Building Better Analytics Teams

What makes a good analytics team?

I have had a number of opportunities to build and manage teams of domain experts in my career, and true to form — I have noticed some common patterns.

Rather than a “how-to”, I instead want to highlight the traits and scenarios I have seen make some teams succeed better than others.

The Better Teams Define Their Success Parameters

A touchy subject and oftentimes a source of criticism is the purpose of the analytics team. I have built teams within corporations and a common expression is: “avoid science experiments, give us actionable results.” Although the advice seems like sound “corporate-speak” — what I found is the pendulum actually swings in both directions. You actually need an intersection of at least 4 things:

  • Analytic purpose: Is it targeted business insights, or data-driven discovery?
  • Is the team’s composition and objectives aligned to the analytic purpose?
  • Business maturity curve: what is the most appropriate capability your company requires at the moment?
  • Is the team able to execute on its mandate?

The problem arises when one or more of these areas are out of sync with the rest, which results in sloppy results, lack of focus, and worse: disillusionment of management on the value of analytics and a backslide back to gut-feel decision-making.

Define The Box, To Look Out Of The Box

In some cases you will encounter teams that segment themselves by analytical process, or segmenting analytics teams into procedures: e.g. the data quality team, the reporting team, the analysis team, the forecasting team, the modeling team, etc.

Analytics is actually the aggregation of all these processes. While there is occasional argument for and against this “commoditized” approach, the more successful teams are able to holistically integrate various parts of the analytical process to drive outcomes, regardless of their place in the analytical process. In some cases it does help to have distinct teams handling different processes, but each team should be able to contribute and interact with other processes: i.e. the “reporters” can help drive analysis, the “analysts” help police data quality, the “forecasters” refer to reporting for drivers of their models, etc.

Some of the most interesting expressions I have heard uttered in the workplace are multiple versions of: “I don’t write commentary, I just produce reports” and “I just comment on the numbers, go ask the reporting team why the number looks funny.”

Quite tragic.

The Better Teams Hire Character, Then Train Skill

We are probably at the height of the trend (or hype) in “analytical occupations”. Roles like Data Scientist, Data Engineer, Chief Data Officer and Chief Digital Officer are now the hot jobs under recruitment and most job descriptions for these roles put a high emphasis on technical competencies and skillsets required from candidates. After going through repeated recruitment, hundreds of CVs and interviews, in my experience, the successful analysts are not so much a function of skill but of their attitudes and values.

A baseline of technical capacity is definitely required to handle the analytical tools and make sense of data trends and patterns — but the stellar analysts also possess the drive and ambition to see projects through, the courage to point out anomalies or insights, and the initiative to take ownership over their data (warts and all) to drive business outcomes. Analysts who are naturally collaborative — and willing to share effort and credit on tasks also end up achieving more than the solo jockeys, regardless of brilliance.

There’s also the self-deprecating attitude — despite the body of knowledge they will usually possess, all the high-performing analysts I’ve met are humble enough to admit they don’t know everything — and this drives them to a path of curiosity and lifelong learning. All of them are voracious consumers of information, new trends, technology, and happily interact with other analysts to swap techniques. Contrast this with the occasional “big-wig” complex of some experts and academics who let ego get in their way — some of whom I’ve seen crash and burn miserably in the workplace.

Executive Leader
Leaders create an encouraging but intense environment that requires people’s best work

An Expert Leader Is First And Foremost A Leader

A follow-up point to hiring is the characteristic of a good analytical leader — and the typical assumption is that the big data boss needs to be the smartest, most-academically-decorated genius in the room. Technical skill and academic qualification tend to help with tackling analytical scenarios, but successful analytical leaders are still distinguished by their soft skills. The best characteristics (including those I have learned myself, the hard way) I think can be summarized by the Multipliers framework by the Wiseman Group — because they are true for analytics leaders:

Good Analytical Leaders (Multipliers):

  • Attract talented people and maximize them
  • Create an encouraging but intense environment that requires people’s best work
  • Define success and areas for people to stretch themselves
  • Drive sound decisions through rigorous debate
  • Give people ownership of results and invests in their development

Bad Analytical Leaders (Diminishers):

  • Hoard resources and underutilize and lose talent to boredom
  • Suppress people’s creativity and thinking
  • Must always be the smartest person in the room
  • Make centralized decisions and disempowers
  • Micromanage everyone

The Better Teams Encourage the Qualitative As Much As the Quantitative

Analytics roles generally attract individuals from quantitative disciplines — computer sciences, economics, statistics, mathematics — and generally the ability to crunch numbers and formulae will not be in short supply. However, especially in a work environment, analysts will not just be called upon to process data, but also to present this in meaningful form to peers and management. This requires a set of disciplines closer to the arts and humanities: public speaking, 2D and 3D visualization, colors and representations, and written communication — all of which will usually come counter-intuitive to the quants.

In forming analytics teams, a healthy balance between science and humanities (reminiscent of the Apple philosophy) is needed — through deliberate recruitment of mixed backgrounds and complimentary training to fill out the gaps for each analyst. The successful teams I’ve handled had good mixes of Statistics and Accounting, Mathematics and Finance, Economics, Business Management, and Computer Science.

The Better Teams Never Let Technology Define Capability

Depending on where they were minted, many analysts tend to be rabidly loyal to a particular tool, or software, or framework. We have all seen the nth iteration of the “Is R better than SAS or SPSS” or “SQL vs. NoSQL” or “RDBMS vs. Hadoop” debates but the reality I have seen is technology is rarely a differentiator of results. Granted some forms of analysis are easier to perform on certain platforms than others — the results come from teams that are able to formulate a business question or problem in terms that technology (whatever available) can interpret, process, and consequently resolve.

When I moved from being a business user into IT consulting, I was appalled at the number of instances teams or companies I encountered acquired tools they rarely use, or use incorrectly — and since most commercially available analytical software and hardware do not come cheap, this can become an invisible sinkhole of profits for many organizations. In these cases it is the analytics teams more than IT that should drive the applicability of their current or desired technology toolkit to their business environment.

Analytics Maturity
Executive commitment to analytics drives maturity
Source: Davenport and Harris

The Better Teams Recognize And Manage The Politics Of Information

Over the years I remained in contact with many prior colleagues who worked in the teams I’ve set up and am always humbled to hear about their present successes and how they traced much of their learning to those formative years spent together working with me. What I’ve found is the key to longevity in an analytics career is the deceptively simple recognition of the saying: information is power.

Careers, companies, and strategies have been built and destroyed on the back of information. Oddly, most analytical teams prefer to be “apolitical” or “amoral” about information — they will tend to dwell on the “how” rather than the “why”. I think this is both a costly mistake and an opportunity loss. Analytics has the ability to transform organizations, and analytics teams should take ownership of that power and use it, but be conscious of the massive power on their hands.

I am not suggesting analytics teams become active and malicious power brokers (some do) but going back to the first trait above: in defining success parameters become a trusted advisor to all parties in the organization.

Final Thought: The Best Analytical Teams Cease To Exist

According to Davenport and Harris’ analytical maturity stages: true maturity in analytics happens when the analytical functions no longer exist as a centralized capability, but become a natural part of each organizational function. This implies that a truly successful analytical team actually ceases to exist independently — and effectively each team in the organization becomes an analytical team.

Or to put it another way, the truly successful teams in any organization are the ones that become truly analytical.

I like the sound of that.

Published by

Dominic Ligot

Founder | Trainer | Fintech | Analytics