Accelerate Marketing Growth With These 3 Data Science Tips
Sure, data scientists can be portrayed as the mad scientist type, toiling away in a dark room with numbers, data and models, day and night. So why is The Harvard Business Review touting data science as the sexiest job of the 21st century? Data scientists are sexy because the work they are doing is extremely practical for anyone in business. Yes, they are extremely smart, and better with numbers than your average joe, but data scientists and their methods are more relatable than you may think.
They are working to solve real-world problems and create the ultimate customer experience. Many of the most successful consumer companies like Amazon and Netflix are employing data scientists to do just that. They are leveraging data to recommend what shows you should watch, what books you should read and what stocks to buy. Businesses pay huge salaries for this kind of insight. And while you likely won’t ever have to dissect the Stefan-Boltzmann formula, the truth is that anyone can learn these problem-solving techniques from the work of data scientists to help reach your goals.
I connected with our data scientists to gather three problem-solving tips that can be used to accelerate marketing growth. Here they are:
Beware! The Tunnel Is Closed – Avoid Bias
As we are moving through the workday, trying to balance our workload with our personal lives, we often move very quickly and simply fixate on the end goal. We operate on full speed and make decisions in the blink of an eye. Back in 2008 the FDA determined that bisphenol A (BPA) from plastic containers was safe when leeched into food. By taking a look at independent research studies, this is not the case. In fact the research shows that in over 90 percent of studies, negative health effects were stemming from small amounts of BPA.
As humans, we have a tendency to put more faith in information that agrees with what we already believe or more faith in information coming from funding sources. Similarly, it is natural to discount opinions and data that disagree with our beliefs. This tunnel vision is a form of bias and data scientists know to avoid it. When problem solving, data scientists focus on challenging any solution with supporting AND contrary data to ensure they are taking in all of the information available to draw their own conclusions, rather than having blind faith in data.
Keep It Clean – Ensure Data Cleanliness
The amount of data publicly available is growing exponentially. There are an infinite number of data points coming from the web, social media and beyond. The problem is that only a miniscule percentage is useful for determining the answer to the question or process it is trying to solve.
In searching for a solution to any problem, we must consider the largest volume of information available, but also ensure its cleanliness and validity. We must account for noisy and random data. Data Scientists have discovered many techniques to automatically detect spurious information and to retain only the meaningful information. Data scientists know to look for and remove outliers and they use super-computing machines to scan all data and remove weak signals. Data scientists know that data sets must be clean, relevant and correct. We too should reject data that is inaccurate or old because it may no longer be relevant. Poor-quality data can lead to significant costs and wasted time.
When It’s Too Good To Be True, It Probably Is
Nate Silver of FiveThirtyEight offered a great example of over-fitting or over-generalizing data back in the 2012 election. Automated polling systems were exclusively targeting landline phones, mostly leaving out the young population who tend to only have mobile devices. The sample wasn’t well represented and as a result, the surveys were skewed.
In machine learning and data science, over-fitting occurs when your data fits the model too well. Your hypothesis is so good that it describes your sample nearly down to the last detail. This sounds perfect right? Wrong. This model is actually useless for any other sample because it will never be flexible enough to meet other needs. Just like in the polling example, the shoe is over-fitted and is in fact, skewed. Data scientists know to always focus on building a representative sample and avoid doctoring it to achieve the results they desire.
Data scientists are in such high demand because the skill set is so specific and so rare – they understand statistics, programming, modeling and can communicate in a clear and concise manner. A data scientist may come from academia, research or software development, but that should not serve as any intimidation. For you to start applying the principles of data science to your work and life.
Image Credit(s): fotupia