Cabinet de recrutement en Analytics et Data Science

THE CHALLENGES OF TRANSFORMING DATA AND THE TALENTS WHO STEER IT

Interview with Benoît Binachon, founder of Uman Partners, for the publication GUIDE du BIG DATA 2016/2017.

What led you to set up Uman Partners?

We founded Uman Partners around two important elements: first of all recruitment in Data, because we like the sector and because it is an interesting business speciality and a real need for companies; but also with the desire to introduce Data and the quantitative into our own work processes. Our approach in discerning candidates is therefore very “data-driven”. Our objective at Uman Partners is to help our clients to recruit in all senior posts linked to Big Data, Data Science and Machine Learning. We recruit very few traditional quantitative profiles but rather Senior Data Scientists, Data Science bosses, Chief Data Scientists, etc.

 

What kind of profiles are you most asked for?

We recruit three types of profiles: the first, and the most sought-after, is linked to pure Data Science, the core scientific jobs as well as the technical positions in Big Data and the architecture that often goes with it. There are people who have main strengths in Data Science with a minor in Big Data and people whose main strength is Big Data with a minor in Data Science. The primary need is to build the technical and scientific sides, so businesses recruit first of all a Chief Data Scientist and then the other people to develop the project.

The second most sought-after profile are users and intermediaries of Data Science, that is consultants that know both Data Science and business speak: project directors who will lead Data initiatives for the marketing, risk, fraud, consumer insights, manufacturing accounts.

These are people capable of understanding the needs of a business director and the issues involved.

We notice, on the part of manufacturers and businesses in general, a minimising of the importance of recruiting project managers with expertise in Data. Companies are not aware of the importance of raising awareness in project managers and bosses of Data consulting. But it is true that it is possible to shape a Data Science project manager on the condition that they are open to the basics. A person who has not had a career linked to Data Science, will be capable of doing this job if he or she is able to detect the use cases that can be resolved using Data Science via available tools, and that he or she acquires the culture of these tools. We find graduates of Science Po may be capable, while pure mathematicians may be oblivious.

The latest posts for which we have been approached are those requiring a great awareness of Data: Client Marketing Director, Consumer Insight Director, Industrial Performance Director, etc. Today, for example, a Customer Service Director is asked to manage a call centre using a very quantitative approach. A Consumer Insight Director in cosmetics might equally be asked to introduce prediction and prescription into the creation of a perfume or a beauty treatment cream because it is very expensive to carry out trials and develop products without having visibility on the success of a product to a targeted population. The Data provides a long view in business functions.These operatives are not being asked to be Data Scientists, but to understand what can be done with which tools, to listen and to use the whole technical ecosystem. With this family of jobs, we are seeing a real delay in France, especially in comparison with the USA, where there has been an increase in “data-driven” posts. There is naturally less need and also fewer candidates in these posts in France and so great difficulty in finding candidates who really have data-driven experience in their areas.

 

What Big Data integration can be carried out in the management functions of a company?

In France, we are seeing advancements more in the digital at director level. The digital is entering the executive board. This is not necessarily reductive since Data is a digital tool.

On the other hand, there are many digital initiatives that do not use Data, which is completely absurd: “Working with digital without Data, is like building a car without an engine” and it is fairly common to see a Digital Manager or Chief Digital Officer who is not at all aware of Data, which results in a lot of frustration and failure.

Equally, there are still very few Chief Data Officers in France, unlike in the USA. The role of Chief Data Officer is to have a good understanding of each executive board rôle: Marketing Director, Finance Director, R&D, Supply Chain , etc. He or she has to be a sponge to understand the whole Data ecosystem, the needs of the company and the whole offering that exists in Data to meet these needs.

Despite his or her key role, the important thing is not that a Chief Data Officer leads the project necessarily, but that Data is a priority for the executive board of the company.


 

Chief Data Officer: an emerging job in France

The rôle of CDO (Chief Data Officer) in France is very often limited to an IT or BI approach consisting of “making data available to the decision-makers”. This is a good start, but the role of CDO can be a lot more developed if his or her process begins with the expected business issues and insights, goes back to the available data, then the Data Science, and to the necessary algorithms and delivers action plans and tools in production to the decision-makers, rather than data alone.

A CDO in a large group, as a strategic interlocutor to the directors, will possess a strong Data Science and Big Data culture, good practice for managing change and quite an exhaustive vision of the “business” issues to be addressed. And he or she has a team!

The rôle given to the Chief Data Officer is truly to be at the heart of strategic reflexions, to have a real power of influence and means to implement. He or she should be a member of the executive board. In France, Chief Digital Officers not directly dealing with Data issues have been recruited. In general they are not well understood by the executive board and this is a source of great frustration. Data initiatives must be led at the highest level, this is fundamental.

