How to develop & deliver better analytics with Chris Bergh
14 February, 2021
Chris Bergh is the founder and CEO of DataKitchen, a DataOps platform that simplifies data toolchains, environments, and teams. Chris started his career at MIT's Lincoln Laboratory, NASA and Microsoft, working across a range of engineering and leadership roles leading analytical teams. He experienced the barriers in embracing data opportunities first-hand, and translated his insight into founding and advocating for the DataOps movement.
In this interview, we spoke about Chris' personal journey, the experiences that led him to start DataKitchen, the concept of DataOps and why companies should give it a try.
What is DataOps?
We started our discussion talking about DataOps - a term that Chris has coined. As an Analytics lead, Chris experienced the struggles of delivering business value through analytics first-hand. He talks about the lack of standardization and the need to unify tooling and skillsets within his team. These ultimately led him to think about better ways of developing and delivering analytics. Chris designed a set of principles and grouped them under the umbrella of a DataOps manifesto, specifically:
- Individuals and interactions over processes and tools
- Working analytics over comprehensive documentation
- Customer collaboration over contract negotiation
- Experimentation, iteration, and feedback over extensive upfront design
- Cross-functional ownership of operations over siloed responsibilities
In our conversation, Chris highlighted his view that the focus of analytics should shift from the object to the process. "The what is not as important, it's how you put the pieces together with people and a set of interlocking processes that's key to fostering innovation."
What companies get wrong when building a data culture
Chris talks about the typical failure modes that companies hit when embracing data opportunities. He highlighted trust, cycle time, and relationships as the things that companies often get wrong.
Firstly, he highlights there is often a lack of trust in data from the people who are receiving analytical product. "Analytics teams often get beaten down by follow-up questions and mistrust". Errors can happen anywhere, from the data issues itself to the analytics, anywhere in the pipeline. Chris talks about a mindset shift is needed to establish more trust in data - "companies should embrace error rates and see them as an opportunity to improve and perfect their systems". His view is that error rates are opportunities for process automation, and even marginal improvements can lead to improved trust and ultimately, a healthier data culture.
Secondly, Chris mentioned the time it takes to put a model into a production, adjust and iterate as a major failure mode. "The longer it takes for the team to move ideas into a production, the worse".
Finally, we talked about the relationship between the people that build data products - data analysts, data scientists and IT enablement teams - and the business users. "This relationship is often challenging and there's a lot of finger pointing", Chris says. In his work at DataKitchen and through the DataOps advocacy, these are the key points Chris seeks to address.
"Love your error rates & fall forward. Error rates mean that there's an opportunity to improve and that you can shine a light on the issues that you have."
DataOps success stories
We asked Chris about the stories of companies that impress him. He spoke about the importance of fostering collaboration and making your employees happy, productive and empowered. "These things are key to building teams and companies that do good things and deliver value for your customers". Converesely, "you can tell when organisations don't have the vibe of success, and feel like they are burdened by the data", Chris says.
Data work can also be thought about as a process of influence. Companies are using the data not only to deliver the right answers, but to inform decision-making and trigger the right action. "The service mindset is key - it's less about the model and the tools you use, and more about who your customers are, and what their needs are". Chris also notes that sometimes, analytics professionals fall prey to focusing on the tools and methods, when that should not be the focus. "There are plenty tools at your disposal, the important thing is to use them effectively to build the conviction."
"If we can create influence and use data to make impactful decisions, that's the definition of success."
Focus on the how
In our discussion, Chris also argues that the analytics world is currently in somewhat of a boom and bust phase, moving fast forward to hit a wall shortly. "Algorithm are commodities, the applications of the algorithms is what you should be perfecting", he says.
He predicts data scientists might end up in a world saturated with tools, but should instead focus on the system as a whole and the efficiency of teams they're working within. "Data analytics is the team sport, and how you get team to work together in a productive way, that's what matters."
"Instead of focusing on what you do - focus on how you do it. It is an investment into the future - the what will change overtime, but the how is what you can control and what pertains. "
Building a data product
We asked Chris about how they use data at DataKitchen, and how he personally deploys the DataOps principles in his own teams.
At a company serving Data engineers and Data scientists, Chris highlighted the importance of talking to their users regularly to understand and build on the value of the product. He also shared his view, that his and his co-founders' hands-on experience with analytics and being deeply technical was crucial not only for building the product and hiring, but also for empathizing with users.
What is next?
Finally, we asked Chris to share an emerging trend in the data landscape that makes him excited. Here, he reflected on the idea of flipping the equation in data ownership, consumers reclaiming control over their data, and the vision of companies creating a distinction between "data we produce and the data we share". If the ownership of our digital footprint changes, what would that mean for the market dynamics, consumers and our economy?