Improvement projects should have a small, balanced set of measures that are tracked over time.  These should include:

• One or more outcome measures (associated with the aim of the project)
• Process measures (things that have to happen reliably to achieve the desired outcomes), and
• Balancing measures (to check for possible consequences elsewhere in the system). 

It can be tempting to set out to measure a lot of things, but too much time spent collecting data could be at the expense of testing, implementing and spreading changes. The aim is improvement, not just measurement. Only measure what is useful and try to keep it as straightforward as possible.

In general, a small set (containing one or two outcome measures, a few process measures and a balancing measure) is likely to be about right for most improvement projects. 

Work out and specify how the data will be collected and monitored, and whose responsibility it is to do it.  It’s important to think at the start about how you would like to analyse the data, to make sure you collect all the information you need, in the right format. Use the measurement plan tools to help you develop and record your measures.  

Start by developing ideas for your measures – 

1. clarify the improvement aim and how this can best be measured;
2. use the driver diagram to help assess the key processes that need to be measured to ensure they are happening reliably;
3. identify any potential unintended consequences of the work and set up balancing measures for them.

Have you picked specific measures or concepts?  For example ‘staff experience’ is not a measure, it is a concept. A measure for this concept might be ‘% of staff who report they had a good day’ or ‘average rating of staff on how supported they feel in their work’.

There are many options of what to measure for each concept so you need to decide what is going to be the best for you.  Do not underestimate how challenging this, often overlooked, step can be and build time to have the necessary discussions into your project plan and test the measures if necessary.      

You will need to decide what the most appropriate type of data will be for your measures, depending on what you want to achieve and what you need to learn.  For example, will it be a numerical measure, a percentage, a rate or a simple count?  When you want to look at your data in an SPC chart [insert link to tool] you need to understand what type of data you have for each measure so that you pick the best chart for the job.  The terms used for these are described below.

Usually, for improvement, if you have a variable measure like the amount of time something takes, it is best to use the actual data in your measure rather than convert it into a percentage.  This approach to classifying a measure into more than or less than a specific value, and calculating a % is often used for performance or accountability– e.g. compliance with a target – but this loses valuable information in the data.  The charts below illustrate this point. Using the actual time in the chart on the right the team can see improvement, whereas it is hidden in the chart on the left.

It is important to have clear operational definitions for each measure that are agreed and understood by the team.  If variability is being introduced to your measures because of them being collected differently each time then it will be very difficult to understand the variation over time. 

Operational definitions set out what method of measurement will be used, and agreed descriptions of how we will recognise whatever it is we seek to measure. They specify what will be included and excluded in the measures – including specifying a denominator if relevant.