Michael Tambe is a Manager at LinkedIn, covering Business Operations and Analytics. He is a Data Driven Intrepreneur who uses Big Data, rigorous analyses, and human narrative to identify new business opportunities. In his work he prides himself on three key things: Business Analytics, Operational Depth and End-to-end ownership. Before LinkedIn, Michael was a Senior Associate at McKinsey. He holds a PhD from MIT in Materials Science and Engineering.
I often ask myself “How can I better utilize data?” After all, I have access to terabytes of data. Who knows what I could achieve if I unlocked its secrets? Make better business decisions? Build a culture of measurable meritocracy? The possibilities of being data-driven seem endless.
While I was excited to be data driven, I did not know what that entailed. Asking myself questions like “Which decisions should I look to data to answer?” never spurred greater insight in me. In theory I should use data to make all decisions, but in practice I know that’s impossible. In fact, I nearly wrote off Big Data as hype until I decided to approach the problem from a different angle
(Dilbert is the intellectual property of Scott Adams)
I define being data-driven as being able to manage data and systems like I would manage a person working for me – I figure out what I want it to do and help it develop the skills needed to do the job. Let me explain
I figure out what I want it to do
I started by asking which tasks are better entrusted to computers than humans. I see the role of a manager as building a team to run the business. My definition of team was limited to humans, so I focused on identifying the right people for every task. In today’s world, a mix of humans and computers run the business. As a result I redefined my role as first delegating work among man and machine, then managing each accordingly.
This forced me to think about what computers do better than humans. Looking back I’ve only seen computers do three things consistently better than humans – track human (customer or employee) activities, do mathematics, and find correlations between things I didn’t know were interrelated. So I asked myself three questions
- “What activities do I want to track?” (i.e., If I could look over the shoulder of my employees and customers, what would I take notes on?)
- “What math do I repeatedly do?” (e.g., forecasts, KPIs, scenario analyses)
- “Which metrics do I not know how to drive?” (e.g., To better drive marketing email open rate, I’d look for correlations with attributes like length of the email subject)
These questions spurred greater insight in me and helped me build a to-do list for my data. I then prioritized the list in the order above, because, as the New York Times recently reported, one can’t do whiz-bang math or predictive modeling without proper data. This brings me to my next point
I help it develop the skills needed to do the job
I found my data isn’t capable of doing its job on day one just like any other employee I hire or manage. Just because our company stores data does not mean it’s the data I need to make a decision. In fact, I’ve found that unless someone is already using a piece of data to drive a business process (e.g., quarterly earnings)it’s a fair bet that the data will be incomplete or measuring something different than what I think it is measuring.
I work hand-in-hand with my data until it has built my trust. Even the best new employee lacks context in my business, so I need to work hand-in-hand with her until I’m confident she knows how to do her job. For data this means understanding precisely how the data is generated, then working through a few examples to ensure it’s done the way I need it. For example churn is tracked by contracts that are not renewed. However contracts may also not get renewed if a customer consolidates spend on multiple contracts to one contract. Only by working through these examples hand-in-hand with my data was I able to distinguish that and get a help get to data I trust
Getting the data right takes time. I manage this by always seeking alignment between data and human narrative. Even now, after I’ve worked to build a data set I trust, if my data says churn is high, I ask an operational lead. If he disagrees with the data, we dig deeper and figure out the truth. I neither take his word for it nor do I trust data over his judgment. This is exactly what I would do for two members of my team telling me conflicting information.
Bottom line is that I didn’t figure out how to utilize my data until I changed my mindset to think of data and systems as another member of my team. So, the next time you find yourself asking “How can I better utilize data?”, try thinking of data and systems as an employee you manage and ask “What role do I want data to play in my team?” and “How can I better manage its performance?”
Does this new framing lead you to new insights?
This is the first part in a series on being data driven, stay tuned for future posts
(Image Credit: Howard Lake)