Data Scientists. Post 2 of 3.
Do I need one?
Having discussed what a data scientist is in my previous post, the million dollar question is do you need one?
Don’t believe the hype
Many of us remember the resulting scramble that occurred when the hype surrounding Big Data made its way up to the boardroom. CEOs who had never previously shown much of an interest in data or analytics suddenly started demanding to know what their organisation was ‘doing with big data’. A similar trend seems to be emerging with data scientists, whereby senior executives are requesting that the business hires in data scientists to solve all of their problems. This has caused many organisations to hire data scientists without giving sufficient thought to what they are going to do and whether they are actually required.
I have come across data scientists (or at least people who identify themselves as data scientists), being hired to do standard BI dashboard development and reporting. What a waste. It’s crucial to not get swept up in the big data/ data science hype and believe that data scientists are the silver bullet to all of your business problems, or that wherever you have data you need a data scientist.
It’s not all about size
A key consideration is the nature of data that you need your data scientist to work with. For me, it’s less about size and more about structure and cleanliness. If your data is already well structured in a relational database then it is more likely that a data analyst will be capable of working with it and a data scientist will not be required. If your data is dirty, unstructured or generally requires a whole load of wrangling before you can begin to do anything useful with it, then the programing skills of a data scientist may well be required.
(Un)structure the problem
A second key consideration is the nature of the problem you are trying to solve. As discussed previously, for me a key difference between data analysts and data scientists is that the latter is more comfortable with unstructured problems and approaches that require the development of novel solutions.
If your problem is fairly well structured and can be solved by the application of an out of the box technique then it more likely that a skilled data analyst will meet your requirement. However, if your problem is more nebulous and requires the development of a novel approach a data scientist may be required. For example, if your problem is effectively targeting your marketing campaigns to realise the most bang for your marketing buck, then this is likely to be solvable using historic sales data and standard customer segmentation techniques, perhaps with a bit of predictive modelling around retention thrown in for good measure. This is bread and butter stuff for a skilled data analyst.
However, if your problem is protecting your brand against attack by disgruntled customers who are influential on social media and also highly connected to existing high value customers: then you may well need to scrape some social media data from the web and text mine it to identify said disgruntled customers; use a graph database to understand their connectivity to existing clients; and throw a measure of social media influence into the analytics pot to understand if they could put off potential customers. This kind of problem is more suited to a data scientist.
The spice of life
Variety is also well worth considering when deciding if you need to hire a data scientist. If your requirement is for a set of processes to be set up and then repeated, then perhaps you need to parachute a data scientist in short-term to set up the process (if it involves dirty data and/ or requires the development of a novel approach), but then hand this over to your data analysts to manage long term. Employing a data scientist to manage a problem that has already been solved is a waste of your money and their time. However, if you are constantly battling a whole of host of problems that vary in their complexity, size and the nature of the solutions required to solve them then a data scientist may be required.
Don’t try to catch a unicorn
If you believe that you do need the specialist skill of a data scientist, don’t get too hung up on trying to find one that has every single skill you think you need or have been told that data scientists have. There is much to be said for creating diverse teams that collectively have the requisite skills, knowledge and experience — a sort of crowdsourced data scientist. (Of course this assumes that you have the luxury of hiring more than one person, which is often not the case).
Can you give them what they need?
Finally, even if the nature of your problem is such that a data scientist is required, if you haven’t got the capacity to give them what they need you are unlikely to retain them long term. More on this coming up in the next post!
