Decoded GA4 in Evidence Studio —
same connector, no local Node,
auto-publish on save.

Evidence Studio is the managed cloud version of Evidence. Same SQL-in-Markdown reports, same BigQuery connector, but the runtime is hosted, the editor is in-browser, and saves auto-publish. Decoded GA4 keeps the SQL inside your reports short — and that matters more, not less, when other people are editing them.

Connection: BigQuery Hosting: managed cloud Editor: in-browser Publish: on save or push
Summarize This ChatGPT Perplexity

Evidence Studio takes the Evidence framework and removes the local-setup tax. Decoded GA4 takes the schema tax off the SQL the reports contain. Together that is the whole picture.

What Evidence Studio is for

Evidence OSS is great if you want to commit reports to Git and run a Node build pipeline. Evidence Studio is the same thing without the operational overhead — a hosted runtime, an integrated editor, automatic publishing. You either save in the editor or push to a connected Git repo, and the report rebuilds.

Why nested GA4 hurts a managed editor more than it hurts local dev

In local dev, the analyst writing the SQL has the tooling and the patience to deal with UNNEST. In a managed editor used by PMs, marketers, and analysts who would rather be writing the prose, every CROSS JOIN UNNEST is a barrier. They open the file, see twenty lines of flatten before the actual question, and close it again.

What changes with Decode GA4

The events table is flat inside BigQuery. Evidence Studio's BigQuery connector reads it like any other table. The SQL blocks inside the Markdown are short. Non-engineers can read them. Engineers can write new ones in-browser without context-switching to a different tool to figure out the right UNNEST pattern.

Option A

Connect Evidence Studio to the raw export

Point the BigQuery connector at the events_YYYYMMDD table directly. Every SQL block in the editor needs UNNEST. Authors paste the same flattening pattern into every new report. The in-browser editor surface fills up with scaffolding before the question even starts.

Editor full of UNNEST scaffolding
Option B

Maintain a flattening view in BigQuery

Build and own a SQL view that flattens GA4 once, register that with Studio. Cleaner SQL inside the reports, but the view becomes a hidden dependency. Updates to the view do not trigger Studio rebuilds, and schema drift breaks the view silently.

Hidden upstream that drifts
Option C

Run dbt before Evidence Studio

Stand up a dbt project to flatten GA4, materialise a clean events table, and have Studio read that. Tidy, but you have just added an entire transformation tool to a workflow that was supposed to be lightweight. Now there are two scheduling systems and two failure modes.

Two tools doing one job
Feature Decode GA4 source Raw GA4 export
SQL block length in editor3-5 lines20+ lines per report
Auto-publish behaviourSaves render fastUNNEST inflates build time
Non-engineer authoringSQL is readableSQL is not
Build cost in BigQueryPartition-pruned scansUNNEST inflation per build
Schema change recoveryAuto-detected upstreamManual fix in every report
Time to first published reportMinutesHours of UNNEST plumbing

Sign up for Evidence Studio, connect BigQuery in the platform, write a Markdown file in the browser. Save publishes.

  1. [ 1 ]

    Subscribe via Google Cloud Marketplace

    Decode GA4 ships as a Marketplace listing. Usage-based pricing, no monthly minimum, billed through your existing GCP invoice. Subscription takes about a minute.

  2. [ 2 ]

    Sign up for Evidence Studio

    Create an Evidence Studio account. The managed runtime, the in-browser editor, and the deploy pipeline come pre-configured — no Node.js to install, no build server to maintain.

  3. [ 3 ]

    Connect BigQuery in-platform

    In Evidence Studio's connector UI, add a BigQuery source. Provide the GCP project ID and a service account JSON with BigQuery Data Viewer plus Job User. The decoded events table becomes available to your reports.

  4. [ 4 ]

    Write a Markdown report and save

    In the in-browser editor, write a SQL block referencing the events table and a chart component below it. Save — Studio rebuilds and publishes. Or push to a connected Git repo and the same rebuild happens on commit.

Wire decoded GA4 into Evidence Studio in four steps. The runtime is managed, so most of the local-Evidence setup falls away — what is left is the BigQuery connector and the SQL.

01

GCP

Run the Decode GA4 installer with the events_external template. Pick the dataset Studio will read from.

02

GCP

Create a service account with BigQuery Data Viewer and Job User. Download the JSON key.

03

Evidence Studio

Add a BigQuery connector with project ID and service account JSON.

04

Evidence Studio

Write a SQL block in a Markdown file in the in-browser editor. Save to publish.

Evidence Studio reads the decoded events table through the service account you provide. The data stays in BigQuery — Studio handles rendering and hosting. Standard GCP IAM applies; rotate the key the same way you would for any other connector.

01

No local Node setup

Studio hosts the runtime. Authors who do not want to install Node, manage versions, or run a build server simply do not have to. The trade-off is less local control — for most teams, that is fine.

02

In-browser editor

Markdown and SQL are edited together in Studio's UI. Save and the report rebuilds. Useful for non-engineers who would rather not learn a Git workflow to update a chart.

03

Auto-publish on save or push

No deploy step. Save in the editor or push to a connected Git repo and the report goes live. The decoded events table feeds the same fresh data on every rebuild.

04

SQL inside Markdown that authors can read

Decoded events mean each SQL block is a normal GROUP BY against direct columns. PMs and marketers reading a report can follow the SQL above the chart.

05

Schema evolution that just works

When GA4 adds a new event parameter, the next decode run picks it up. Studio reports keep building. The new column is available the next time someone wants to use it.

06

Same source as the rest of your stack

The decoded events table also feeds dbt, Looker Studio, Steep, Rill, and Lightdash. Studio is one consumer of many on the same upstream.

01

Client-facing reports for agencies

Agencies running multiple GA4 properties write one Studio report per client. Decoded events make the SQL identical across clients — only the project ID changes. The hosted runtime means no client-specific build server.

02

Internal weekly business reviews

A single Markdown file in Studio that updates every Monday morning with the week's revenue, conversion, and traffic numbers. The SQL is committed; the prose is editable in-browser; the chart auto-refreshes on save.

03

Marketing experimentation reports

Each campaign or A/B test gets its own Studio page with SQL pulling specific event_param dimensions out of the decoded events table. The hosted publish pipeline means stakeholders see the analysis the moment the analyst saves.

How is Evidence Studio different from Evidence OSS?

Same framework, same SQL-in-Markdown model, same BigQuery connector. Studio adds managed hosting, an in-browser editor, and auto-publish on save or push. The trade-off is less local control over the build environment. See Evidence OSS →

Do I need a Git repo to use Evidence Studio?

No. You can edit and publish entirely in-browser. If you want pull requests, code review, or CI on top, connect a Git repo and Studio will rebuild on push as well as on save. See setup →

What permissions does the BigQuery connector need?

BigQuery Data Viewer and BigQuery Job User on the project containing the decoded events table. No write access. Provide the service account JSON when you set up the connector.

Where do the published reports actually live?

On Evidence Studio's hosted infrastructure. Authentication, custom domains, and access controls are managed inside Studio. The data behind the reports stays in your BigQuery project — only the rendered output is hosted by Studio. Full prerequisites →

Deploy in under 5 minutes

Hosted Evidence,
without the UNNEST inside.

Subscribe via Google Cloud Marketplace, point destination_dataset_id at the dataset Studio reads, and publish your first SQL-in-Markdown report from a browser tab — no Node setup, no build pipeline.

Get Started on Marketplace → Read the documentation

Google Cloud Marketplace · Usage-based · No monthly minimum