Operational Analytics Trailblazers

We recently sat down with Alex Schwarm Ph.D., Head of Data and Analytics at Arrive Logistics. Prior, Dr. Schwarm led data science, data engineering and analytics teams at Dun & Bradstreet and Applied Materials.

We talked with Dr. Schwarm about his experiences as a leader in the data science revolution at the heart of modern businesses. Check out our interview about successfully operationalizing analytics!

---

In your career, you've had the opportunity to see all sides of data and analytics teams, spanning data science, engineering, and analytics operations. Can you tell us a bit about your journey?

Dr. Alex Schwarm: I have had the unique opportunity to work in many very different areas: semiconductor manufacturing software, solar manufacturing software, manufacturing consulting services, digital advertising, sales acceleration software, and logistics. The exciting thing I have found is that all these industries dramatically benefit from software, data, and modeling. I have also been lucky to have worked in multiple roles: strategy, consulting, pre-sales, product, new product development, and engineering.

I have found, for all of these roles, that focusing on business results creates clarity and focus that is difficult to achieve in any other way. By combining software, data and models, and concentrating on delivering business results, we can provide outsized results from smaller teams with low investment.

With the push for data science, machine learning, and "AI," many industries now see the same potential. Unfortunately, many organizations fall in love with technology and forget that revenue and profitability are the oxygen that feeds the engine to drive innovation and excellence. Teams can accomplish great things by ruthlessly focusing on delivering business results while also pursuing technical excellence.

You currently have broad responsibilities across data science, data engineering and operational analytics. How do you bring those functions together? And how do you measure success?

Dr. Alex Schwarm: There is a natural alignment across these groups in that they are usually interconnected and interdependent. At Arrive, data engineering owns the data platform from a technology and infrastructure perspective, along with data pipelines and secure integrations. This makes data available to and from the data platform's internal and external integrations.

Analytics engineering leverages the data-platform-ecosystem technology stack, as well as the data integrations from data engineering, to deliver content within and manage the content of the data platform. This includes data definitions, guarantees of data quality, and consistency in support of all relevant stakeholders.

Lastly, data science leverages data assets within the data platform to develop and evaluate machine learning systems that support assisted or automated decision-making.

Machine learning engineers also play a crucial role in operationalizing and scaling those systems. ML engineers operationalize and scale data science implementations in collaboration with data engineering and data science.

Ideally, measuring team success is based on business impact. Specifically: Cost, profit, and revenue. These can be measured through time saved by automating manual processes or by identifying the impact of having inaccurate or "stale" data when decisions are made. In other cases, where the direct monetary lift isn't easily measurable, secondary business metrics that correlate with monetary metrics can be used – like achieving annual goals or other related quantitative measures.

With companies focusing on actionable data insights, what obstacles have you seen in operationalizing analytics inside companies?

Dr. Alex Schwarm: We were challenged with having one organization responsible for delivering and supporting reporting dashboards, while also modernizing our data platform technology, tooling, and processes. This challenge was compounded by our growing size and complexity (we exited last year with $1.6B/yr in revenue and are doubling annually).

New data engineering projects – spanning both analytics and data science – were constantly being added to the data engineering backlog, which exasperated this problem more. This huge number of projects overwhelmed the data engineering team, requiring very tough prioritization and leaving many projects indefinitely stuck as "lower priority." Data engineering became the bottleneck for our analytics teams that lacked the tools for self-sufficiency.

What advice would you give to help modern businesses successfully implement operational analytics?

Dr. Alex Schwarm: Firstly, organize so budget owners can make decisions for scaling their organization based on their needs.

Secondly, separate the strategic analytics engineering function from the more tactically-focused reporting analyst function. By separating these groups, stakeholders can control how they scale their reporting analyst team.

To accomplish this, we defined a new team structure with two groups. Analytics engineering is a technology- and process-oriented group that reports to the data org, and is focused on data management, leveraging best-in-class technology, and software engineering principles to produce reliable and quality data assets. Department analysts make up the other group and report to the stakeholder organizations to cover metrics, dashboards, and reporting.

This design allows stakeholders to choose how they budget for reporting and dashboards, while the data org budgets based on strategic data management needs. One challenge we encountered was that not all our stakeholder organizations had experience hiring and managing reporting analysts. To overcome this, we supported the hiring process by sourcing and screening candidates with the stakeholder organization having the final hiring decision. We also helped stakeholder organizations set up processes and working models with their analysts.

Finally, we found analytics automation platforms, like Savant, to be game changers by alleviating the data engineering bottleneck. These tools enable analytics teams to complete projects end-to-end without needing other teams. They also help drive cross-team collaboration, which is key in our organization model with analytics engineers and department analysts.