CFO of Sandline Global & author of Deep Finance, Glenn has spent the past two decades helping startups prepare for funding or acquisition.
With the ever-increasing amount of data available to businesses of all sizes, it is imperative that companies work to get a handle on what data they have on hand and what other data might be available to them. For those not currently using data and analytics as business tools, it may be overwhelming to consider where to start. But there is a simple algorithm to move your business from analog to digital.
The process starts with understanding the business problem you’re trying to solve. What do you want to accomplish?
From financial planning and analysis (FP&A) to product development, the most successful companies are using data to drive better decisions to become more efficient and achieve overall business objectives. They can balance those objectives with predictive data that identifies and mitigates risks and drives everything from top-line revenue to bottom-line income.
Once you’ve identified your goal, your first action step is to gather and organize whatever data is available. This is going to be a mix of internal data that is proprietary to your organization and publicly available external data. The data will come in all forms—from structured to unstructured, from plain text to geodata—and will come from sources as varied as internal software systems to public tweets and output from IoT (internet of things) devices.
But this data alone is not enough. While data may well be the fuel that drives the next industrial revolution, it is of no use in its raw form.
One of the greatest quotes that explain the data science process is something Clifford Stoll is often credited with.
“Data is not information. Information is not knowledge. Knowledge is not understanding. Understanding is not wisdom.”
Once you’ve wrangled the available data, it is time to organize and consolidate that information into a workable collection. From here, we are ready to begin the first steps of data analytics.
Step One: Descriptive Analytics
Descriptive analytics is the step where we start to glean information from what we’ve gathered. We do this by wrangling, consolidating, inventorying and organizing to provide context for what has happened historically. We can visualize historical sales methods, profit margins, operational performance and other metrics over time so that we can identify trends and patterns in the data.
As these patterns emerge, we can move into the next phase of data science.
Step Two: Diagnostic Analytics
For intellectually curious business leaders, visualization and explanation of historical data should lead to immediate questions. Looking at sales trends over several years, for example, a seasonal dip in sales may be evident in a particular month or quarter. Or you may see that a great number of customers who canceled their service had a higher-than-average number of support calls in the month prior to their cancellation.
Diagnostic analytics attempt to answer why historical events have happened. This is the phase where data analysts attempt to move from hunch to hypothesis. Here, we find correlations between activities and try to prove or disprove our theories about why.
Now we’re turning data into information, but what do we do with that information? We work to turn it to our advantage. We work to turn it into knowledge. This is an interim step toward understanding.
Step Three: Predictive Analytics
By looking at past trends, we can start to model out future activities through trend analysis and correlations. This may be as simple as basic linear regression or—for more complex issues—could require the use of machine learning algorithms.
As machine learning becomes more accessible to all businesses, there is an increased opportunity for all of us to gain value with these powerful tools. Regardless of the methodology used, there is a process for training a machine learning model to predict the future by identifying trends and patterns in the data.
The machine learning process involves building a model, then using your existing data to train the model to make predictions. In machine learning, these three subsets of data involve a training set that is used to establish the model’s predictive capabilities, then a validation set that is used to test for bias and other errors. Lastly, it establishes a test set that confirms the model is properly tuned before being put in place to make predictions.
At each step along the way, we are adding greater and greater value through analytics. We have moved from visualizing the past (descriptive) to understanding why past events happened (diagnostic) to now being able to model out the future (predictive).
We’ve seen how to use data to create information we can use to drive business questions and then progress to knowledge, which, as we delve deeper, drives understanding.
How then do we ultimately turn that understanding into wisdom?
Step Four: Prescriptive Analytics
The ability to glimpse into the future gives companies great power. But sight without action is of little use. The power comes in the ability to not only identify but also impact trends—stopping negative momentum or enhancing upward movement.
This highest order of analytics is the culmination of the three prior stages and is where the real power of analytics truly shines. When data scientists have moved through the stages of analytics and have a true understanding of their data, they can then both ask and answer the right questions. They understand the correlations between the myriad of factors that impact their business and are able to control the future by instantiating activities that influence it.
Putting It All Together
We’ve looked at the full data landscape from the initial capture of available data, through processing and interpretation of it, all the way through using it to predict the future and make corrective actions.
We have seen how to convert data to information, information to knowledge and knowledge to understanding. This leaves only the final step: to achieve the wisdom that comes when we pair our human intellect with the science of analytics to drive decisions.