Home » Technical Topics » Data Science

Analytics Translators: Fact or Fiction?

It’s been two years since Mckinsey invented the term analytics translator, called it the ‘new must-have role’ and predicted we’d need around 5 million of them.

FIVE MILLION

That’s a bit more than the entire population of Los Angeles, but just slightly less than all of Norway.

For the past ten years, we’ve struggled with the ambiguous title ‘data scientist’, then ‘citizen data scientist’. Now it’s ‘analytics translator’.

Although I’ve seen many ‘data scientists’ change their Linkedin titles to ‘analytics translator’, the problem remains that no one knows what ‘analytics translator’ really means. Mckinsey seems to have slipped this term into a Harvard Business Review article, and it has somehow taken root. What’s more, people seem truly excited by the term.

“we’ve struggled with the ambiguous ‘data scientist’, then ‘citizen data scientist’. Now it’s ‘analytics translator’”

A Bit of Background

When the University of Amsterdam asked me to begin training professionals in how to become ‘analytics translators’, I had to start formalizing the skill set required to be an ‘analytics translator’.

Since Mckinsey was responsible for inventing the term, I wanted to understand their thinking.

After a bit of digging, I found they were recycling content from a 2016 paper they’d published, which discussed a shortage of ‘business analysts’.

So it seemed that during the 14 months between publication of these two Mckinsey articles, the term ‘business analyst’ had evolved into ‘analytics translator’. They must have decided it was time to introduce a new job title, never mind that we still haven’t clearly defined what a ‘data scientist’ is (whereas half the planet is, was, or will soon be ‘data scientists’).

“the term ‘business analyst’ had evolved’ into ‘analytics translator’”

 

What Is An Analytics Translator?

Quite simply, then, an analytics translator is someone who can understand business requirements and ‘data science’ possibilities.  In this sense, ‘analytics translator’ is a skill set and not necessarily a role or a job title.  

This skill set is actually extremely important. In a recent survey by O’reilly, 47% of respondents indicated this as one of the biggest challenges holding back adoption of AI / ML.

“ 47% of respondents indicated this as one of the biggest

challenges holding back adoption of AI / ML “

Some companies may have roles dedicated to gather requirements, and in this case the role itself could rightly be titled ‘analytics translator’. It is not, however, an equivalent label for a ‘data scientist’, ‘machine learning engineer’, ‘statistician’ or ‘product owner’. At best, the term ‘analytics translator’ could substitute for the traditionally nebulous title ‘business analyst’.

Even if your company doesn’t use this job title, the skills related to the analytics translator concept are crucial, and you should make an effort to ensure that a large number of your staff are ‘fluent in analytics translator’

What Skills Does An Analytic Translator Need?

Facilitating the execution of data science projects within a business context generally requires an understanding of business goals and processes; a high-level understanding of analytic vocabulary, techniques, technologies, and processes; and the ability to communicate cross-functionally. I’ve grouped these into 3 categories and 9 subjects.  OK, maybe 11.

“Analytic translators should have specific skill sets”

Foundational Technical Understanding

  1. Basics of classical statistics (regression, exploratory data analysis, hypothesis testing, correlation, etc)
  2. Overview of common machine learning techniques (deep learning, SVMs, decision trees, adaptive boosting, clustering algorithms, etc)
  3. Overview of technologies commonly used (programming languages, database concepts, deployment tools (docker, cloud), etc)
  4. Understanding the life cycle of model building, training, deployment, and maintenance

Foundations for Collaboration and Communication

  1. Understanding of frameworks and tooling used by data science teams (scrum, kanban, Jira, git, etc)
  2. Stakeholder management: setting expectations, building trust, change management
  3. Techniques of communicating analytic results (as per Stephen Few, Cole Knaflic, etc)

Foundational Business Understanding

  1. Understanding the goals and priorities of the diverse horizontal and vertical elements of organizations in which data science is only one team or department
  2. Choosing the analytics projects most likely to deliver business value in the current economic and corporate environment

Regarding skills 1-4, a large number of data science teams are run by non-technical managers (4). These managers may not need to understand the technical details of team projects, but it is critical that they understand the difference between a high-risk, high-effort technique and a low-risk technique. They should understand the background terminology of project reports so they can focus on the elements that are unique to their business case. They should be able to see through any BS thrown at them.

Regarding skills 5-7, it’s important to understand that much of the common tooling and methodologies (such as scrum) in use today were developed over the past 20 years by and for software developers, rather than data scientists. Data science teams must adopt and adapt the elements which will make them most effective in their unique tasks. Likewise, visual design and communication skills require special considerations when the subject matter is quantitative and especially when the audience is non-technical. I’ve seen very few data scientists communicate well to non-technical audiences without special training.

I sometimes add 2 additional items to this list of skills, as part of a soft-skills training for data scientists, these are

  1. Working in a multi-cultural environment
  2. Dealing with office politics

Why Are Analytics Translator Skills So Important?

Analytics translator skills are crucial for two roles in particular: data science product owner and data science team manager. Without analytics translator skills, neither of these two roles will be able to bring to bear the full potential of data science within a business context. Analytics translator skills are also extremely valuable for data scientists (individuals specialized in advanced analytics), both to steer them in producing business value and in helping them work more effectively with business counterparts.

How Do We Fill Analytics Translator Roles?

I’d agree with the Mckinsey article that the best solution is to train existing staff, rather than hire into an analytics translator role. Existing staff will already have a deep, proprietary knowledge of your business and have already formed relationships with key stakeholders. To illustrate, consider an AI healthcare company here in Amsterdam, who employees licensed medical doctors as product owners for their data science teams. In such cases it’s especially clear that adding analytics translator skills to the hiring conditions is not reasonable, and that providing trainings is the appropriate solution.

“the best solution is to train existing staff,

 rather than hire into the role.”

Also, analytics translator skills are sufficiently general that most staff with a healthy dose of curiosity can learn them fairly quickly, and my experience is that a large number of experienced professionals are indeed eager to learn, given the opportunity. Many companies are setting up internal analytics translator trainings. For those without such trainings, part of the skill set can be learned through online courses, but there are relatively few open enrollment programs that teach the combination of business, communication and stakeholder management skills required. I do teach some of these skills at the University of Amsterdam’s Business School, and I often give in-house training for analytics translator skills. During these trainings I also have the privilege of hearing participants’ own thoughts on what an analytics translator could or should be, and how the concept differs from other roles.

Would love to hear the opinions of others here. I welcome your comments below.

Originally posted here