Recently, Mary Shacklett, a colleague and contributor to the TechRepublic Big Data Analytics Blog, produced
a concise and accurate treatment of a long-standing debate in the IT industry
on centralization vs. decentralization. Notwithstanding my professional
disagreement with the philosophy that drove her thesis, Mary’s right on target
with the classic challenges that still haunt the leader who’s trying to
structurally align their data science team with the goals of the organization.
In brief, here’s what she surfaces:
- IT doesn’t fully
understand what the business needs
- The business doesn’t fully
grasp the depth of technology required to succeed
- Dotted line reporting
relationships don’t work
- Joint reporting
relationships don’t work
- There’s no universal right
way to structure the organization
Data science teams only add more complexity to an already
perplexing problem. Should they report to IT, as Mary suggests? Should they
report into the lines of business, as I would probably suggest. Or, should they
be on their own: perhaps assign a Chief Data Science Officer that reports to
the CEO or President (as if we need another CxO to talk about)?
The right answer depends on whether or not it works for your
organization (how’s that for a typical consultant’s response). More
specifically, the size and maturity of your organization has a lot to do with
how your data science team should be structured within your organization.
When you understand how organizations grow, you’ll be better
equipped to make a good decision on where your data science team should go. One
of my favorite articles
on organizational growth was published in the Harvard Business Review by Larry
E. Greiner, a professor of management and organization at USC’s Marshall
School of Business. Greiner asserts that there are five stages of organizational
evolution and maturity:
- Coordination, and
Bridging each stage of evolution are revolutions, or crises:
leadership, autonomy, control, and red tape, respectively. For instance,
nascent companies are driven by creativity until they face a crisis of
leadership; if they successfully navigate this crisis, they enter an
evolutionary phase driven by direction.
This is very important to understand as a leader, because
many who struggle with the misalignment of their data science team(s) are
either trying to model what they did in the past, or emulate a paragon that’s
beyond their current level of maturity. When I was consulting for PayPal, many of their decisions were
influenced by what Visa was doing. That’s a bad idea! Visa’s been around for 37
years and has about 65% market share. PayPal’s been around for about 14 years
and has – a lot less than 65% market share.
In this case, PayPal decided to separate their analytics
team from their IT team; however, they were facing a crisis of control. They
had over-delegated and now top management needed to rein things in. This is the
point at which they should have started bringing IT and analytics closer
together, so that top management could get better cohesion with the lines of
business and more direct visibility. But the bigger problem I see is when
companies grow too big.
Too Big to Flail
So decentralizing at the wrong time is a bad move, but so is
centralizing at the wrong time. This is the problem I saw when I consulted for
Visa. Visa wasn’t really dealing
with a crisis; however, they were approaching one – the crisis of red tape. The
privilege of experiencing a red tape crisis is only reserved for the larger
organizations – and it’s a tough one to cross. In my experience, it’s the
toughest, because the organization has already been successful with a
centralized model, which has had the opportunity to take deep roots and blossom
nicely – until the weight of bureaucracy starts pulling it down.
The only way to survive a red tape crisis is to
decentralize. Your data science team has probably lost touch with what the
business really needs, so your initial emphasis should be to bring that back.
Put data science teams in each of the major lines of business, while maintaining
a center of excellence that supports them. What’s important at this stage is
that you foster an environment where objectives are clear, but procedures are
flexible. Creativity, innovation, and self-organization must be reintroduced
without losing the best practices that got you where you are today. If you wait
until the crisis hits then it will be a painful transition – so don’t! You can
usually see this coming; start making the transition before the burden of
bureaucracy overwhelms the organization.
To centralize, or decentralize, that is the question. The
answer has troubled leaders for decades and you can never be assured of the
right structure until you see it work in your organization. That said, you can
improve your chances by understanding how organizations grow and successfully
navigate growth crises. To get through a crisis of control, you should
centralize. To get through a crisis of red tape, you should decentralize. In
all cases, leadership matters. If you’re struggling with where to place your
data science team, think about what your organization is dealing with today.
Then, make an educated decision, and move on. It’s time to enjoy your next
phase of evolution.