Do not let your data lake become a data swamp! This timely warning has been heard frequently over the past years as companies have invested heavily in novel ways of capturing and collating data. While it might feel like one buzzword accosting another, it cannot be denied that some organisations have fallen into the big data trap: storing vast amounts of data, yet failing to turn this into action that bolsters their bottom line. How do we prevent this?

Typically, one would start with either a hypothesis (I postulate that) or a hunch (I believe that) with the goal of demonstrating that they hold true. After all, while you can make decisions based on intuition, the house always wins, particularly in a world of increasingly complexity. The goal therefore, is to move up the analytics value chain from hypotheses and hunches to insight. To start this process, we need to collect data (from Latin datum “a given”). This data can come from any sensor or system, as long as it is of good quality.

Collecting data should never be the end goal. Photo by Tony Webster.

The idea of a value chain is that every step in the process increases the value of a product. In the case of analytics, a hypothesis or hunch will inform what data you need to collect. Once you have collected your data, you will want to define metrics. Metrics are the signals, extracted from data, that measure some value you care about. Most important are those metrics that capture the goals you have set (e.g. profit); we call these outcome metrics. Metrics that capture progress towards your goal (e.g. sales) are called process metrics.

Many businesses stop here; they report the metrics and make decisions based on these raw numbers alone. In doing so, they lose out on much of the value in their data. To truly unlock this, they must perform analysis. This involves understanding how the metrics affect each other, and why this is the case. We can look backward at historical factors to explain what has already happened, or look forward to predict what will happen; this is called predictive analytics.

Analysis often involves pattern detection and statistics. Photo by Jorge Jaramillo.

The risk with analysis is that we become stuck in trying to identify the perfect explanation for our metrics, losing sight of our original goal. Therefore, we need to process the outcome of our analysis into an insight: a finding that is credible (I believe that!), novel (aha!), and actionable (so what?). Insights can be as emotionally compelling as an opinion, but are based on rigorous analysis of data and should stand up to close inspection, inspiring confident action.

Naturally, all businesses take actions, but these are often not driven by insight. Making data-driven decisions requires an organisational culture attuned to the importance of analysis and insight. Because an organisation that takes insightful decisions can be more confident in its decision-making processes, it can be assertive in the execution of its work, making less mistakes over time. In other words, following the analytics value chain builds better companies.

— Colm & Ryan