Data can be a key driver for business growth as it enables the integration of business processes, helps measure success, and ensures that an organization works like well-oiled machinery. In order for businesses to use the power of machine learning models and build AI capabilities, a comprehensive data strategy is necessary, which goes beyond data transformation and operational reporting. Organizations must create data endpoints that facilitate the scalable sharing of data both internally and externally.
In the present day, people generate a larger volume of data than ever before. Determining what data is essential and matters to businesses is very challenging. It requires a significant investment from both data teams and business subject matter experts. This is because every business team owns data but not everyone fully understands it. Additionally, it’s challenging that no one comprehends why it takes so long to obtain cleaner data. What goes behind the scenes?
Jasmeet and Data Strategy
Singh has implemented winning data strategies for organizations with a varied set of industries, org structures, and different needs. Prior to joining HackerRank, he started his career in IT. He worked with medical device manufacturer Lake Region Medical based in Chaska, Minnesota. In this sort of environment, a small error in data processing could endanger patients’ lives. This could also cause lengthy FDA audits.
“At Lake Region Medical we had to ensure the highest quality standards for compliance,” Singh Says. “But without modern data governance and quality tools available today, one had to rely on creating data definitions in Excel, writing custom SQL scripts for data quality checks, and using Windows scheduler to get email alerts for failures.”
Data Strategy in Business Systems
The data strategy at the global restaurant chain On the Border must integrate multiple business systems. This includes point of sale, online ordering, 3rd party catering system, employee scheduling, attendance, etc.
“At the time most, people would call and order,” Singh said. “There were no mobile apps where customers create their profile so coming up with personalization and segmentation to give the best customer experience meant a lot of data mining.”
Data Strategy and Analytics Teams
At Amazon, data processes were already mature. However, there were different challenges of scale, complexity, and org structure. Each business had its data and analytics team.
“Amazon operates in so many businesses, so it was like working for multiple companies at once. Each business had its own data and analytics team,” Singh says. “So it was shared ownership with data contracts needed to be in place, traditional systems and architectures could not scale so everything had to be on a distributed model including people.”
Data Strategy with Machine Learning & AI.
HUSCO International is a global manufacturing company based in Wisconsin. They tasked Jasmeet with leading digital transformation and industry 4.0 efforts using data, machine learning, and AI.
“HUSCO was using cutting edge robotics and automation technology to manufacture parts at scale, 90 percent of data was generated by sensors vs humans, it all needed to be ingested and processed in real-time to realize maximum value, data needed to be shared asset among suppliers, the manufacturer (HUSCO) and customers.”
Insights and Data Exports
Contrary to previous companies, HackerRank is a software-as-a-service company with an AI-first strategy. HackerRank caters the data strategy towards users at thousands of enterprise customers, in addition to internal business operation analytics. HackerRank operates with a POD-based model where each product area assigns a product owner and a self-sufficient engineering team.
“Insights and data exports needed to be near real-time, perform insanely fast and actionable as recruiters needed to make hiring decisions within a few minutes of conducting campus recruitment drives, it was not just HackerRank but some of the largest tech company’s reputation is at stake.”
Jasmeet’s mantra of winning data strategy is that there is no one-size-fits-all. Align your data strategy with the company’s overall strategy, organizational structure, and culture to ensure coherence and effectiveness. Often, data teams try to align the operating model with the architecture they pick, but Jasmeet suggests doing it the other way around. Pick the operating model first and then pick the architecture aligned with it. If you have distributed analytic teams like Amazon, pick the Distributed Data Mesh architecture. If most of the departments in the company are centralized, pick centralized data lake architecture, and pick hybrid if that makes more sense.
Jasmeet also recommends investing in data governance, data quality, and observability initiatives early on, as delaying these can sabotage the entire strategy. If users lose trust in the presented data, they stop using the reports. Therefore, Jasmeet recommends separating out the data architecture team focusing on deployment, configuration, and best practices, so data delivery teams can focus on development and meeting deadlines. He also believes that one important aspect of winning data strategy is to leverage self-service models so that precious data engineering resources deal less with ad hoc requests, allowing them to focus on big-ticket items.