Most of us seem to remember the time when one of the main issues when making business-related decisions was a lack of relevant data. How the times have changed! We now face a completely different challenge, with overpowering and overwhelming data, which makes it equally difficult for decision-makers to make the right call. Finding the right or the best data that can serve your purpose and their correct interpretation is what it’s all about nowadays. Make no mistake – that is an extremely challenging task.
With the global obsession with “big data”, there are many tools and techniques supposed to help organizations use data in the best possible ways, but more often than not, the results achieved fail to meet the expected ones. Knowing what data is available and how to use them can easily be the difference between a success and a failure. That’s why we need to know how data are analyzed and what the advantages and disadvantages of each method are.
The simplest method, familiar to everyone, is the arithmetic mean, a.k.a. “the average”. You add all the numbers, divide the result with the number of items and what you get is the overall trend of a data set, which provides a quick snapshot of your data. It doesn’t take a genius to do the maths, but it takes a smart decision-maker to know that they can hardly rely on the information obtained in this way, at least without combining this with some other methods. Namely, if you perform an analysis without a proper context, you can easily be misled and make wrong conclusions. It is often closely related to the mode and the media (two other measurements near the average), but you need to be aware that if you use a data set with a high number of outliers, the results will not be accurate enough.
When you see the Greek letter sigma, you know that it’s likely used to represent the standard deviation, i.e. the spread of data around the mean. Basically, if the standard deviation is high, it means that the data are spread more widely from the mean. Still, the data obtained in this way can give you a quick overview of the dispersion of data points and, as such, are useful. The problem is that they can be deceptive if analyzed in isolation. Such a problem becomes apparent especially when there is a strange pattern, such as an abnormal curve or a large number of outliers.
This method of data collection can help you find out how dependent and explanatory variables affect each other. The regression line shows whether the relationships are strong or weak. Businesses use this model to determine trends over time. For example, if you wish to see how the Australian media report on your company, you have to turn to the data provided by social media analytics, among others. However, you need to know that regression is not very nuanced and you need experts in the field to help you combine the data obtained using this method with others in order to get the correct information, which can help you make informed decisions.
Sample size determination
Measuring a large data set is tricky business and most problems occur when information is obtained from every single member. Instead, you should focus on a sample. Provided you get a good sample, you’ll save not only a lot of time but also money and other resources. It’s always a better idea to have a relevant sample than to use data that relates to the things outside your sphere of interest. In that case, you can get overwhelmed easily and the result may not reflect the real answers you’re after. You have to be careful, though, to select the right size of a sample if you want to be accurate and that’s where proportion and standard deviation can help you. The problem could be that your assumptions might be wrong in the first place, thus making your analysis totally useless.
Also known as t testing, hypothesis testing determines whether a certain premise is true for your data set or population. The results of a hypothesis test are considered to be statistically significant if you establish that the results couldn’t have happened by random choice. Actually, the randomness of data is probably the biggest problem you can encounter in data analysis, which is why you really need to have an expert decide on the strategy and interpretation of results.
As you can see, each method has its merits and pitfalls and you need to approach data analysis very carefully. If you’re supposed to make decisions based on the results, you want them to be correct and relevant. Otherwise, you’ll be wasting a lot of resources for nothing.