Why companies can fail in integrating Data Science?
The quantitative explosion of data and the issue of mastering it have generated a need for specialised skills to exploit this high value resource. This trend has not only led to the profusion of new technologies (infrastructures, learning libraries..) and training in Data Science (more than 150 schools and universities in France provide it today) but has also led to enormous expectations on the companies side.
Even though they are convinced that good data exploitation provides a strong competitive advantage, many of these companies are disillusioned. Too often, the expectation of creating value around data turns out to be inversely proportional to the accuracy of the means used. To create value indeed Data Science must be part of the overall strategy of the company and of the reality of its organisation with the aim of serving its business.
In other words, the prior requirement definition of the company is a necessary condition for the implementation of relevant and efficient models. These needs range from the monitoring of the company to the management of operations, the evolution of the offer (product, pricing) to the optimisation of customer segmentation. The adoption of business intelligence platforms will therefore meet the need for management by optimising decision-making and operational efficiency. In contrast, the statistical and mathematical research work on data will be part of a longer period, for example to drive product evolution.
The missions of the data scientist will be defined with sufficient precision to serve the expectations of companies as soon as the needs have been well identified. Besides, it will attract the candidates who do not like vagueness! As the term “data scientist” is both new and broad, it is better to take the time to clearly define what is expected of them. Too many job opportunities are drowning in a confusing mix of tasks or skills related to Data Science without specifying the purpose (the “why”) and the missions (the “how”) of the job on offer. In this sense, Lyft, the leading rival of Uber worldwide, clarified the semantics of the positions related to Data Science within its organisation by distinguishing:
on one side, the analysts who extract information from the data, monitor the health of the enterprise and contribute to better decision-making;
on the other side, the scientists who build mathematical models and algorithms and feed the main components of the product.
While defining clearly the assignments is an efficient way to attract applicants, there is no doubt about the responsibility of job sites, headhunters and freelance platforms in the role of intermediation between the company and applicants. Facing the emergence of these jobs, new actors like Welcome To The Jungle, Malt or Comet are playing a remarkable part.
Once recruited, data scientists must be integrated in an organisation which makes best use of their skills and know-how. If each company uses its own model, three major organisational templates have already shown their effectiveness.
In the standalone template, applied in companies such as LinkedIn and Facebook, data scientists operate in an autonomous unit alongside technical teams. As they are detached from daily operations, data scientists have the opportunity to invest in long-term research and development projects even if there is a significant risk of disconnection with operational teams. This organisational model can only work within companies with a strong technological culture, where the operational entities are already completely impregnated with the data dimension.
In the integrated template, data scientists are affiliated with teams which need support and are part of the operational business. On the other hand, long-term projects are more difficult to conduct and team managers must understand the issues involved.
Finally, in an embedded organisation, which is certainly the most widespread model, data scientist teams help the different units of the company (commercial, operational, marketing…) to identify, understand and solve their various issues. They therefore take actions on concrete projects with high added value, but lose autonomy.
In the end, it is probably by demystifying and integrating Data Science into a strategic and managerial vision that companies will take full advantage of it. However, dealing with an explosion in demand for data scientists, organisations face the challenge of attracting and retaining these talents, which are rather prone to nomadism. In 2017, Kaggle founder Anthony Goldbloom claimed that “59% of people working in the field say they spend 1–2 hours a week looking for a new job”.
