10 best practices for building a modern data stack

Companies are tightening the belts with one exception: data. 

The modern data stack shows no signs of a recession with Frank Slootman the CEO of Snowflake saying at their annual user conference they will not slow hiring plans. CNBC Tech Council executives noted how technology investments are business drivers, not cost centers. And Foundry’s annual cloud computing survey cites cloud analytics as a top growth driver. 

A digital economy is fueling this growth in the data sector. But for the modern data stack in particular, it is the agility and time to value that is inspiring technologists and business leaders alike. 

In this article, I’ll share ten best practices from early adopters of the modern data stack.

1. Align to your why

Sargento is a $1.5 billion cheese manufacturer based in Wisconsin. They have pioneered innovations in creating such things as packaged shredded cheese, cheese sticks and snack packs. Although Sargento has an innovative culture, the company’s data platform had not evolved much beyond operational reports and spreadsheet-based analysis. Demand for data was outpacing available resources, and the company wanted to speed insights and decisions. In making the case to modernize its data stack, the modernization journey was tied to the overall business strategy in terms of company growth and operational efficiency. People are core to the Sargento culture so ensuring business users – not only data analysts - could adopt a new platform was core to the strategy.    

SaaS brand management platform, Frontify, echoes a similar approach. The company has aggressive growth goals, and the small data team could not keep up with demand for new data. The legacy data platform was based on custom ETL scripts, my SQL, and Tableau dashboards. New requests took a month to fulfill.

2. Break bad habits and promote change

Centralized data teams built around an order-taker mindset need to shift to more agile work processes. From a technology perspective, too, the idea of only loading a subset of data, and aggregating tables may no longer be necessary in a cloud environment. Manik Gupta, Chief Analytics and Insights Officer at Bayer said, “The days where our data and analytics teams would prepare for 6-18 months are now gone. You have to run 2-6 week sprints to come up with new programs or new functions to maintain momentum.”

Changing the mentality of gut-feel decision-making in a company with long-time employees, based on decades of experience, requires trust building. An ah-ha-moment at Sargento came when a plant manager was able to identify an unexpected reason for plant down-time. The company uses  a “Cheese Block” newsletter to showcase new insights. Sargento uses gamification through pop quizzes to encourage usage, such as, “What were Kroger sales in 2019.” Correct answers get entered into a raffle, with winners receiving a prize (and bragging rights!)

Travis Lehn, Senior Manager shares Best Practices for Driving Change

3. Leverage best of breed platforms from across the ecosystem

Sargento conducted a competitive bake off of their modern data stack vendors alongside two other fullstack vendors, ultimately choosing Matillion, Snowflake, and ThoughtSpot. The modern data stack had better performance, user experience, and faster time to value. 

Similarly, global B2B tech company CarTrawler is dealing with rapid growth in data volumes in linking airlines and rental car companies to enable customers to book a seamless travel experience. CarTrawler modernized their tech stack on Airflow, Snowflake, ThoughtSpot.  

The data integration and cloud data platforms vary by customer based on their distinct requirements. For example, for integration, Frontify is leveraging FiveTran.  At JustEat Takeway and CNA Insurance, the data platform is based on Google Big Query.  Chick Fil A leverages Amazon Redshift.  

Across these customers, ThoughtSpot is the experience layer for the modern data stack.

4. Partner the analyst and the business user

Domain expertise is essential in understanding the heart of a business question. In rolling out ThoughtSpot, CarTrawler partnered the business analysts with business users to help focus on the right business questions. Organizations are increasingly embracing federated organizational models and squad concepts as data is truly democratized. 

5. Focus on business problems, not features and functions

Cartrawler offers a 90-minute enablement for onboarding users. The enablement is not positioned as, “come learn a tool” but rather “bring us your business questions.”  So the business users leave the enablement session solving a business problem. 

Similarly at Frontify, the team established an internal portal of content, promoting success stories, or ways to perform cohort analysis. The emphasis on learning a technology is secondary to understanding the business data and answering business questions. 

6. Identify what’s in it for users

CarTrawler focuses on how they will make the business users and analysts lives better, or “WIIFM,” what's in it for me. At Schneider Electric, the message has been to “free up your energy,” which is aligned to the company’s overall mission of providing sustainable energy.  Motivation will differ depending on someone’s role and current approach in interacting with data.

It’s important to recognize that state of the art keeps changing at an increasingly faster pace. Mastering new technologies in the modern data stack may mean letting go of legacy technologies that an analyst or data engineer was proud to be an expert in. There may be an emotional connection to these products. Acklowedge this emotional journey, create new role models and communities, and communicate how these changes will influence career progression.

7. Host a hackathon

At ThoughtSpot, the shared services (CORE)  team held a hackathon to identify use cases and replace what may have been spreadsheet-based analysis to ThoughtSpot. For example in focusing on the finance use case, our CFO is able to closely track top line growth, bookings, and free cash flow.  Our controller is able to see detailed travel expenses, reduce the aging of accounts receivables, and end manual spreadsheet analysis.

This hackathon has led to a number of new applications including employee retention, JIRA ticket tracking, diversity and inclusion liveboards.

8. Use live data in meetings

It takes an evidence-based culture to support Live Analytics in an executive meeting. It also requires a modern data stack such that data really is available at a business user's finger tips.  

At ThoughtSpot, board meetings, sales meetings, and finance meetings are run based on live data and a single version of the truth in seconds. This also requires organizations to break the bad habit of dead end PowerPoints, which can cost up to $40K for an analyst to prepare manually. Speaking at a recent modern analytics event, Vamsi G, the VP, Global Head of Finance at Western Union described, “Before ThoughtSpot, we used to run our weekly business executive meetings with slides, PowerPoint. Now we run everything through ThoughtSpot."

9. Proactively decommission legacy technology

It’s often easier to bring in a new tool than to decommission the old. CNA Insurance has described their investment in Google Big Query and ThoughtSpot, and the importance in decommissioning legacy data platforms. Otherwise, a data team gets dragged into maintaining technical debt, supporting multiple platforms, with too much time debating personal favorites rather than value delivered. Some organizations will allow legacy platforms to die a slow death. Others, will more aggressively remove access, then decommission as Bank of New York Mellon and Bagel Brands described in this episode of The Data Chief Live.

At American Express, CDO Pascale Hutz shares how they proactively decommision legacy technology and incentivize looking for opportunities. Teams are rewarded for saving costs and freeing up resources to enable innovation.

10. Measure and promote success

Your ultimate goal in deploying the modern data stack may be to drive top-line revenues or create operating efficiencies. And yet, there are a number of leading indicators that reflect your progress along the way. Measure these impacts and promote them continuously to ensure stakeholders stay engaged on the business value of the modern data stack. Examples of leading indicators:

  • Frontify boasts a 40% BI adoption rate and Schneider Electric has achieved 78%, compared to the industry average of 25%

  • More advanced requests now take only 30 minutes and the data team has shifted from being a report factory to higher value work.  

  • Likewise, at Afterpay time to deliver new data has gone from years and months to minutes. The modern data stack has enabled tighter collaboration between the data team and the line of business.

  • Roche has been able to decrease open requisitions in for data analysts, saving headcount within procurement analytics.

Embrace Innovation

Across the board, what these early adopters of the modern data stack share in common is rapid experimentation and a bias for action. The defining decade of data is here, and they are embracing innovation to lead their companies in leveraging data. 

With these ten best practices in hand, you can too. As part of that experimentation, start your free trial of ThoughtSpot today. See for yourself how natural language search and AI lets anyone create Live analytics, freeing you from dashboard backlogs. You can connect to a cloud data platform or load your own CSV file.