“How can we just have all the data in one place?” The answer to that simple, common question, recounted by a member of a treasury team in the throes of a transformation process designed to support their company’s rapid growth, has a relatively simple, predictable answer that can be difficult to achieve: build a data lake.
- The member shared lessons treasury has learned in building a data lake at a fall meeting of NeuGroup for Technology Advancement in San Francisco. Their presentation included detailed charts and tables explaining the approach as well as steps taken and planned to achieve the key objectives.
- “We are trying to build a solid foundation for the future and we wanted a centralized, single source of truth for all treasury data,” they said. “The goal is actionable insights, discoveries and observations that will scale and incorporate AI and machine learning models that offer predictive analysis.”
- Matt Thomas, a NeuGroup director of peer groups who led the session, observed that “member companies are evaluating data structures and systems to access actionable data that is as real-time as possible in an automated process.” He added, “Organizations will need to bring in outside experts or hire team members with implementation expertise. Technology is only one element; planning, assigning roles and continuity of updated processes are critical.”
Problems and goals. The motivation for the member’s project is expressed clearly in a statement listing problems that are widespread among NeuGroup members: “Treasury’s current operational and reporting processes are time-consuming and lack insights due to data availability and integrity issues.
- “The data requires manual processes which do not scale. Combining data from multiple sources makes building dashboards, reports and analysis challenging and complex.”
- In addition to building a data lake to address the problems, treasury’s goals include the creation of dashboards for 90% of use cases and deploying a chatbot next year, as well as improving forecasting “by increasing data quality and reducing manual work.”
- Implementation of the treasury data lake to address most reporting issues will result in saving the hours of nearly one full-time employee, much of it “soul-crushing work,” the member said. “They can then focus on being more innovative and more analytical.”
Planning, building and architecture. The treasury team did extensive research and “homework” in a process of planning before building, the member explained. “Instead of ‘let’s start building,’ we wanted to plan backwards to inventory all the treasury use cases, treasury data, the raw data inputs, how they’re processed, what’s the output? Once we inventory all that, then we will build forwards. Once the data is there, we can really just focus on reporting or analytics.”
- The member’s slides showed the current and future states of treasury’s system architecture. And while the flow of information in the future state is linear and fully automated, it remains complex and involves multiple systems. Those include the company’s treasury management system, Kyriba; its ERP, SAP S/4Hana; and tools used by treasury for forecasting.
- Those are some of the input sources that flow into the data lake within Snowflake, a data warehousing platform. “The team is currently working through challenges to ensure that there is quality data flowing into Snowflake from Kyriba,” according to the member’s presentation. That will allow treasury to ultimately improve cash forecasting accuracy to 95%+.
- The new system architecture makes use of middleware managed by the company’s IT department that holds all the treasury data before routing it to Kyriba or SAP for processing and then to Snowflake. “We believe in maintaining full ownership and access to our data, partnering with our vendors who align with our data autonomy principles,” the member explained.
- Treasury uses Power BI to create dashboards and reports to “analyze data and make decisions,” according to the presentation. These, among other uses, will allow treasury to view projected balances, identify funding and automate monthly and quarterly reporting.
Lessons learned. Treasury’s list of lessons learned included a gentle warning to make sure you hire carefully when embarking on a data lake project. “We didn’t have the right resources initially. We reassessed and brought in new resources to consult on the design and configuration of Snowflake,” the member said. Other lessons from the presentation:
- “Perform a gap analysis on people, processes and technology and ask for expert opinions—seek help from existing resources like banking partners.”
- “Storytelling to secure buy-in is important; bring non-technical stakeholders along on the journey, especially telling the story to the leadership team. Overcommunication is always better for achieving alignment among everyone.”
- “Data governance needs to be considered and documented throughout the project: templates, tools, data dictionary, data catalog, data policies, etc.”
