Close

Thank you for your comments. Our team will moderate and please expect it to show up on the page within 72 hours. Thank you.

Leading Data Science Teams: What Executive Sponsors Should Know

The leadership of an executive sponsor is critical for data science teams. Because data science is technical and specialized, the details of these teams’ work can seem opaque or confusing to other departments. Executive sponsors help companies understand the unique needs of their data scientists, so that a productive working relationship can thrive.

Usually, the executive sponsor is a high-ranking leader within the company who interfaces between the data science team and the rest of the business, but they have many other responsibilities. For example, the executive sponsor role could be held by the VP of Marketing, Chief Product Officer, or Chief Human Resources Officer. 

The executive sponsor passes on information to other leaders about how the data science team will help them achieve their goals. Competent executive sponsors are accountable, both to their team and to the executives who assess their work. They set strategic team objectives that align with company goals, then advocate for the budget and resources needed to realize them. 

Best practices for data science executive sponsors

The executive sponsor has a demanding role, and unfortunately, many issues faced by corporate data science teams arise at the leadership level. Too often, executive sponsors are unprepared or lack the knowledge to support their team. In practice, this can look like sharing information about the team’s activities without demonstrating how they benefit the company or, when the data science team shares their needs, relaying them to the rest of the company without conveying the reasons behind them.

Here, I break down a few of the problems I’ve most commonly witnessed on data science teams, and how executive sponsors can be prepared to address them:

1. Communicate key business objectives

Too often, leadership doesn’t let data science teams in on their business objectives. Instead, they task them with narrower questions they feel are more relevant to their work. But for data scientists to help companies reach their goals, they must understand the problems they are working to address, and the knowledge they must gain to do so. 

Asking the right questions is crucial to extracting value and solutions from data. From an outside perspective, it’s not always obvious how data can help a company reach its goals. That’s why data science teams should understand your greater objectives—the big questions you are really looking to answer. They might not have quick or easy solutions, but they’ll develop their own lines of inquiry and come back with approaches you might not expect. 

When you communicate your company’s entire vision and all the obstacles that stand in your way, you enable the data science team to deliver the best possible return on your investment. By making teams aware of your business objectives, you give them a guiding principle around which to shape their research and make sure it produces value. 

“Executive sponsors help companies who hire data scientists understand their needs, so that a productive working relationship can thrive.”

2. Choose the right team  

The makeup of a data science team is another important factor in its success, but building out a flexible, agile team that can interpret and respond to business problems is more challenging than it might seem. 

Usually, the executive sponsor is responsible for hiring the lead data scientist. This lead scientist, in turn, will hire the rest of the team. As such, this first hire has significant power to shape the outcome of the entire team. Executive sponsors should approach this responsibility with care, as it can be especially challenging to know what traits to look for if they themselves don’t have a data science background.

Often, hiring managers with no background in data science themselves hire based on candidates’ impressive degrees, publications, and other credentials—even though these give surprisingly little insight into the candidate’s practical working style. 

One hiring strategy for executive sponsors is to administer a writing test. While this may seem irrelevant, the ability to relay complex ideas in a simple way indicates that a candidate has the communication skills to translate between their specialty and its practical applications. Another clue could lie in the applicant’s professional and academic history: Have they worked in a solitary manner, or do they have a history of collaboration?  

Strong communication and the ability to work with others, especially those from other disciplines  are ‘soft skills’ that are useful in a very wide range of situations. For data scientists, they imply an ability to understand business objectives and integrate them into their work. These are the kinds of data scientists executive sponsors want on their team.

3. Build a team with all kinds of diversity in mind  

Data science is a constantly changing field, but overall data scientists fall into two broad categories—generalists, and machine learning (ML) experts with a narrower specialization. 

While there are situations that require ML experts, generalist data scientists are usually the better choice as first hires to build a strong corporate team that delivers value. These are the scientists who can effectively apply their core competencies in statistics, machine learning, and programming to business objectives. While the expertise of ML specialists certainly has value, in most cases it’s this agility that leads to real insights and solutions. 

As your team grows, diversity is a strength that you should advocate for and support by hiring a mix of generalists and specialists. More importantly, the team’s leader should ideally draw from diverse academic backgrounds, rather than sticking to their own discipline. (Visier is also a great tool that helps you increase overall diversity and inclusion on all teams). 

4. Work with imperfect data

Many outside the data science field don’t realize that the dataset available to scientists is just as important, if not more so, than the techniques and algorithms used to analyze it. Working with limited or imperfect data is common. However, this can become an issue for executive sponsors when expectations for their team don’t match up with what their training data will allow. 

For example, a company’s leadership may want their team to develop a sophisticated AI system that predicts when employees will leave the company. But if there’s insufficient data to train that system on, such as if all data over three months old is protected by a privacy policy or data is only available on employees who voluntarily provided it, the system won’t produce the accurate results executives are looking for. 

In these situations, executive sponsors need to make sure that those working in other areas of the company understand what their team can realistically accomplish, and ensure they do not over-promise impossible results their team cannot deliver. 

This means staying in close communication with the data science team, and trusting them when they explain what is and is not possible with the available data. They may also have to advocate for their team to other employees, explaining why they should have access to more data, or why the data available isn’t applicable to the task at hand, perhaps because it’s inconsistent or “dirty.” 

5. Provide the team with good data access 

Equally as important as the data itself is how easily scientists can access it. Unfortunately, many scientists arrive on the job to realize no one has given much thought to the data pipeline, and they’re expected to fill in the gaps, playing the role of both scientist and data engineer. 

Infrastructure issues can cripple data science projects before they start. Unless the needed infrastructure is extremely simple, it can be very difficult to retrofit into an organization where it wasn’t considered. If the team is trying to build and then refine the pipeline as they work, their results will lack the consistency they need to be useful and reliable. This is why, ideally, the data pipeline is in place before the team begins work.

“When you communicate your company’s entire vision and all the obstacles that stand in your way, you enable the data science team to deliver the best possible return on your investment.”

There are many factors that create infrastructural issues, from policies restricting data to unintentional logistical barriers. But they often boil down to a neglect of the data science team’s needs, or a lack of preparation for their hiring. Often, this occurs if the team is a late addition to a well-established company, rather than being built into its operations from the start. 

If executive sponsors anticipate infrastructure issues they should prepare to advocate for the team to other leaders on these  challenges, and articulate key goals and timelines towards getting the team the data they need. Sponsors can also look into past data projects at the company they can emulate when creating an implementation strategy. Of course, teams working in people analytics can avoid many data pipeline issues by extracting data directly from a solution like Visier People.

Executive sponsors must ask the right questions

All these seemingly disparate issues connect back to good communication—between data science teams, those who lead them, and the companies who value their work. Enabling teams to succeed means asking them the right questions, but also listening to their needs, and truly taking them into account. 

Answering the right questions is a guiding principle behind Visier’s approach to analytics. Even more than finding the right answers, great questions are how leaders  define the right business strategies, take action, and deliver exceptional results

Author Photo
Miles Steininger |
Miles's passion for data and enabling analytics brought him to Visier. This was a change for him as he previously practiced intellectual property law, as a patent agent, in startups and multinational public companies in areas like machine learning, robotics, wearable technology, and quantum computing. Miles hikes, skis, and bikes occasionally, poorly, and slowly.