How Statsig unlocks rapid customer investigations with Deepnote

Statsig was growing fast, but the data team was bottlenecked by limitations of their previous data platform around organizing and searching prior work and collaborating in notebooks. The team runs two kinds of investigations: nested customer-level investigations that reproduce and diagnose issues by company and experiment, and internal marketing and business analyses that put data scientists at the forefront of the company’s online presence and sales support.

By adopting Deepnote, Statsig unified collaboration, accelerated time-to-insight in investigations, and turned data quality alerts into reusable workflows, enabling the team to move from alert to root cause quickly and keep shipping trusted experiments and methods.

Finding the right data platform

Statsig sought to move away from an existing solution as usage scaled, driven by a need for better collaboration and more predictable operational overhead.

The team explored a range of options, including warehouse-centric approaches, but their workflows required more than query execution. For customer reproductions and method validation, they needed a collaborative, searchable notebook environment with shared workspaces, version history, and a way to find and reuse SQL queries and prior deep dives. For customer reproductions and method validation, they needed Python notebooks and reusable templates, which alternatives could not provide.

Extensibility was just as important as collaboration. Statsig’s data lives across multiple systems: product data in MongoDB, configuration files in GitHub and the Statsig SDK, and a roadmap that could include engines such as Snowflake, Trino, ClickHouse, or Redshift. Given the pace of growth, the team needed a notebook environment that could connect to whatever data platform they adopted next, without being locked into a single vendor, platform, or cloud.

Difficulty finding and reproducing prior work

Without a unified workspace, picking up where a teammate left off was challenging. Valuable analytical queries were often impossible to retrieve, causing unnecessary duplication of work or issues when rerunning code on a new machine.

Method validation on real data

Real-world validation of the statistical methods Statsig offers to its users required production-grade Python & SQL notebooks that could act as reusable workflows within the company. Achieving this with in-house Jupyter or ad hoc Colab notebooks was difficult, and warehouse-centric tools didn’t provide the flexibility the team needed.

Plug and play integration across your stack

Deepnote connects to MongoDB, Trino, Spark, and Statsig’s own SDKs. The team can open a fresh notebook, attach the required integrations, and get moving without the overhead of environment setup.

Google Docs style collaboration

With a multi-player data notebook, anyone from the team could pick up where their colleagues left off, and the team was able to complete projects faster while collaborating on the same notebook together - with enhanced visibility and audit logs of who changed what.

Deepnote Agent for coding assistance and EDA

Going from a query to a tailored chart now takes seconds, helping explore data faster. With Deepnote Agent, analysts can skip API references for prominent Python frameworks and use text-to-SQL to generate and customize charts, without context switching.

Reusable, parameterized investigations

With a single parameter change, the team can reuse the SQL template for customer investigations, visualize key metrics, and run in-depth analysis while avoiding much boilerplate.

Alerts based on scheduled workflows

In a simple Deepnote-powered internal data app built by Statsig, an on-call engineer pastes a data-quality alert into a notebook and hits Run. Deepnote reproduces the alert, highlights the violating rows, and sets the stage for investigation.

From there, analysts group and filter the data to uncover patterns, such as a specific experiment type or a recent code change. Once the root cause is identified, they route fixes to the appropriate owner and ensure data lands on time.

Faster customer investigations

Frontline data support can kick off and complete deeper investigations more quickly, keeping focus on insights rather than tool friction. One notebook handles reproduction, exploration, and communication, so customers get answers faster.

Faster, collaborative experimentation

The team uses Monte Carlo simulations on production-like data to compare new methods against the current one, including edge cases. Deepnote’s shared history, multi-player editing, and templates make the workflow fast, realistic, and deployment-ready.

Flexibility as the stack evolves

With numerous native data integrations, the team can ingest data from anywhere and easily switch data warehouses or compute engines as needed.

With AI, teams are building richer experiences and shipping faster than ever. Statsig is committed to helping teams ship faster and smarter - and solving complex data problems with Deepnote is key to their journey.