3 Ways to Ruin Your Business with Data Science Mistakes
Nowadays, companies of all sizes rely heavily on data science and business analytics to solve their business problems.
While data science has worked wonders for many companies, the same can’t be said for others. When applied to any business, data science can be a double-edged sword. If not implemented correctly, data science can lead to mistakes that cost your business thousands if not millions.
Before we can start addressing the ways in which data science can ruin your business, let’s see how it can help it. After all, despite the possible risks, the ROI on data science and machine learning can be quite high (which is why most companies are implementing it).
Let’s consider two companies (Amazon and Google AdWords) that managed to reap the benefits of integrating data science in business.
In 2017, Amazon had a total revenue of $177 billion. Many insiders would argue that around 20% (roughly $35 billion) of their revenue that year was generated as a result of their cross-selling and recommendation efforts, which are based on data science.
In 2017, Google AdWords’ total revenue was $95 billion. This is actually a bigger success story than Amazon. How so? Almost 100% of that revenue was generated through machine learning and business analytics.
When you use your browser to search for something, Google displays an advertisement that matches your interests, making you more likely to click on – and purchase – the advertised product.
However, there are cases in which data science in business can backfire. Like in the case of Tesco.
As early as the 1990s, Tescobegan using prediction models and big data analytics to improve their advertising efforts. The British company was able to provide better, more personalized ads, which impressively grew their profits more than 7x within two decades.
Several years ago, however, things started to go south. Tesco’s customers felt like the amount of data they had to share continued to grow, while the return on value for them was low. In a nutshell, the predictive models Tesco created, ultimately turned their own customers against them.
Tesco’s failure can’t only be attributed to their machine learning models as they also failed to enter the US market. In the end, this example shows that businesses should consider how the models they are creating are utilized with regard to creating value for their customers.
These 3 examples show how machine learning models and business analytics can boost or hurt a business.
There are several mistakes that you should avoid when applying data science in business but we will focus on three specifically.
The chart below shows the total number of UFO sightings per year since 1963. The data was taken from the database of the National UFO Reporting Center and this work was originally done by work by Dan Henebery and Josiah Davis.
Do you notice anything unusual? The number of sightings was steady for several years and then increased dramatically in 1993.
In September 1993, the first episode of The X-Files aired. At its peak, more than 25 million people in the US watched it. So what can we deduce from this chart? Without any additional information, the most likely explanation has to be that aliens were big fans of The X-Files and came to Earth to watch the series with us.
Next let’s take a look at the data below that shows the frequency of UFO sightings based on the time and day of the week (Monday through Sunday). Yellow-orange represents more frequent sightings.
You probably notice that most UFO sightings happen on Saturday nights. And it can’t be a coincidence that most parties happen on Saturday nights as well.
We already knew that they like to watchThe X-Files, but now we also know they like to party with us too. Next, let’s look at the average number of reported UFO sightings per week since 2010.
Most sightings happen during the week of 4th of July. Perhaps aliens love America and enjoy fireworks as well.
