Wanna become a data scientist within 3 months, and get a guaranteed job? Then you need to check this out !
What is a data strategy?
Data strategy is the one thing many companies need, especially startups, even if they don’t know it.
So, what is a data strategy? How can we define it?
A data strategy is defined as the strategy around the collection, storage and usage of a data, in a way that data can serve not only the purpose behind the selling point a startup, but also open up additional potential monetisation avenues in the future.
Let me explain a bit what this is all about.
Throughout my work, I have seen many times two types of situations:
Data strategy case study 1
A company wants to conduct a statistical analysis. As the problem is investigated, it becomes clear that there are data that could have had been collected and could have improved the analysis, but are not available.
I had worked with a car rental company that wanted to forecast demand for the upcoming months and then create customised deals for each client. The company is not keeping information regarding the marital status and the age of its users. This information could have been very easily collected on its platform throughout all these years of operation. However, since this was not done the company will have to go with a suboptimal model.
Data strategy case study 2
A startup requires some sort of data science service. As the discussion about the system progresses, it becomes clear that the requested service depends on many different kinds of data. The collection and storage of this data exposes new avenues for monetisation that can help improve the startup’s revenue strategy.
A common case is a recommender system. A recommender system can benefit from all kinds of information about the users: age, gender, purchases and possibly other things as well. Designing the platform in a way that improves information collection from its users, results in a big comprehensive database that can be used to improve the recommender system but can also be used for other purposes. For example, this information could be used for better managing discount deals, improving advertising or even the user experience on the platform.
So, what do these two cases have in common? In both cases a data strategy can save the day! In the first case, a data strategy would have provided the company with tons of additional revenue. In the second case, a correctly implemented data strategy could provide the startup with a competitive advantage in addition to increased revenues.
However, a data science strategy is less effective if it is not in place since an early stage. This is why it is important to consider it as an integral plan of the product roadmap and the business plan. One of my favourite sayings is “failing to plan is planning to fail”, and this couldn’t be more true for a data strategy!
Tips for designing a good data strategy
I believe that the ability to design a good data strategy is one of the points that distinguishes between a great data scientist and someone who is simply good at analysing data. While both of the problems described above could be easily solved without any additional modifications to the business, a data strategy can add immense value to a company, but designing it requires a good understanding of both the problem, the current market conditions and the business itself.
So, what are a few tips for designing a good data strategy?
First of all, you can’t beat experience. If a data scientist has worked with companies that faced similar problems in the past, then it will be easier to think of things that went wrong or well and add them to the strategy. However, an experienced data scientist (someone with 5+ years of work experience who has solved many different problems) can come up with data strategies for domains in which they are not familiar.
Secondly, you can always find case studies on the web which might give you some inspiration. Just googling “machine learning case studies” gives you 166 million results (and the list is growing)! I have also written about some use cases here as well. One such use case for example is the use case around predicting football injuries.
Thirdly, academic journals can be a good source of what can or can’t be done. Navigating the academic landscape can be, however, very time consuming and tiring, if you do not have doctoral training. I would suggest that you start with some famous conferences that publish papers on applied machine learning such as ICML or ECML-PKDD and then move on from there. Also, focus on publications in the last 4-5 years, since the state of the art changes rapidly, so older publications might not be as relevant.
Finally, just ask a data scientist! I have provided this service many times in the past. Sometimes, I’ve even done it for free, since it took me less than 30 minutes to figure out what the client was doing wrong, and I felt compelled to just point out their mistakes. That way, I know that in 6 months time, the client will be ready to work with me, as they will have collected data the right way. This allows for a smoother and efficient collaboration.
Also, if you are interested in educating yourself, then you can learn more from the Tesseract Academy. This is a company I started to specifically educate non-technical decision makers (CEOs, managers, business development, startup founders, etc.) in how to develop a data strategy, and use data science within their organisation.