It may be evident that data analysis has become a buzz phrase. From human resources to supply chain management, from marketing to finance, analytics has become a key tool in these business decisions. Businesses today know that without this understanding of data, they will be left behind.

Data analytics enables companies to arm themselves with data to reduce costs, increase sales, and streamline operations. They can also use analytics to predict future events. While previously executive decision making may have been based on instinct or tradition, today they ask themselves: “What do the numbers tell us?”

From this question, the data can tell companies a little bit about their customers and operations. Analytics can indicate how many calls a salesperson has to make before getting an interested prospect to review services or if a certain product could take off in a few months. But Analytics is useless if the data quality is poor.

The challenges of data-driven companies

Businesses are benefiting from data-driven decision making, but there is a steep learning curve. Struggling with large volumes of data, coming from multiple data silos and in different formats, is challenging. The ability to handle large amounts of information, integrate it from different areas of the company and combine it to obtain actionable data in real time is easier said than done.

One of the main challenges is data quality: Without high quality decision making is likely to fail. As with any data-dependent process, decision-making depends on the quality of the information. As the saying goes, “trash comes in, trash comes out.” Incorrect or incomplete information will lead to incorrect predictions and misleading descriptions.

Where do data quality problems originate? One of the problems concerns the initial assumptions on which an analytical model is built. In marketing, predictive models could apply to next year’s marketing budget. You could try to make marketing expenses more efficient by analyzing customers in new groups: those who are going to buy, regardless of advertising; those who will buy only after seeing compelling advertising; and those who will not buy. The idea is to spend resources only in that middle group because the other two are a waste of money.

But what if the customer profiles in these other groups is not correct? What if the demographic definition of a customer’s “going to buy anyway” category is based on misinformation, such as brand loyalty that ignores competitive technology? A mistake like this can ruin a marketing campaign, no matter the quality of the predictive analytics.

What is the solution? Test all hypotheses before incorporating it into the model. Be sure to get in, because even the business truths that are taken for granted could be off base.

Do we sign up?

The example above involves an error at the front end while building the model. But another significant mistake often occurs at the end of the process: failure to perform an autopsy on the results of a predictive model. In the example I just presented, a more effective marketing campaign could come out of the rough numbers: sales improved, while marketing spent less.

But it is not that simple. Measurable improvement is all well and good, but success in analysis is measured by the amount of an improved process compared to potential success. One number going up while another goes down actually says one thing: that a process is moving in the right direction. If left at that, the company cannot yet assess the effectiveness of the analysis process.

What’s missing? First, specific goals should define the modeling process: optimal sales goals that can be compared to sales, for example. Having those numbers makes it possible to make the analysis process a success, not only against past performance, but against future potential.

So, for example, if a food distributor wants to increase sales by 8% over the next year, they first have to look at their current sales number and compare that number to past growth, for example, five years. to see if this objective prediction has merit.

With this type of thinking in place, analysis can improve a process today, with the potential to continually improve, fine-tuning not only the results, but the quality of inputs to the process.