How Cisco built a data-driven culture at scale

  • 28 March 2022
  • 0 replies

Userlevel 5

This content, written by Daniel Mintz, was initially posted in Looker Blog on Dec 10, 2020. The content is subject to limited support.

At JOIN 2020, I had the pleasure of talking with Jennifer Redmon, Chief Data & Analytics Evangelist at Cisco. We talked about what it means to be data evangelists, how to encourage a , and how to measure success. Our conversation started with a big picture perspective but zoomed all the way down to the nitty-gritty details and tactical how-tos of the role. I hope you’ll find her insights as valuable as I did.

Defining what it means to be a data evangelist

Jennifer describes the role of data evangelist as a combination of education, enablement, and development of technical platforms. In her words, it’s about asking: “What products, services, and platforms do we need to help people use high-quality data in their day-to-day roles?”

Beyond establishing the technical foundation, evangelism is about empowering everyone in the company — from the individual contributor level to the executive level, and from the technical user to the business user — to incorporate data in their day-to-day decision making.

In Jennifer’s view, the biggest factor in getting people to successfully use data is in reducing friction. “It’s not at all about increasing motivation,” she says. “Many psychologists will say that the brain is actually one big prediction machine just looking to incorporate new data sets.” In other words, we are intrinsically wired to want to use data in our decisions.

Though the desire is there, most people have not been trained in the skills of data science and data analysis, and they don’t have the time to figure out where the data they need lives and how to access it. So a big part of being a data evangelist is developing that education and putting the right tools in people’s hands. It’s our job to help people understand what is possible. There may be tools out there that are perfectly suited for people’s needs, unbeknownst to those that need them.

Common misconceptions about being data-driven

Being data-driven: goal or journey?

During the course of our conversation, Jennifer and I uncovered a number of common misconceptions about using data. One is that being data-driven is an end state. In reality, it’s a continuous journey. Jennifer pointed out that, in the third industrial revolution, certain mechanical and analog successes enabled us to “check a box,” so to say.

This fourth industrial revolution, however, is quite different: “when it comes to digital successes, we’re never going to have an end point.” Being data-driven is not a box that you can check and then move on from. It’s a journey that will continue as long as advancements in the field are developed.

That being said, there are certain foundational elements to this journey that cannot be skipped over. For example, high-quality data collection (preferably automated) is foundational. As Jennifer puts it, “You can’t jump to an advanced AI or ML algorithm unless you know that your data is clean. Otherwise, you’re going to be creating some really strange or unrealistic insights.”

Determinations vs. probabilities

This leads us to another misconception that people often have: that high-quality data will help us get to a more deterministic forecast. Unfortunately, there’s only so much that we can possibly account for — so we need to look at the world in a probabilistic, not deterministic, way. It’s natural for us humans to want certainty. That’s how we’re built. But we need to maintain a healthy amount of caution in our analyses because there will always be outcomes we can’t predict. In fact, knowing what we don’t know is often as important as knowing what we do.

Replacing vs. adding to established decision-making methods

A third misconception people have is that using data to drive decisions is a full paradigm shift that replaces the old ways of making decisions. However, we should be careful to not throw the baby out with the bathwater. As data analysts and data scientists, we understand the data; but that doesn’t mean that we necessarily understand what its purpose is, or what the business process that generated it is, or what the person receiving it is actually trying to do.

Jennifer cautions that, “As we pivot from using tribal or experiential data to structured or semi-structured data, we risk losing the domain knowledge or depth of understanding that comes with it., Without sufficient context, we can actually make poorer decisions when we think we’re being data-driven.” This underscores that being data-driven is not so much about totally replacing the old ways of making decisions, but adding to them in the mix to make them better.

Specific tactics for encouraging the use of data in your company

We both find that sometimes people approach using data with a data-first mindset, rather than a question-first mindset. In other words, they look at what data sets they can use first, rather than starting with the question, “How do we solve this problem? What is the data we need to marshal to make an impact?” The appropriate mindset is to come at it with a very clear motivation that’s aimed at having impact.

With this context in mind, Jennifer shared great tips for nurturing a data-driven company culture.

1. Have clear key performance indicators (KPIs) and benchmarks

What are you trying to achieve, and how do you know that you’re actually achieving it? What’s the main thing that matters at the end of the day?

To illustrate this concept, Jennifer brought up the example of an initiative she is working on that aims to reduce suicides. The World Health Organization has identified “reporting on suicide” (learn more at as one of 7 priority areas. They ask reporters to avoid the term “commit,” avoid blame, and avoid talking about the method. In this example, the main benchmark that matters is whether the guidelines for suicide reporting are being followed. What doesn’t matter is the number of visits to the website or the number of application programming interface (API) calls.

2. Identify a high performer and replicate their methods

There are jobs that have historically relied on people’s own intuition, experiences, and domain knowledge rather than on data. The best way to figure out what it means to be data-driven in that context is to find somebody successful in that role who is using data, find out what they’re doing, and then teach that to the rest of the team.

3. Use Jennifer’s framework for measuring how data-driven people are

Her model is based on a quadrant.

Along the left side we have data enablement, meaning the tools, the , and the culture — all the extra things on top of domain knowledge — that are needed to be successful.

Along the bottom we have data IQ, meaning the data skills one has divided by the data skills one needs to succeed at a particular role, task, or job. Starting in the lower left quadrant, we have the “data illiterate.” These folks have both low data IQ and low data enablement.

Opposite them in the upper right quadrant are the “data-driven” people. Reiterating the point made earlier, this is not an ideal, end-state checkbox that is ever going to be “done.” What it means to be data-driven evolves over time, and someone who at one time would be considered “data-driven” might find themselves in one of the other three quadrants very quickly when the market shifts.

The upper left quadrant describes people that Jennifer refers to as the “data enthusiasts.” These people may not understand how the technology works, but they want to learn and are excited about it. They’re great change agents.

In the lower right quadrant are the “siloed high performers.” These people have high data IQ but low enablement, and among them is where Jennifer spends most of her time. She says that these people can actually do a lot more than the tools, platforms, and culture currently allow them to do. She recommends really listening to this group and diving deep to understand what they think is possible. Ask these people, “What could we achieve if we gave you X, Y, or Z?”

4. A/B test your approach

Jennifer says if you want to see if a given approach or model is working, A/B test your users. Whether they are internal employees or customers, provide whatever new data-driven approach you’re trying out to maybe 50% of the population. Give the other 50% a placebo, making sure they think they’re doing the actual approach. Then see if it makes a difference.

Personally, one of the ways I evaluate if my data-enablement suggestions are working or not is from the reaction I get from people. If I give someone a workbook or new dashboard, and their response is, “Oh, thank you. This is great.” I know that’s their way of saying, “I am never going to use this.”

Positive indicators that my efforts will be used are when someone’s response is along the lines of, “Oh, where did you get that number?” or, “Can I slice this differently?” or, “These tiles seem great, but can I change this one out?” or, “Is there any way I can get this delivered into Slack every Monday morning so I don’t have to load it up?” These kinds of responses tell me that they’re actually excited and are actually going to use the tool I’ve put in their hands. They’re engaged and asking for more.

As data evangelists, these are the kinds of responses that we’re looking for. Kudos to Jennifer for sharing so many useful tips from her data journey, and for reminding us that it’s a journey that keeps us learning new things every year.

For more expert tips on fostering a data culture at your company, make sure to check out and .

0 replies

Be the first to reply!