluate improvement carefully.
📁 Ready! Click or drop a CSV file to begin.
Candidate Stage Summary — stage-specific confidence is not shown
Look at the dots and the change dates first. These checks can help you decide what needs a closer look; they cannot prove a change or its cause.
How to start checking your results →
Look for a possible relationship between neighbouring measurements. A trend or seasonal pattern can also affect the result.
Repeat the whole analysis 10 times using the same data and settings, with new random shuffles each time.
The government aims to eradicate corridor care by end of Parliament. Four initiatives — Joiner level analysis applied in advance. The pre-committed prediction: Bootstrap CUSUM will not find a structural change point unless the discharge constraint is addressed at system level.
England exceeded the dementia diagnosis target from 2017–2020, then COVID produced the largest single-year fall in the series. By March 2026 the rate had recovered to 66.3%, and NHS England reported the target met on its monthly series. The annual points still cannot confirm a structural change.
More recorded appointments do not necessarily mean better access. National data, patient surveys and Stowhealth show why activity and patient experience need to be considered together.
How X-mR, Run Chart, and Bootstrap CUSUM tell radically different stories from 488 weeks of clinical data — and why it matters for governance reporting.
What Bootstrap CUSUM reveals about the long deceleration of UK GDP — and why conventional analysis has been telling us the wrong story.
Two identical plants, a 17% efficiency gap, and the residual CUSUM monitoring system — built on one reading per shift, pen and paper — that solved it.
Bootstrap CUSUM applied to 15 years of monthly data reveals four structural stages of decline. Not one policy intervention is detectable as an improvement. Tony Blair promised four-hour waits in 2000. Twenty-five years later, performance is worse.
Bootstrap CUSUM on a decade of Marylebone Road NO2 data finds the ULEZ signal appeared 18 months before the scheme launched. But the biggest single change was COVID — and the London-wide expansion improved air quality at a site already in the zone for four years.
Bootstrap CUSUM finds three structural change points in NHS anticoagulation safety data. The DOAC adverse event rate fell 96% from peak. But even by 2019, half of DOAC prescriptions were still not adjusted for renal function. The improvement is real. The job is not done.
The engineering solution has existed since 2010. The NHS declared wrong-route medication administration “wholly preventable.” Bootstrap CUSUM on six years of Never Events data finds one stage, no change point, mean 17.5 per year. Paul O’Neill fixed this at Alcoa. The difference is method.
The UK cut emissions 53% since 1990. Bootstrap CUSUM explores changes alongside major events, including energy market liberalisation (1998), the global financial crisis (2008), and the carbon price floor (2015). Timing alone does not establish cause. Transport shows no structural change at 95% confidence across 34 years.
Six bedside actions. 96% of NHS hospitals. 37 countries. Bootstrap CUSUM on 22 years of ONS mortality data finds one stage and no change point in either clinical series. The ratio between the two series suggests a possible change around 2013, the year the CQUIN coding incentive was introduced. Timing alone does not establish the cause.