How Teachable Reimagined Their Analytics Workflow With Sigma
Sr. Content Marketing Manager, Sigma
After catching themselves in the endless cycle of “report factory hell”, Teachable, the leader in online courses and coaching development, decided to make the switch to Sigma. In this Q&A you’ll learn how Sigma can help your line of business teams improve productivity through quick, scalable solutions and eliminate siloes reminiscent of traditional analytics workflow cycles.
Describe your analytics workflow before Sigma.
Peter Jaffe, Head of Data at Teachable: It was frustrating for all parties involved. We honestly struggled to build a truly self-service analytics and BI stack, which meant that most of the company depended on us to build reports – no matter how straightforward or predictable, predictable requests had to go through the data team. We had to dedicate two full-time employees to answer all these ad-hoc requests that came up. It took days, if not weeks, to provide our business users insights, which wasn’t sustainable.
Teachable was using another BI tool prior to Sigma. What were some of the issues you faced that Sigma solved?
Peter: We used Looker and adoption throughout the company was low because of their proprietary coding language, LookML. Power users could do their own work in Looker or directly in SQL, but the complexity of Looker was too high a bar for most users to get past. They could use the platform to consume reports prepared by the data team, but virtually no one was able to generate or even edit their own dashboards. We spend so much time and money investing in our analytics stack, but Looker was a wall that kept us from taking advantage of the new architecture.
One of the key things with Sigma is that after our team built out analytics functions for different business units, end users could get at the data they want immediately without having to come to us for help – we’re no longer a blocker.
How easy was it for end-users to jump in and start using Sigma? What are some core differences between a traditional spreadsheet and Sigma?
Peter: It’s very easy; Sigma’s spreadsheet interface allows for a shallow learning curve. Someone on our team will typically sit down with a new user for a 30-minute onboarding session where the users are oriented on the basics and the data sources available for them. These datasets act as a jumping off point for end users so folks can jump straight into the data on their own… It’s made my life so much easier.
And if we do have any questions or ad hoc requests, they are related to the data sources not Sigma.
Charles Jin, Financial Analyst at Teachable: Yeah, as an end-user, I didn’t have a lot of SQL experience in my career but I know excel, so seeing how easy it is to export data into existing models I already had made adoption really easy.
How do you use dashboards now versus before Sigma?
Peter: You know, the beauty of Sigma is that it’s a sandbox that allows users to play with data as they wish. So for us it may be better to flip that question on its head. At Teachable, our end users don’t start at top-level dashboards but the other way around in a more sophisticated, exploratory way with data itself – going directly to the workbook to join and manipulate the data for their own question-answer analyses.
And because of that, sharing work across the company is easier. So for example, Product has these things called “atoms,” which are essentially datasets that answer a specific question – let’s say the adoption rate of a specific feature. Atoms are then used as the basis for a bunch of different dashboards that are distributed via slack or on a general channel so folks across the org are always in the know.
Charles: Along those lines, from my perspective, it’s a lot easier to prebuild monthly reports to quickly export and share at the beginning of each month. So for example, we prepare monthly reports for investors… It’s hard to quantify but without Sigma, tasks like these would take far longer. This gives our team time to work on more strategic projects that can guide the company and have a greater impact holistically.
How are teams or departments working together to solve problems in Sigma?
Peter: Continuing with our example from the product team above, our VP of Product will build worksheets of elements and features he likes to track (e.g. adoption stats). A lot of end users will use these as building blocks for their own work.
Charles: On my end, I find myself reverse engineering colleagues work so that we’re all aligned on the same metrics to use and how we want to organize data. This helps us create more accurate forecasts and benchmarks that the entire org can measure itself against. This is only possible because Sigma is our single source of truth for all our teams.
What kind of analyses are you using Sigma for specifically?
Charles: At a high-level, regularly recurring tasks and ad hoc work. For the former, it’s tasks like monthly GMV forecasts, aggregating relevant data for subscription revenue and payments revenue, etc. We also update our budget and growth targets so we can compare and track our growth projections vs. realized – which we share with the company so they can see our performance over time.
But in terms of ad hoc work, we do a lot of user cohort analyses to see how different groups are using the product. So,at a high level, we did an analysis on the performance of users on our platform in relation to the speed at which they could sell their first course. These types of analyses wouldn’t be possible in excel because of the amount of data we’re working with (~ 400k+ rows of data).
What notable business outcomes have you noticed as data-driven decision-making has expanded throughout your organization?
- Over 80% of the most frequently-used workbooks and dashboards have been created by business users.
- We have about 15x the number of people across the org creating reporting in Sigma compared to our previous BI platform.
- Out of all Sigma users in the org (112) almost 90% use Sigma every month.
Charles: We’ve used Sigma to build monthly Gross Merchandise Value forecasts that are shared with other teams. These reports are ~15% more accurate than our prior forecasts. This increase in accuracy has helped us set more realistic and achievable financial goals, and better plan our strategies or initiatives that will grow our GMV.
What is the BI team able to do or focus on now that they save time?
Peter: Ad hoc requests still come in, but they’re usually more complicated questions involving multiple layers of research and hunting down data that isn’t in the CDW. It’s a great learning opportunity for our junior analysts that helps develop skills and gets them visibility in the company. We turn insights from these requests that are broadly applicable into data reports. These reports are published 1-2x per month and shared with the company on Slack.