Quality Improvement Zone

A funnel plot helps identify where observed differences are likely to be meaningful and where they may simply reflect random variation.

Funnel plots are particularly useful when comparing units of different sizes, such as hospitals, schools, GP practices, NHS boards or Local Authorities. Without accounting for sample size, comparisons can be misleading and may result in inappropriate ranking of performance.

A funnel plot typically includes:

• a performance measure on the vertical (Y) axis, such as a rate, percentage or average
• a sample size or denominator on the horizontal (X) axis
• a centre line showing the average performance across the system
• funnel limits, which show the range of variation expected by chance alone

The funnel limits widen as sample size gets smaller and narrow as sample size increases. This creates the funnel shape and reflects the fact that smaller samples naturally show more variation than larger samples.

Funnel plots help teams compare performance fairly by taking differences in sample size into account. They support understanding of variation and help identify observations that may warrant further investigation.

Using a funnel plot can help you:

• compare performance across organisations, services or teams fairly
• avoid misleading rankings that do not account for sample size
• identify potential non-random variation
• focus attention on areas that may warrant further investigation
• identify opportunities for shared learning across a system
• support evidence-informed improvement activity

Funnel plots help teams understand variation and distinguish observations that may differ from what would normally be expected by chance alone.

Funnel plots are most useful when comparing performance across multiple organisations, services or teams that differ in size. The aim is to understand variation and identify opportunities for learning, not to produce rankings or make assumptions about performance.

1. Define the measure

Choose a measure that can be compared across multiple units. This is often a:

• proportion
• percentage
• rate
• average

The measure should be clearly defined and calculated consistently across all units being compared.

2. Identify the units to compare

Decide which organisations, teams, departments or services will be included in the analysis. Ensure that the units being compared are sufficiently similar for the comparison to be meaningful.

3. Gather the measure and denominator

For each unit, collect:

• the measure of interest
• the relevant denominator or sample size

For example, this might be a percentage of patients receiving a service alongside the total number of patients eligible for that service.

4. Create the funnel plot

Plot the measure on the vertical axis and the denominator or sample size on the horizontal axis.

Add:

• the average performance line
• upper funnel limits
• lower funnel limits

The funnel limits represent the range of variation expected if only common cause variation were present.

5. Identify points outside the funnel limits

Review the chart to identify units that fall outside the funnel limits.

Where a point lies outside the funnel limits, the observed result is unlikely to be explained by chance alone and may represent special cause variation.

6. Investigate before drawing conclusions

Being outside the funnel limits does not automatically indicate good or poor performance. It suggests that the result differs from what would normally be expected and warrants further investigation.

Before drawing conclusions, consider:

• differences in populations served
• differences in case mix
• data quality issues
• variations in definitions or measurement methods
• differences in service design or practice

7. Use findings to support learning and improvement

Use the funnel plot to identify areas where further inquiry may be useful.

For example:

• units with unusually positive results may provide opportunities for shared learning
• units with unusually different results may benefit from further exploration to understand contributing factors
• patterns across the system may help identify priorities for improvement

The goal is to understand variation and learn from it, rather than to rank organisations or assign blame.

Interpreting a funnel plot

Points inside the funnel limits

Points that fall within the funnel limits show variation that could reasonably be expected due to chance alone. This suggests random variation.

Points outside the funnel limits

Points outside the funnel limits show variation that is unlikely to be explained by chance alone. This may indicate non-random variation and should prompt further investigation.

Important note

A funnel plot does not explain why a result differs. It identifies where differences are unlikely to be due to chance and where further learning or investigation may be worthwhile.

Creating a funnel plot

Funnel plots can be created in Excel, specialist QI software or statistical software. Some QI software packages can generate funnel plots automatically once the measure and denominator are provided.

If you are unsure how to create a funnel plot, seek support from your local improvement advisor, analyst or data specialist.