---
title: "How Skene works | Skene"
description: "Three steps: run a one-time analysis of what you collect, agree the customer journey with GTM, then check the code against it on every release."
url: https://www.skene.ai/product/how-it-works
doc_version: "7c68abe313d0"
last_updated: 2026-08-29
---

How it works

# Know if what you shipped actually landed.

You are the one who has to say whether activation moved, or where the funnel is leaking. Skene grades the change you shipped against your own data, and it checks first that your tracking is telling the truth, so the grade means something.

[Start free](https://www.skene.ai/signup) [See pricing](https://www.skene.ai/pricing)

Sign in, connect one source, and the one-time analysis returns what you already collect.

How it works / Steps 2 of 3

See what's tracked today

Sign in and connect one source: the GitHub App or your Supabase project.

1. See what's tracked

2. Define landing

3. Protect it

- **One source is enough to start**
  Sign in and connect one source, and the one-time analysis returns what you already collect.

- **Your team keeps shipping**
  Every release is checked against the plan, and a break is named where it happened.

- **Two source kinds, that's all**
  Today Skene connects to a GitHub repository and a Supabase project. There is no other connection yet.

Supabase PostHog Mixpanel BigQuery Resend

The three steps

Sign in and connect one source. Either is enough.

[Start free](https://www.skene.ai/signup) [See pricing](https://www.skene.ai/pricing)

## See what's tracked today.

Run the one-time analysis. Skene maps what exists, what is tracked but not placed, and what is missing.

- Regex over 16 analytics platforms, no model in it

- Finds what is tracked but not placed anywhere

Overview · 78%

✓ Signals confirmed · 78% (28 of 36)

✓ Journey stages covered · 5 of 7

→ Not measured · 8

→ Last checked · Today

## Define what landing looks like.

GTM names the funnels and lifecycle moments it needs to trust. Skene derives the collection plan.

- A schema agent walks your database schema

- A code agent walks the repository

- Milestones merge by name into seven lifecycle stages

The number, counted Deterministic

- **Comparison**: Same inputs, same answer

- **Fails a release**: A missing event, or a table that is not there

- **Reported only**: A table that is there with no sample

- **Model calls**: Separate and optional

## Protect that as your team keeps shipping.

Skene checks every release against that plan and names the event, file and line where tracking went missing.

- Install the GitHub App, turn on pull request reviews

- Every pull request gets a review with comments on changed lines

SkeneTechnologies/skene-dashboard  #287 changes requested

S skene-ai bot

Skene Analytics Drift Review

This pull request has 1 tracking issue (1 wider than planned). See inline comments.

What breaks

Medium: `trial_ended` fires for a wider audience than planned at `trial.ts:77`

src/billing/trial.ts 1 comment

## See what's tracked. Define landing. Protect it as you ship.

The three steps are one loop, not a setup wizard. You run them in order once, and from then on the third one runs on every release without being asked.

**See what's tracked.** The one-time analysis reads the repository and the schema and returns what you already collect, what is tracked but never placed, and what is missing.

**Define landing.** You name the journey the business actually cares about. That definition becomes the collection plan every later check is measured against.

**Protect it.** Every release is checked against the plan, and a break is reported where it happened, in the pull request that caused it.

FAQ

## What people ask before they connect.

More questions? Talk to us.

[Talk to us](https://www.skene.ai/contact) [Start free](https://www.skene.ai/signup)

Today Skene connects to a GitHub repository and a Supabase project. There is no other connection yet.

Sign in, connect one source, and the one-time analysis returns what you already collect. No card required.

Yes. Skene OSS is the MIT-licensed command-line tool at [github.com/SkeneTechnologies/skene](https://www.skene.ai/community/open-source). It runs on your machine and needs no Skene account.

## Run the one-time analysis on your own repository.

[Start free](https://www.skene.ai/signup) [See pricing](https://www.skene.ai/pricing)

## Sitemap

- [Markdown sitemap](https://www.skene.ai/sitemap.md): every page on this site, with its markdown counterpart.
- [Index for language models](https://www.skene.ai/llms.txt): the same set as a link-dense index.
- [Full corpus](https://www.skene.ai/llms-full.txt): reference content concatenated as plain text.