In the previous post, I explained how I view data science and its value added in the business context. Despite it clearly offers a huge potential for many companies, I think data science is not used successfully and widely enough as it should be. One important barrier is that it stays very abstract. And surprisingly, it is the small/medium businesses that can overcome the barriers in its implementation the easiest.
Why is Data Science so difficult to grasp?
It really does sound quite glamorous – being able to read and predict from data, having automated dashboards and apps…. All the fancy buzz-words that go together with data science such as big data, artificial intelligence or deep learning show that it is something that attracts attention. But as one nice quote put it, many of these words are like teenage sex – everyone talks about it, nobody really knows how to do it, everyone thinks everyone else is doing it, so everyone claims they are doing it…
So why it is so difficult to make Data Science happen when it so great?
The underlying business question might not be articulated well.
Sometimes it seems that the main focus of companies is on the technical aspect of storing the ever-growing data. The term big data really is being used in ridiculous expressions – personally my favorite one being ‘big data at the speed of big data’.
As I mentioned earlier, one should never forget that it is the underlying business question that still matters the most. The definition of performance can vary between industries and companies – it can be a financial goal such as increase in profit, decrease in costs, increase in turnover, preventing customer churn or even non-financial, like improving customer satisfaction, increasing safety or even being fit. Obviously one should be able to dig deeper to understand why the performance is low, but the high level performance is usually not about more than 3 key numbers.
Some businesses actually don’t need it.
It might not be, what you’d expect me to say, but really, not everyone needs data science. As I already suggested, data science is not necessary for actual operations of most companies, but it is very useful to get an edge over competition.
When there is no competition, there is no actual pressure to look for efficient solutions and the effort to implement data science can very easily fall through. To illustrate this, I again use the metaphor of data scientists being doctors of businesses – when there is no real pressure, patients (businesses) want to know all the details about the procedure and they might not even decide they want it in the end, on the other hand, when there is an emergency, the only thing, which is important is that you will save lives and no one is interested in knowing exactly how.
Results of data-derived techniques are not always intuitive.
Strictly speaking, this is obviously a great property as counter-intuitive relations can’t be just easily found by other means, but in the same time it is very challenging for the business representatives, who tend to look for some underlying logic and do not accept the results as long as they don’t find it. It is always surprising to me how many people expect data science to “only confirm, what I already know (as I am the expert, so I have to know everything)”.
People do not recognize that Data Science is a skill.
This is a weird contradiction. On one hand, people don‘t understand it, on the other, they refuse to acknowledge that it takes a serious skill to be able to develop data science solutions. It happens quite often that marketing executives have this idea that they will just hire some random university student, who will sit down at the computer and just like that develop a functional machine learning model. Well, bad news, it does not work like that. Imagine that you would just say to a senior accountant that you would rather look for some student to do their job. Only a very irresponsible financial director would do that, isn’t it right?
Why is it easier to successfully use Data Science in SMBs?
In my experience, it is much easier to overcome the above mentioned barriers for small or medium businesses than for large corporations. The main reason why I think it is so is that SMBs are typically more centralized in their decision making. Simply put, the same person (or a tight group of persons) holds the company vision, decides the strategy, bears the costs and uses tools that they invest in. The emphasis is then very often put on profits of the business (in a more pronounced way than by larger companies), all the more so that the pressure to succeed (competition) is usually bigger for starting and/or small companies. This creates ideal conditions for the success of performance-oriented data science solutions.
Moreover, SMBs are usually less regulated (both internally and externally) therefore more efficient and less costly tools such as cloud data hosting or open-source software can be used, together with externally purchased data science services, rather than maintaining costly in-house data science teams. The efficiency of data science solutions that are built on this infrastructure stays high, which means that data science potentially brings higher ROI for SMBs than for other types of businesses.
In the next posts: How does an implementation of a data science solution typically look like?