When you're trying to measure a result and just can't seem to find the perfect measure, you're probably in need of a proxy!
Imagine for a moment that you're the Director for a government agency responsible for building and maintaining the roads in your state or country. And the politicians are asking you to prove that what you're doing is improving the quality of roads. How do you measure the quality of roads, exactly?
Imagine once more, that you own a business whose values include being environmentally friendly. Are you really going to put the effort into understanding, counting and calculating all your employee's carbon footprints?
Good proxy measures are stand-ins that can help you understand the results that matter to you, when you don't yet have the perfect way to measure them. Usually they are measures of correlated results, measures that you'd expect to behave reasonably in sync with a good measure of the real thing.
Why would you use a proxy measure?
Measuring results like those mentioned above poses several challenges that are big enough to drain your commendable intentions to objectively know, and reduce you to a feeble lump that utters "it isn't possible to measure this!"
One challenge is that the result is so broad it's almost a motherhood statement and thus near impossible to make it tangible enough to measure. What is road quality anyway? Is it how long the infrastructure lasts? Is it how comfortable and kind the road surface is for cars and their drivers? Is it a combination of these things?
Another challenge is how to get the data in a way that costs less (hopefully far less) than the value having the data would bring. Imagine the cost of following all those teenagers that attended your program, to track their incidence of road accidents. It probably costs more money they you have just to run the program!
And a third challenge is measuring a result in a way that the right people understand what it means and how they can do something about it. If you told each employee what their carbon footprint was and that they must reduce it, would they really be inspired to action?
Good proxy measures side step these challenges.
How do you find good proxy measures?
An ideal measure of the quality of roads would encompass all the dimensions of road quality - surface roughness, corner camber, corner sharpness, road shoulder condition, traction (there are loads of them). It takes a very long time to construct a meaningful measure like this. In the meantime, a proxy measure might be something like the mean number of kilometres between suspension repairs for cars registered in the locality. A well designed survey of local car owners and/or suspensions shops could inexpensively estimate this.
To get an ideal measure for the impact that a driving accident awareness program has on teenagers' road accidents would involve a study that followed participants in the program and compared their incident rate and severity with a control group of teenagers that didn't participate in the program. Costly. Instead, a proxy measure might be to track the trend in road accidents involving drivers aged between 17 and 19 years. Data readily available from police records.
And even though measuring your carbon footprint is all the rage for the green crowd, it's not nearly as practical and easily understandable as proxy measures like percentage of lunches that are take-away (think of all the packaging) or average number of days to fill the waste basket (if only full waste baskets are emptied).
Good proxy measures aren't perfect, but their trends help you know something about the result you're interested in, quickly and easily.
What are the risks of using proxy measures?
There is a price to pay for proxy measures. Often they only tell you part of the story, and often they are influenced by other forces that can bias or distort your understanding. Often you have little or no control over the data or calculation of the proxy measures.
But when you appreciate their limitations, they can still speak volumes to you about those hard-to-measure results.
Choose them consciously, from a starting point that involves really understanding the different angles or sides of the result you want to measure. Describe your result as tangibly as you can, even if it's not a complete description. Leave your worries about data availability until after you've listed and considered several possible proxy measures. And open your mind as much as possible, by involving out-of-the-box thinkers and giving yourself a few short sessions (instead of one meeting) to consider options.
