How Teachable Used Metaplane and Sigma to Achieve Their Data Quality and Visibility Goals
Without Metaplane and Sigma, we’d need twice as many engineers and analysts to try and keep up with data quality issues and the data analysis that the rest of the company requires.
Teachable is a platform that enables creators to monetize their knowledge by developing and selling online content, courses, and coaching. The company is highly data-oriented—in fact, a full 70% of the company consumes important company metrics on any given day. From customer operations to support, from product to finance, every department at Teachable uses data from the warehouse to monitor product usage, support ticket SLAs, weekly sales, SOX audits—and that’s just the start.
William, Teachable’s Senior Business Intelligence Manager, leads a team of five data engineers and data analysts. His team is charged with providing a single source of truth for company data and making it easy for the rest of the company to use that data.
Before using Metaplane and Sigma, achieving that mission involved manual workarounds and enormous investments of time and energy. The team lacked the tools to understand the quality of their data in real time, and their BI software didn’t allow less technical employees to explore the data themselves. With those goals in mind, William decided to implement Metaplane and Sigma.
Challenges: A lack of trust and access
Before Teachable adopted Metaplane, the data team had a common problem: their main source of data anomaly detection was their business stakeholders. When tables didn’t update as expected, or rows were dropped from a table, William’s team frequently didn’t find out about it until an end user sent them a message wondering if the data was stale, or worse—downright wrong.
Each uncertain Slack message degraded trust in the overall data a little bit more. And with such a large surface area for their data environment, Teachable’s data team lacked the tools to be able to quickly and confidently explain when issues started, what was impacted, and why they occurred.
William had set up some monitoring in Airflow to ensure that he’d know when jobs failed. But for more complex use cases, like building aggregate monitors for data added within the last few days, would have taken significant work to build and maintain manually. William knew that he’d need a more robust and easier-to-manage data observability tool.
Before Teachable adopted Sigma, the company used Looker as its business intelligence platform. This caused massive bottlenecks in data consumption, because all their metrics were built inside of Looker. And because few in the company knew how to write SQL, few could manipulate the data and build their own dashboards. As such, 19 out of 20 of the most-used Looker dashboards had been built by William himself in his days as Teachable’s Senior Data Analyst. Each had required him to write huge amounts of custom LookML, and business stakeholders faced a long queue for their requests for new dashboards and insights.
William knew that whatever BI tool Teachable implemented next needed to empower his stakeholders to build their own dashboards and uncover their own insights—without code.
Timely data anomaly detection with Metaplane
When William signed up for Metaplane, he quickly plugged in Redshift, Postgres, and Sigma. Then he got to work adding monitors to Teachable’s most important tables and columns. With those monitors, William wanted to ensure that the data that was delivered to their stakeholders was correct, complete, and on time. William particularly liked how easy Metaplane’s setup was:
❝If I’d had to build something even approaching Metaplane, it would’ve involved some ugly SQL with window functions for every table—and we have close to 50 schemas, some with over 100 tables. It would have taken weeks.
William set up automated alerts to the data team’s Slack channel, so that if any of those conditions were not met, the data team would be the first to know. Shortly after setting up Metaplane, that very thing happened: a data anomaly surfaced, and the data team was able to tackle it before end users even realized that the data was incorrect.
❝Metaplane gives us direct insight into what’s happening. We can tell stakeholders that we’re not only looking into it, but that we know when the data stopped updating, how many rows it’s missing, when the problem will be resolved…all before they ask. Before, we were looking into issues blindly, with little information about what had gone wrong.
Teachable has a large product team, and while the data team and the product team try to stay in sync, that’s not always possible. Using Metaplane, Teachable’s data team can check in with engineering when they see something unexpected, such as a large number of deleted records. Then, they can use Metaplane’s lineage tool to see what might have been impacted for downstream stakeholders.
Metaplane’s rich integration with Sigma also helps Teachable’s data team ensure that end business users can trust the data that they’re seeing. When Metaplane spots a data anomaly, or the data is slower than usual to update, they can let those end users know which Sigma worksheets are impacted.
Self-serve data insights with Sigma
Sigma was a game changer in enabling self-serve data exploration. Its interface is easy to use and employs a familiar spreadsheet paradigm for manipulating and exploring data, making it easy for Teachable's teams to adopt. In Sigma, users can join tables, create visualizations by dragging and dropping, and add filtering and sorting logic to columns. One particularly useful feature is the ability to add columns with custom logic, allowing users to view the data in ways that are impossible to replicate with other tools. And for those who need extra power, Sigma's custom SQL builder is available.
❝Before Sigma, one data analyst had created and maintained 95% of the company’s most used dashboards. Now, the data team has only built 15% of the dashboards that the majority of the company uses to do their jobs.
The ability for stakeholders to break their data into useful segments has proven to be a particularly powerful capability. Before Sigma, those business users would have had to ask someone on the data team to perform that t analysis for them. However, with the data at their fingertips, those business users can explore the data more deeply and uncover insights that they wouldn’t have otherwise.
Some of those insights have been so valuable that the data team has built them back into aggregate tables for wider use within the company. This has created a virtuous feedback loop, where the data team solicits feedback and observes the behaviors of the teams they serve, then enhances the data environment to be more valuable. As more people at Teachable utilize the data, more insights are unearthed—and the cycle continues.
- Since adopting Sigma, Teachable has experienced a 70% reduction in ad-hoc data requests. The company has been able to grow without needing to scale its data team because stakeholders are empowered to get their feet wet with company data. “It’s essentially like Sigma is an additional analyst (or perhaps even several additional analysts) on our team,” William says.
- Since adopting Metaplane, the data team can quickly see if anything is amiss when they log on in the morning and assign someone on the team to look into it. William estimates that this saves his team 10 hours per week; previously several teammates frequently helped identify and triage every issue that came in, which took time away from other work and forced the data team to frequently switch contexts. William loves that Metaplane requires no maintenance and provides better reporting than his team would have been able to accomplish on their own.
- Additionally, with Metaplane, William and his team noticed that there were some alerts that tended to pop up at the same time each day, helping them realize that there were certain parts of their environment that were breaking daily. This let them fill their backlog with issues that were invisible to much of the company, which let them tackle them proactively on their own time before they snowballed into larger issues.
Teachable aims to add forecasting tools to its self-serve platform, expand its data science resources for the company, and provide support for every department. The company plans to grow its team to seven or eight members.