What is multi-vari analysis?

Wikipedia says that:

Multivariate analysis (MVA) is based on the statistical principle of multivariate statistics, which involves observation and analysis of more than one statistical variable at a time. In design and analysis, the technique is used to perform trade studies across multiple dimensions while taking into account the effects of all variables on the responses of interest.

So multi-vari analysis is a technique for viewing multiple sources of process variation. Different sources of variation are categorized into families of related causes and quantified to reveal the largest causes. Instead of telling you just that shift is a factor and machine is a factor, multi-vari analysis provides a breakdown for you to say, for example, that machine B on shift 1 is causing the most variation. It won’t quantify the variation…just show you where it is. Once you’ve narrowed your list of inputs, you know where to go take more data, do more training, plan further studies, etc. Pareto analysis won’t get you to that level of detail.

Multi-Vari is the perfect tool to determine where the variability is coming from in your process (lot-to-lot, shift-to-shift, machine-to-machine, etc.), because it does not require to manipulate the independent variables (or process parameters) as you would with design of experiments.

Because it provides a great way of analysing the effects of multiple factors multi-vari analysis is widely used in six sigma projects.

You can display the effect of categorical type inputs on a response on a multi-vari chart. It is one of the tools used to reduce the trivial many inputs to the vital few. In other words it is used to identify possible Xs or families of variation, such as variation within a subgroup, between subgroups, or over time.

Although simple in nature, such a chart has considerable power. It can help you isolate the major (dominant) family of variation. For example, does most of the variation in downtime occur within-machine, machine-to-machine, within-operator, operator-to-operator, within shift, shift-to-shift, within-location, or location-to-location?

Once you have determined where the major source of variability is coming from, then you can target more efficiently your experimental designs (DOE) to reduce that particular source of variation.