A data workspace is the place where an analyst actually does the work of analysis. It is the IDE for the data team. You connect it to the warehouse, write SQL or Python in cells, see results inline, build a chart, write a paragraph of explanation, and share a link with whoever asked the question. Workspaces are where data analysis is authored, in contrast to BI dashboards, which are where finished analysis is consumed.

The simple way to think about it: a BI tool like Looker or Tableau is a magazine. A workspace like Hex or Mode is the writer's word processor. Magazines are for readers; word processors are for writers. Both matter, but they are different products built for different jobs.

Where the Category Came From

Before workspaces existed, an analyst's life looked like this: write a SQL query in a desktop client (SQL Workbench, DataGrip, or the dreaded Microsoft SQL Server Management Studio), copy the result into Excel, build a chart, paste the chart into a Google Doc, and email the doc. Sharing a query meant sharing a .sql file. Sharing context meant writing a wiki page that nobody read. Versioning meant nothing.

Mode Analytics, founded in 2013 by Derek Steer (ex-Facebook) and others, was the first product to put SQL, Python, visualizations, and a sharable URL into one collaborative web app. That combination — write a query, get a shareable analysis — defined the category. For most of the 2010s, Mode was synonymous with "the place data teams do work."

In parallel, Jupyter notebooks (the open-source descendant of IPython) became the default tool for data scientists working in Python. Jupyter was great for individuals and terrible for teams: notebooks lived in random folders, had no real-time collaboration, and produced JSON files that broke in Git. Cloud-hosted Jupyter alternatives like Deepnote (founded 2019) and Google Colab tried to fix this, with mixed success.

Then in 2019, Hex launched and quietly ate the category. Hex took the best of Mode (SQL-first, shareable, opinionated UX) and the best of Jupyter (Python cells, full data science toolkit) and added the thing both were missing: a real reactive execution model and a data app builder that turned notebooks into interactive tools non-analysts could actually use. By 2024, Hex had become the default new-team choice and Mode had been acquired by ThoughtSpot.

What Makes a Workspace a Workspace

Three properties separate a real data workspace from a glorified SQL client:

1. Multi-language cells, one runtime. A workspace lets you mix SQL (against the warehouse) and Python (against the result of the SQL) in the same document, with results from one cell flowing into the next. This is the killer feature: pull data with SQL, transform it with pandas, model it with scikit-learn, chart it with Plotly, all in one place.

2. Collaboration as a first-class concept. Multiple people can open the same notebook, leave comments, fork it, or watch it run. Every analysis has a URL. Every URL has permissions. This sounds obvious in 2026, but it was the entire wedge that killed desktop SQL clients.

3. The author/consumer split. A workspace produces two artifacts from the same document: the notebook (for other analysts to inspect and fork) and the app (for stakeholders to use without seeing the code). Hex was the clearest articulator of this idea — they call the consumer view a "data app" — but it's now table stakes for the category.

The Opinionated Take: Hex Won

By 2026, the workspace war is effectively over and Hex won. Three reasons:

1. Reactive execution model. Hex notebooks recompute downstream cells automatically when an upstream cell changes, the same way a spreadsheet recalculates when you edit a number. Jupyter and Mode both used a linear "run cells in order" model that broke constantly when analysts ran things out of sequence. Reactivity sounds like a small detail; it isn't. It eliminates an entire category of bugs.

1. The data app builder is genuinely good. Most "turn your notebook into an app" features are toys. Hex's app view is well-designed enough that PMs and ops people actually use the apps analysts ship. This means analysts can deliver self-serve tools without bothering the BI team to build a Looker explore.

1. AI was integrated early and well. Hex Magic (their AI assistant) shipped before most competitors and was tightly coupled to the warehouse schema, which made it actually useful instead of generating hallucinated SQL.

Mode is still a good product. Deepnote is still a good product. But the gravity has moved.

Where This Fits in the Stack

Workspaces sit above the warehouse and next to BI tools. The data flows like this:

Source databases / SaaS apps
        ↓ (Fivetran, Airbyte)
Data warehouse (Snowflake, BigQuery, Databricks)
        ↓ (dbt models)
        ├──→ BI tool (Looker, Tableau, Power BI)   ← consumed by execs
        ├──→ Workspace (Hex, Mode, Deepnote)        ← authored by analysts
        └──→ Reverse ETL (Hightouch, Census)        ← pushed to ops tools

A workspace is not a replacement for a BI tool. The two coexist. BI tools are better for governed metrics, scheduled dashboards, and self-serve exploration by non-technical users. Workspaces are better for ad hoc deep dives, statistical analysis, ML prototyping, and one-off interactive tools that don't justify a full Looker model.

Tools in This Category

• Hex — The current leader. Reactive execution, Python + SQL, polished data app builder, strong AI features.
• Mode — The category-defining product. Acquired by ThoughtSpot in 2023. Still widely deployed but no longer the default new choice.
• Deepnote — Cloud Jupyter done right. Strong with Python-first data science teams.