Big data is a fairly loosely used term that has become mainstream in the past couple of years. There’s more data than we know what to do with currently, and there is a mad dash to find “alternative” data sets in an effort to gain an edge over competition whether in actual industry or within investment markets.
I’ve spent a large portion of my career procuring real estate and economic data and validating/invalidating its usefulness in financial decision-making processes. I was doing it before it went mainstream. Along the way, I’ve learned a lot about data, decision-making and the importance of having a process. I highlight four key principles I’ve learned during my big data journey:
- Avoid over-optimization: With so many data sets available to decision makers, it’s very challenging to not attempt to use too many individual and potentially redundant data sets in decision-making models. Combining data sets without proper robustness tests can inadvertently create false-positives leading to greater reliability on a group of data sets than is warranted. Not all data is created equally and past predictive performance is by no means a guarantee of future accuracy or lack thereof. Focusing on key performance indicators (KPIs) and the one or two data sets that are most accurate in predicting those indicators is critical. Otherwise, decision-makers find themselves with too broad of a focus and using too many data sets, both of which are massive time sucks. I speak from experience.
- There isn’t always a needle in the haystack: Not every KPI has a corresponding predictive data set. Avoid forcing a predictive match between a data set and a KPI. This is easier said than done. Our own inherent biases and desire for a predictive data set can lead us to incorrectly rely on suboptimal data. Maintaining discipline is imperative if we are to derive the enhancement benefits of incorporating data into our decision-making processes.
- Know your data and its flaws: To maximize the power and utility of a data set, understanding what the data is supposed to represent is critical along with a firm understanding of how the data fits into your decision-making framework. Of possibly greater importance is understanding the limitations of the data in question. For example, is the data more useful in identifying inflection points in trends versus the magnitude of the continuation of an in-place trend? Knowing the ins and outs of the relationship between KPIs and supporting data allows for proper decision calibration at different decision-making points throughout a cycle.
- Process is still king: The best financial decision-makers have always used data in some form or fashion to assist in making decisions. However, data wasn’t and still isn’t the most important component to the decision. The actual process is what matters most. It’s always what has mattered most. If you don’t have a good decision-making process, there is no data set, no matter how apparently predictive, that will compensate for the lack of process over the long-run. In other words, if you don’t have a refined process that you consistently follow then data may end up being more of a detriment than an enhancement.