Skip to content
databrushDocs
  • OverviewDatabrush’s churn and retention analysis for your Supabase database, in the app and over MCP.
  • QuickstartConnect Supabase, choose your database and confirm the setup to publish your first analysis.
  • Supabase integrationHow Databrush connects to Supabase, what it reads and keeps, and how to revoke its access.
  • Connect your agentConnect Databrush to Claude, Claude Code, ChatGPT, Codex or another agent over MCP.
  • Tool referenceEvery tool Databrush’s MCP server offers, what it does, and who can use it.
  • How it worksHow Databrush analyzes your data, what Home shows, and how sure each finding is.
  • DefinitionsCustomers, who is left out, actions, churn and the dormant window, and how to change them.
  • Security & dataHow Databrush protects your database, your customers’ details and your account.
Connect Supabase
How it works

Product

How it works

Databrush keeps your analysis up to date as your customers change. Every number comes from fixed formulas and the definitions you confirmed.

On this page

The analysis

Databrush keeps reading the schemas you chose and adds what changed to each customer’s history: their status, their activity and their warnings. It works out churn and retention with fixed formulas from your definitions, so the same data always gives the same numbers.

AI helps propose your setup, name new values it finds and word product suggestions. It doesn’t compute any number. See AI providers.

Home

Under the Overview chart, Home has four sections. Your agent reads the same views over MCP.

SectionWhat it shows
Waiting for youDecisions to make: new product suggestions to review, new actions to keep or remove, proposed exclusions and re-confirmations.
Needs attentionCustomers likely to churn, with the reason, their warnings and whether someone has contacted them.
Retention SignalsBehaviors in customers’ first week that are linked to staying, with the statistics behind each one.
Product suggestionsSuggestions you accepted or shipped, with how sure Databrush is and what each could gain.

Each customer has a record with their status and the reason for it, their warnings, their first week and their daily activity.

How sure a finding is

Retention Signals and most product suggestions carry a label for how sure Databrush is, strongest first:

LabelWhat it means
ProvenA clear difference that held again in customers who joined later.
Early signalA clear difference that holds after accounting for testing many signals at once. Proven needs it to repeat.
Worth testingClear on its own, but it could still be chance once many signals are tested.
Early hintThe estimate points this way, but the data can’t yet rule out no difference.

A signal can also show No clear difference or Not enough data.

Gain and results

Each product suggestion says what acting on it could gain. Usually that is the share of customers Databrush expects you to keep, next to today’s share. Otherwise it is the number of customers at stake. A suggestion shows money only when your database holds revenue, and it says what it assumed.

When you ship a suggestion, Databrush fixes how it will measure the result and keeps tracking it until the results are in.

Emails

Databrush emails you when something needs you, and sends a weekly digest when there’s news. Emails never name a customer. Choose which emails you get in Settings › Notifications, or use the unsubscribe link in any email.

Questions? Email support@databrush.ai.