How to make friends and influence people in data.

Leadership strategies to get the best from your data team

Working as a quantitative analyst, I was once referred to by a client as the ‘IT team’. Most of my colleagues would consider the IT team as several network engineers and infrastructure specialists, who certainly could not care less about what we do with our data. What gave me a giggle was the fact that if our actual IT team had heard that comment, they too would be offended! The comment made me reflect on what makes a good data leader — not only because of the mischaracterisation I had experienced by someone senior and well versed in their business, but also because I imagine the opposite problem also exists. Here are some of the best qualities that I would look for in a leader of a data science team, and subsequently someone that can really influence without authority.

Stand For Quality.

Often, data leaders do not trust the quality of the data that they work with. Still, they produce reporting with the data that they are given and add in pages of disclaimers which their stakeholders barely read, until someone dares ask a question about the calculations. In fact, these questions are rarely even answered properly by the data team and are often met with a complaint of how long they took to construct in the first place. Author and consultancy president Thomas Redman (@thedatadoc1) recommends his measure: The Friday Afternoon Measurement. This involves printing the most recent 100 fundamental records that you have worked with, with 10 or 15 of your favourite features (things like customer records, sales records or asset records), and asking a data team on a Friday afternoon to identify inaccuracies within the data set. The surprisingly low number of perfect records shocks most companies who participate.

Standing for quality means taking steps to correct these records, and effectively creating a miniature data lake where this information is accessible and used by the business without pages of disclaimer in accompaniment.

Active Root Cause Analysis.

If our managers were to raise a question about numbers which we produce, we are called to run around in panic and frenzy, until the answer is something technical and outside of our control. “Because the data in THIS table isn’t great” comes to mind as a typical excuse my colleagues and I have used before. What this might have achieved is some aversion of tragedy, but more likely only delayed it. We can all agree that this may turn into a vicious cycle of major incidents being raised in lieu of only minor issues months earlier.

As seasoned professionals in data science, we can also agree that it is rarely just the quality of data in some table which is the issue. Had we scratched just a little further we may have found that there was either poor configuration, or worse, a risk culture issue. It takes courage to push and lead an improvement or transformation project to eliminate the risk and address the root cause — but a good leader will always do it.

Separate The Management Of Data And Technology.

It is no longer feasible to just throw any new data from the advent of new products or services with the old, combined, IT team/data warehouse. Doing so without proper strategies can lead to disconnect in business strategies, a lack of clarity in responsibility, and frustration from all parties involved. Understandably, the IT team have the right skills to monitor, maintain, and oversee flow of the new assets. However, if management of the data and management of the technology blur their boundaries, then suddenly those that are the main users of the data are not enabled to report to the consumers of the data. Bottlenecks in data and technology are predominantly caused by exactly this — the lack of ownership by people who care the most, delegated to those who ration their care enough across their entire portfolio. If data assets are owned by data analysts, then resourcing will be appropriately allocated across the business and the network engineers/software engineers will be able to answer technical questions as often as they should come up (rarely!), while focusing on maintaining the infrastructure instead of the data itself.

Good leaders will push to include more data assets into their team’s ownership, whilst simultaneously aligning to my first and second points above. This way, their stakeholders will have absolute transparency on data quality, and be aware of any issues at the infrastructure level.

Hire Smarter People Than Yourself.

A good leader in any data science team will fill a business mandate to hire data scientists who are capable of really understanding and manipulating the data. Additionally, they will lean on some business people who are able to frame the problems they intend to solve, or the solutions they intend to present. To support them, they require a capable IT team to maintain sufficient infrastructure for their services to run, and subject matter experts to ensure nuances are exploited in their formulation. These people must also effectively work together in the same team to make sure every project links back to the business objectives.

Rather than discussing what could happen when these people are not included in a team, we can analyse how we can attempt to combine some of the skills listed above. You are likely reading this article and working in a team where there is a heavy weighting of employees in the first criteria, and little to no employees who possess skills in the other areas. You are also likely to have met just a handful of people in your travels around conferences and industry groups who possess some serious data wrangling capability and speak well to business objectives. I have spoken about these people before, called “soft quants”, because they exist in droves around different industry groups not just focused in the data space. The key to hiring well is striking a balance between your business capability, “soft quant” capability, and technical capability — how many business people do you need to understand the data? How many data scientists do you need to understand the infrastructure? How many business people do you need to influence your stakeholder group? Point being, a team of only data scientists leaves much to be desired from what is within our control to deliver on a business objective.

References:

“Top-Down Leadership for Data: Seven Ways to Get Started”, 2020, MIT Sloan Management Review (T Redman),

<https://sloanreview.mit.edu/article/top-down-leadership-for-data-seven-ways-to-get-started/>

“What’s your data strategy?”, 2017, Harvard Business Review (T Davenport, L DalleMule),

<https://hbr.org/2017/05/whats-your-data-strategy>

“Top 10 attributes of a true data leader”, 2017, Information Age (B Rossi),

<https://www.information-age.com/top-10-attributes-true-data-leader-123466684/ >