The Chief Data Officer thus detects the issues to be addressed by Analytics for each business area and coordinates the implementation of the architecture that facilitates the industrialisation of the process, the collection of data through to the delivery of recommendations to directors. His or her mission is to implement and then manage the analytics “factory” that will best exploit the available data.

The Chief Data Officer must constantly monitor the needs expressed by the executive committee to find the tools that are most appropriate, but he or she must also be capable of delivering a need expressed by one of the executive board members, because he or she has detected a new technology in place that will meet it. He or she should always be questioning the different business areas, the Finance Director, Marketing Director… It’s a question of having a permanent circle of innovation between the external needs expressed and the internal needs that the CDO must express.

Many companies, right up to the very largest, possess data whose potential is absolutely extraordinary. But the dispersal of information and the lack of mobilisation means that often this value is not recognised. Yet, the technical means are now completely virtualisable, very few in-situ infrastructures are necessary, only brain power and the mobilising energy of a CDO and his or her team are essential.

And as mentioned above, to be successful in his or her mission, the CDO needs a team. The Data culture of companies is not yet developed enough to allow him or her to complete the mission alone. The CDO needs a Data Lab, an “army”, with a Chief Data Scientist, data scientists, project managers and Big Data architects. A CDO needs a laboratory to test projects and at the very least one person who has a good scientific level to read publications. It is not the role of the CDO, who is too senior. The CDO will build his or her Data Lab in order to lead the transformation.

Likewise, a Data Lab exclusively dedicated to marketing can be a real weakness. Doing Data Science in the area of marketing without considering the risk, for example, in the financial sector, can be sub-optimal. Data is a discipline that requires much transversality.

 


In the face of very attractive giants such as Facebook or Google, how do your clients remain interesting to Data Science and Big Data profiles, which are becoming rarer?

Google or Facebook attract very scientific profiles, specialising, for example, in the classification of indexing of videos. In the end, these are posts with quite a restricted scope, which suit pure geeks.

When we support our clients in the launch of a Data Lab for example, a project such as this will truly transform a company. For banks, telephone operators, insurers, it’s also about building a team with a very wide scope. The scientific issues are just as demanding but there are also human issues, notions of transformation and management.

When Axa launched its Data Innovation Lab, the whole chain, including the whole claim and all the claim data, was transformed, right up to a teleoperator action, and to a client advisor in a call centre. In the same way, for a telecoms operator, one of the use cases is to display on the teleoperator’s screen a very precise instruction that prescribes an action to be applied to the client who is calling. This involves upstream Data Science, and legalities on the exploitable data, IT and trade unions probably, because the ways of working will have to change, training will be needed, due to the impact on the telephone agents, who will obey the algorithm’s instructions.

These two cases show that large groups remain attractive because change management has a wide scope and the human dimension is very interesting for a Data Scientist.

On the other hand, a real handicap for French companies lies in salary pressure exerted by Google, Facebook or even GE (in Paris) who offer employees salaries that are far higher than the majority of French companies. Likewise, thanks to the attractiveness of the salaries offered, American companies are drawing to the USA more and more French Data Scientists, whose technical skills are among the highest in the world. To avoid a brain drain of our talent pool in Data Science, and to continue to attract talent, French companies must review their salary scales.

What new competencies are now requested for a Big Data and Data Science specialist?

In this world, these are profiles that are still rare… despite the fact that we are training many people in Data at the Polytechnique, Supéléc, Centrale and at our universities. These are elitist positions that produce few talents because they require many creative skills and the ability to abstract, qualities that are innate and selective.

The required competencies in this area are of two kinds. New technical competencies are requested, linked with languages, some of which are new, and mathematical formalisms. Data Science is above all made up of mathematicians, people who understand the concepts but who are also capable of scientific impartiality. Developing code (Data Science algorithms) without a scientific safeguard can prove dangerous.

But now talents are required also to have a real capacity for understanding the business and participating in its transformation. These are the competencies that relate to leadership and include, depending on the posts and levels:

– Knowing how to influence a director;

– Knowing how to lead innovation or detect innovations that may be applied on a micro or macro scale;

– Leading and developing teams — knowing how to manage them with methods inspired by Google and others — knowing how to make them shine, connecting them to research, facilitating discussion in meet ups, as well as helping them to excel on a daily basis by enabling them to develop from a strictly scientific point of view;

– Knowing how to rapidly acquire new approaches and technologies;

– Having a customer focus and speaking in simple language about abstract concepts that have very concrete applications in business;

– Listening….

Finding talent in the area of Data Science is like finding the impossible. You  might imagine the super Data Scientist as an introvert who communicates little, whereas in fact communication is a competency of prime importance in this job. Data Science only makes sense when the talent is able to communicate.

