This is the hub page for the Getting Started section. It explains how Cogny works, the order to set things up in, and what a healthy first month looks like. Each step links to a detailed page.

What Cogny is, in one paragraph

Every growth team already agrees on the process: track everything you do (in your channels, on your site, ideally in your CRM), put the data in one place, analyse it, turn analysis into insights, turn insights into experiments, ship the experiments, and measure them. Teams grow as fast as they can complete that loop. In practice the loop stalls at the bottlenecks: nobody looks at the dashboard, the analysis never becomes a brief, the brief waits for a developer. Cogny runs the same loop, with AI doing each step, and closes it on its own.

Cogny is not a content generator and not a dashboard. It is closer to an operating system for growth: it connects to your stack (analytics, search, ads, e-commerce, email, CRM, code repository), reasons over the data, proposes experiments as tickets, executes the ones you approve, and follows up on the results. Everything you do in it becomes context, so it gets better the longer you use it.

Three principles shape how you set it up:

Cogny has no data of its own. It only knows what you connect. The more sources, the smarter it gets. A workspace with Search Console alone can do SEO. A workspace with Search Console, GA4, Google Ads, your CRM and your repo can run the whole funnel.
Cogny never acts without a decision from you. Reads are automatic. Writes (a budget, a campaign, a page) wait for approval unless you deliberately loosen that. Code changes always arrive as a pull request you merge.
Context beats prompting. What separates a useful ticket from AI slop is the business context, the data and the history of what worked. Time spent on context pays back every week.

The four building blocks

Block	What it holds	Where in the app
Data sources	Connected integrations (called MCPs) and your BigQuery warehouse	Settings → Data & Integrations
Context	Industry, company, competitors, growth strategy, the context tree, core metrics	Settings → AI Knowledge
Reports	Scheduled analyses built from 50+ role-based templates or your own instructions	Reports
Tickets	The experiment backlog: proposed, approved, executed, analysed	Tickets

Reports read data and context and produce insights. A second agent turns insights into prioritised tickets. You approve, refine or reject tickets. Approved tickets execute through an integration, the coding agent or a human, then move to analysis, and the learnings feed the next round of reports.

The setup sequence

Do these in order. The whole sequence takes about an hour when the person with access to your tools is in the room. Your onboarding call with Cogny follows the same sequence.

Prepare for onboarding. Decide what the workspace is for (one website, one business model), gather admin access to every tool, and run the Cloud wizard.
Connect your data sources. Search Console first, then GA4 and Tag Manager, then the ad platforms you actually run, then your site (GitHub repository or CMS), then CRM, e-commerce and email. Set tool permissions.
Review and enrich context. Read the three generated documents and correct them. Write a growth strategy. Upload what the agent needs to know into the context tree.
Define core metrics. Name the two to five numbers you run the business on so every report and ticket is judged against them.
Turn on reports. Start with three or four, pick the recipients, and let the first run happen.
Work the ticket board. Review the first twenty tickets properly. Reject with reasons. Then decide how autonomous you want it to be.
Connect your repository if your site is code, so tickets can ship as pull requests with previews.
Add automations and loops once the basics produce good tickets.

What to expect, week by week

Week 1: connect and read. Data only accumulates from the day you connect. The GA4 and Search Console links into BigQuery take 48 to 72 hours to start populating and do not backfill history. Use the week to fix context, define metrics, and confirm tracking actually fires on every funnel step. The first report run happens immediately when you enable a template, but it is working from a thin slice of data.

Week 2: first real tickets. Reports have a week of data. The ticket board fills with proposals. Review them with the person who owns the channel. Expect some misses: a ticket that repeats something you already did, or one that ignores a constraint you never wrote down. Every rejection with a reason and every context edit fixes a class of mistake, not one ticket.

Weeks 3 and 4: the loop runs. Approved tickets execute and move to analysis. Search Console starts showing new pages and rewritten titles. Paid tickets get re-analysed once there is enough data. If you run the SEO and content loop, this is usually when the first new organic clicks appear. Returns on paid search and conversion fixes can show within days; organic takes weeks on an established domain and two to three months on a new one.

Ongoing: about an hour a week. Once set up, the weekly routine is: read the Monday report, clear the New column on the ticket board, glance at Analysis. Teams that are new to the way of working spend more, typically 10 to 20 hours in the first month, because the first tickets deserve real review and because tracking gaps surface and need fixing.

What good looks like

A reference Cloud workspace that has run for two years (a Nordic non-profit with a large donor funnel) looks like this:

Six connected integrations: Search Console, GA4, Tag Manager, Meta Ads, an email platform and an e-commerce platform for donations.
Context: roughly 10,000 characters of industry context, 4,500 of company information, 12,000 on competitors, 3,700 of growth strategy, plus a 13,000-character organisational memory the agent maintains itself.
A context tree of about 20 documents describing the analytics measurement plan: every tracked event, checkout step, KPI definition, dashboard, known tracking problem and tool.
Three core metrics, all conversions by donation type, checked against the metric definitions in every report.
Six active weekly reports, most on adaptive scheduling, plus one monthly.
Around 250 reports and 50 tickets generated, with a steady mix of approved, rejected-with-reason and manually executed items.

A workspace onboarded last week (an international non-profit) got there faster: five integrations on day one, context written on day one, the analytics measurement plan uploaded to the context tree on day four, five tickets by day five, and a first weekly report scheduled. That is the shape to copy.

The mistakes that cost the most

Starting on the wrong domain. Optimise the domain with history. A brand-new domain gets crawled rarely and needs months plus some paid traffic before organic moves.
Tracking gaps. If GA4 does not record your lead or purchase event, no amount of analysis fixes outcomes. Verify tracking before spending on paid.
Skipping the context review. The generated context is roughly right, not exactly right. A mis-framed company or a missing competitor propagates into every report.
Writing rules in chat. Chat is for questions and one-off actions. Rules belong in the growth strategy or the context tree, where every report reads them.
Too many reports. Reports cost credits. Run the three or four you read.
Under-producing content. Founders assume customers already know what they know. One post a month does not move search or LLM visibility.
Expecting a managed service. Cogny is a tool with an optional human in the loop. Someone on your side has to approve tickets weekly.

Where to get help

Book an onboarding session from the Cloud wizard or your welcome email. Bring the person who holds admin access to your tools.
The in-app chat can answer questions about your own data and your setup.
Settings → Help & Videos lists the support channel and walkthrough videos for your workspace.