by Damon Samuel –
Isn’t that the way it sometimes seems? We are in the age of Big Data, and that means that we have massive amounts of information at our fingertips—far more than we had only a few years ago.
The common view is that too much data is better than too little. That is certainly true. But having too much data comes with its own set of problems.
Whether you’re a C-level executive or a junior manager, you’re charged with making decisions based upon the data at hand. But you can’t possibly process all the data available for making every decision. And that problem will only intensify as the swell of data continues to grow.
So how can you sort through all that data, and find the nuggets that are truly relevant to the decision at hand?
If you’re in a position of management, you must make decisions. But if you’re a department manager, you’ll be making very different decisions from those that a CEO will be making. So the metrics that matter to you will be very different from the metrics the CEO cares about.
You must decide upon and define the metrics that matter to your specific focus of responsibility. Do that successfully, and you’ll slash the amount of data that you’re sorting through to run your business better.
Finding the bits of data that specifically informs your decision will require further winnowing.
You might start with a heuristic process, working through a series of questions to help you focus on the most relevant bits of data for the decision at hand:
A heuristic process will narrow the field considerably, but will probably leave far too much data still on your plate. You can use advanced analytics to narrow the field even further. Techniques such as Factor Analysis, Principal Components, Vertical Clusters and so forth are designed for data reduction. These advanced analysis techniques will help identify metrics that move together, and that are representative of an underlying phenomenon in your business that you need to pay attention to.
By working through the above steps, you’ve likely eliminated lots of metrics that were just clutter and distractions—data that will probably be very important for other decisions, but not for this decision.
But you may need to trim the data even further to get down to the few bits of data that truly matter for this decision.
You can do that using regression techniques. You can also use structural equation models (a.k.a. system equation models). These will combine aspects of regression with a principal component analysis to reveal important relationships among the data, helping to identify data that can be informative for this decision.
A company I worked with saved millions of dollars by focusing on the metrics that matter.
This company was very focused upon their net promoter score, as most companies are. This company had a tremendously large marketing budget, and they were planning to expand it significantly in an effort to improve the score. But before spending that money, they wanted to understand how their marketing budget influenced their net promoter score.
Much of the data was generated through analyzing the marketing budget spend: information about how much media was purchased by which channel; how many impressions that media generated; how the messages were received; etc.
There were also large feeds of consumer perceptions on individual marketing collateral – favorabilities, brand and message recall, and so forth, numbering in the hundreds of metrics – which could be sliced and diced any number of ways.
Beyond the media metrics, we would have to control for other impacts to NPS, including metrics that were operational in nature, such as the average speed of answering a phone call; first call resolution; time span between initial service request and service execution.
In all, there was a pool of over 1000 metrics collected from daily, to weekly, to monthly to pull from.
Through a combination of Principal Components and Regression analysis, we narrowed the field down to the few bits of data that really mattered, we found that the media spend had very little impact on customer satisfaction. That insight indicated that the proposed media spend of $30 million would likely be ineffective at significantly improving our net promoter score. Instead, we spent less than half of that $30 million on improving a specific service issue (identified by the metrics-that-matter process).
The result? Millions of dollars saved, and the net promoter score improved by a factor of six times what would have happened if the budget was spent on media as planned.
It’s important to emphasize that finding the metrics that matter is a not a static process. It’s a dynamic process. It’s a process that you must constantly tweak. Few things in the world remain the same over time, and that’s particularly true in the world of business. The metrics that matter today might not matter tomorrow.
But certain metrics will always matter more than others. And that’s why it will always be worth the effort to find them.
#IdeasRealized