For example, in the case of fraud (of an electronic meter, insurance, etc.), the false positive (the fact of detecting someone as a fraudster when he is not) is very detrimental and can have a very negative business impact on the company. The challenge for a Data Scientist is therefore to minimise the number of these and to understand the issues. It is the opposite for bank card fraud when the client is the victim. In this case, the false positive is wasted energy for the bank, but reassuring for the client who feels well protected.

For those functions that use Data, such as Marketing, one of the most important competencies is to know how to lead innovation, detect innovation, evaluate one’s market value, project it and adapt technology in a business which may not be technical.

 

The Big Data gurus talk about the death of false Data Scientists and the birth of converted Data Scientists. Have you observed these phenomena?

I don’t believe that there are false Data Scientists. There are the disciplines of statistics, actuarials, econometrics… and then there is Data Science and Big Data. These are two quite distinct worlds (but which have common areas), meeting different needs. And there are quite simply people less appealing to the world of Data Science among statisticians, econometrics, actuaries, and profiles in financial mathematics, and others who convert. And others who have always been in Data Science (previously called Artificial Intelligence, born in the 1970s).

I think that false Data Scientists (although it is quite insulting to call them “false”) would be people who are excellent users of existing tool boxes, champions of Kaggle competitions, but they do not necessarily have the sufficient scientific impartiality. They are, however, necessary to the Data Science ecosystem. The Data Scientist, the Developer, the Data Engineer and the Big Data Architect are all complementary.

There are however, converted Data Scientists, and various routes can lead to Data Science, with the commonality of always being scientists with a strong capacity for abstraction, among which are:

– University professors (in mathematics and/or Artificial Intelligence) who we find as consultants in large group Data Labs;

– Specialists in signal processing (radar, imagery, telecoms);

– Physicists who have recently completed theses in particle physics, astrophysics or another scientific area and have processed such volumes of data as is used in Data Science and Big Data;

– Specialists in particle physics or high frequency quantitative finance, who have remained mathematics specialists. The first Data Scientists were actually traders;

– Computer developers who have entered the sector via coding.

Do you yourself use a tool to predict success of the candidates (EBI – Evidence Based Interviewer), and if so, how does it work?

The human mind has a tendency to provoke cloning. Firstly, for psychological reasons, and then also because of a lack of capacity to cross-reference and sort data in a relevant way to really predict success. In the Human Resource sector, we recruit mainly based on intuition. And contrary to received ideas, Data Science applied to HR can open new opportunities.

For over 40 years, the Americans in particular have been designing tools for predicting the success of an individual in his or her future job, based on his or her leadership competencies and the characteristics of the company and the post.

The algorithms allow us to determine scenarios for the candidate’s success. In volume and mass, the algorithms will update thousands of different configurations. They will detect niche people who, for common reasons, are successful, reasons that will not necessarily have been noticed by the human eye. The detection of weak signals of these niche successful people will facilite the aggregation of all those with the same characteristics and create winning scenarios for them too. If we detect 15 people who are successful for the same reasons, for example due to a particular training path, we will be able to apply this success factor to 200,000 matching profiles. We are creating clones, but with the aim of creating career paths that could not be generated by chance: Data Science is a tool for making chance fruitful.

On the other hand, a good analytic approach for supporting HR should eliminate database clones. We know that high-level, qualified employees are often correlated. The algorithms therefore delete this correlation in the interests of more complex correlations, for example, those which explain why certain people succeed with little known qualifications. The weak signals for success are detected by removing the upper levels of overwhelming factors.

However, it is the human factor that remains of greatest importance in Human Resources: creative recruiters who trust their intuition will still be essential because Data Science will never detect something that has never happened. It will detect weak signals but not signals that do not exist. The human being, using his or her creativity and not always being aware of why and how, can create the first cases that the machine will then explain and help to reproduce.

We have therefore created (using existing concepts) an approach that is based on the identification of competencies (called “leadership competencies” in the USA and “behavioural competencies” in France) that are predictive of success in a post. We have reconstructed a library of 46 competencies (creativity, capacity to influence, capacity to learn, autonomy, etc.) that we have correlated with a database of individuals containing information about their careers, leadership competencies and success in each one of their posts.

This is a tool that is far from being informal, but structures the face-to-face interview and validates the leadership competencies for success in the post. The candidates excel with this method, as it allows the recruiter to ask the right questions. Rarely will a scientist describe naturally the way in which he or she will convince a director if the question is not asked of him or her. This is therefore a more factual, more human-centred approach, which takes an interest in qualities that are truly important for a post. This is the approach of “A Compétences Egales” (With the same skills), an organisation of which we are members. With the same skills necessary for a post, we give equal chances to people who might be introverts for example.

This is an additional, data-driven tool, which does not make our company a recruitment machine, but, on the contrary, allows us to recruit having seen more clearly the aspirations and capacities of our candidates.

 

Benoît Binachon
Uman Partners
Managing Director – Executive Search – Smart / Big Data Analytics

Download the Big Data Guide 2016/2017 and read the interview.

Share: