Supabase

The challenge: Analytics tools that couldn't keep up with hyper growth

Early on, Supabase's data team was just two engineers, with an analytics stack to match. They used BigQuery as their data warehouse, dbt for data models, Tableau for reporting, and Redash for ad hoc queries. In theory, it would be able to support technical and non-technical users, but in practice, almost nobody used it.

Some tools were powerful but deeply unintuitive. Other tools had bare-bones functionality with disconnected queries and no visualization. With so many disparate data sources, building ad hoc analyses through manual research and spreadsheets was a cumbersome task. Sometimes teams avoided using data altogether.

It was clear, Supabase needed a single tool that could consolidate information from multiple systems and adapt to workflows that looked completely different from one team to the next.

The solution: One platform for dashboards, operations, and prediction

The team at Supabase initially evaluated Hex for its dashboard capabilities, however, Hex could do something their previous tools never could: combine SQL, Python, API calls, and visualizations. It also allowed Supabase to scale data accessibility across the organization for technical and non-technical users.

Today, Hex is Supabase's primary analytics platform. It’s widely adopted across at least nine teams, including: engineering, support, marketing, product, data, operations, HR, finance, and leadership. With faster and easier access to data, the company refreshes 50 dashboards in what they call "Monday Metrics Madness" — a weekly company-wide async business review where product owners flag anomalies and explain changes. The marketing team uses Hex to evaluate the effectiveness of leads sourced from events, webinars for marketing funnel progression, and blog content to increase engagement. Other teams, such as billing and product, took Hex to the next level: orchestrating entire workflows, building predictive models, and surfacing signals that tell Supabase exactly where revenue is at risk.

Hex became Supabase's go-to analytics platform because it was the only tool flexible enough to serve every kind of user from deep technical analysis to conversational analytics.

“The thing about Hex that is absolutely magical is that everyone in the company has fast, easy access to all the data in the business, with an easy to use interface to run data pulls using pre-built prompts. Hex has enabled everyone in the company to be data-driven.” - Prashant Sridharan, Head of Product Marketing

“Hex is absolutely magical. It gives everyone fast, easy access to all the data in the business, with an easy to use interface to run data pulls using pre-built prompts. Hex has enabled the whole company to be data-driven.” - Prashant Sridharan, Head of Product Marketing

Billing command center: support tickets increased by 10x and were resolved 92% faster

Supabase had a lot of niche use cases in the billing department that required tailored insights to support customers whose outreach came through the website and direct messages. As the ticket volume increased by 10x, a single refund meant navigating to five different systems to verify the invoice, check balances, and process payment and three systems to process the refund. Each step was manual and error-prone.

Kevin Grüneberg, Lead Billing Engineer, wanted to reduce the time it took to resolve a single billing issue. To do so, he started with a simple idea: what if support agents could consolidate the relevant data points for a customer in one view?

What started as a simple consolidation idea became something far more ambitious: Kevin built his entire billing command center in Hex. With Hex, the command center pulled everything the team needed into one place: current plan, invoices, charge status, payment methods, dispute history, and Python-derived data from systems without APIs. It auto-flagged common issues, routed the team into nine sub-playbooks,ran the logic and then summarized recommendations for refunds, dunning extensions, and plan changes. Within the Hex dashboard, users could click an approval button, which would prompt Hex to fire an API call to execute a refund. No Postman. No tab-switching. No second-guessing the order of steps.

Hex was key to building this system because the code for billing execution could not live on the backend of Supabase’s product.

The Hex difference? A refund that once took two minutes now takes ten seconds. The team handles 60 tickets a day with the same headcount that previously managed a fraction of that volume — reducing resolution times by 92%, or 480 hours a year.

"Hex replaced the need to open eight systems by consolidating the information and refund actions into one platform view. Resolving a support ticket and issuing a refund that used to take minutes now takes seconds." — Kevin Grüneberg, Lead Billing Engineer, Supabase

Fraud prevention: $300K+ in revenue protected, catching fraudulent cases 97% faster

Before Hex, the billing team was just two engineers and they needed to move fast, but did not have the resources to build out a fraud prevention system from scratch. Fraudulent cases would surface on overdue invoices up to six weeks after a bad actor had already consumed thousands of dollars in compute charges and disappeared.

Kevin built a fraud alerting playbook in Hex that runs multiple times per day. It sends an alert to flag accounts that exhibit suspicious signals. For example, a customer with low lifetime spend who suddenly spins up a $4,000-per-month instance.

When the Hex data app identifies fraudulent activity, it sends a Slack alert to the billing team with the relevant account data. A support engineer can then review the situation, pick an action, and execute the update directly from the playbook.

The shift from catching fraud on past due invoices to catching it on a same-day fraud alert has greatly reduced Supabase’s risk exposure.

Building the system in Hex enabled the team to build faster insights and reduce the impact of fraudulent issues. After just ten months of running the fraud alert workflow, the team prevented approximately $300,000 in fraudulent or misconfigured spend and caught fraudulent instances 97% faster than before.

"Before Hex, fraud would show up as an overdue invoice six weeks later and there was nothing we could do. Now we catch it the same day." — Kevin Grüneberg, Lead Billing Engineer, Supabase

Churn prediction engine: 4 months of work built in 30 hours

In addition to understanding the nuances of billing, Supabase also needed to have a better way to identify which signals were indicative that a customer was at risk of churning. However, they did not have a structured way to identify which enterprise customers were at risk before it was too late. Churn signals were scattered across several systems, making it difficult to compare customers and insights.

To get started, the team had to identify which signals were important. Instead of guessing which signals mattered, Tim Frietas, Product Manager, used Hex's Notebook Agent to backtest dozens of possible signals against historical data and score it on strength and coverage. He also used the Notebook Agent to recommend how to weigh each signal based on its correlation with the potential ARR loss by customer.

Using these insights, Tim and his team built four dashboards targeting four distinct churn patterns: migration risk (customers researching how to export data to other databases), abandonment (declining dashboard visits, fewer API calls), growth ceiling (customers who plateau and slowly fade), and customer sentiment based on customer touch points.

The final output was a composite churn risk score for the top enterprise accounts, tiered into critical and high-risk buckets. Each week, the account teams reviewed the top five accounts, reached out to customers to start conversations, and logged outcomes.

So when an account shows up with a sharp ARR drop and declining API usage, the team knows exactly what to ask: "Did you find the team plan features valuable?" or "Are you migrating workloads elsewhere?"

Even better, to build this churn prediction engine, Tim wrote zero SQL by hand. Instead of taking an estimated four months to manually code, Hex was able to speed up the development process to 30 hours, reducing development time by 96%.

"I wrote zero SQL by hand, absolutely zero. Hex is the single most useful and magical tool that has ever existed to make my job better in 20 years." — Tim Frietas, Product Manager, Supabase

The impact: Hex went beyond analytics cool, it was a workflow orchestrator

Supabase went from a fragmented analytics stack that most people avoided to a single platform that runs the business from board reporting to fraud prevention to billing support operations. Leveraging Hex to optimize workflows has allowed users to focus on more strategic work.

With Hex, Supabase achieved:

• 92% reduction or over 477 hours saved in billing ticket resolution time, cutting single-case resolution from 2 minutes to 10 seconds across 60 daily support tickets
• ~$300K in fraudulent transactions prevented over a 10 month period, by catching suspicious accounts the same day rather than six weeks later on unpaid invoices
• 97% faster fraud detection, moving from invoice-cycle discovery to same-day alerting
• 96% reduction in churn dashboard development time, building a full enterprise churn prediction engine in ~30 hours instead of an estimated 4 months (640 hours)

The team is now working to pipe in unstructured customer text from Slack, support channels, and surveys to analyze customer sentiment in Hex. Tim is also planning to bring in more granular technical performance metrics such as CPU usage and connection pool health to sharpen the churn signals, giving the success team even more context before they reach out to at-risk accounts.

Tim Frietas, Product Manager and Kevin Grüneberg, Lead Billing Support — Supabase