Three Data Strategy Oversights – Part One

Three Data Strategy Oversights – Part One

by Ziko Rajabali

March 28, 2017

In the face of day-to-day operational concerns, it’s easy to adopt the mantra we’ll cross that bridge when we get there. There is wisdom to this philosophy: given limited resources, activities must be prioritized to attain success. But what do we do if the proverbial bridge doesn’t exist?

Data is the raw material of the bridge to cross when we get there
click to tweet that

When that bridge is a strategic decision, it should be made out of data that spans successes and failures, seasonal changes, political and environmental cycles. This doesn’t undermine intuition or automate the decision; rather, it provides an objective background to guide and support the decision. The challenge is that many organizations are crossing a bridge that doesn’t even exist. They are trying to make informed strategic decisions with sub-optimal data, and it’s usually because one of the following three data strategy considerations has been overlooked.

Discover Valuable Sources of Data

Even in organizations where data from all applications & ERPs are not integrated, most have found a way to explore the data through built-in reports or tools that directly access the data in the database. While this has its pitfalls (like grinding a production system to a halt), at least they are aware of the value of that data and it is used when making important decisions. However, there are many sources of data that are overlooked for their potential, including sensors used in operational feedback loops, external sources such as weather, and unstructured sources such as excel sheets for operational reports and even emails.

Emails are a great example of unstructured data because it is usually stored for many years if not indefinitely and when aggregated and siphoned through a machine learning algorithm, can yield some very interesting insights. For example, by using a sentiment analysis on past emails, an organization can get an objective measure of morale over time. This can be highly valuable when planning certain activities that are perceived as highly positive (bonuses, employee appreciations) or highly negative (layoffs, weak earnings). Let’s carry this idea a little bit further. If the objective measure of email sentiment analysis is run through a predictive analytics / machine learning algorithm, we might uncover a leading indicator about how morale shows a rapid dip right before the company loses a group of high-talent employees through attrition, or we might discover that sales always shows a spike after morale jumps instead of the other way around. Whatever the actual discovery, it’s reasonable to expect that individual actions will influence the entire group and therefore the outcome of the business.

When planning long-term data strategy, it’s easy to ignore additional data sources using anecdotal arguments. “There’s just too much data to transmit and store” is common for raw sensor data; “That information is outside our control” is typical for external sources of data; and “It will be too expensive to clean up and organize” is usually argued for unstructured data. The truth is that most of these perceived hurdles have been addressed by technology advancements. Don’t casually ignore these sources of valuable data without objectively considering them as part of your data strategy.
Click here for part two!