The CRM everything else runs on
Canopus is an autonomous growth manager whose CRM is its foundation: customers and prospects in two tables that your own product, your landing page and your Stripe account keep filled, and that every campaign, audience, attribution and report reads.
Nothing in Canopus works without it, and nothing in it is maintained by hand. Six fields are real columns. Everything else is a trait, so a new field costs one row and never a migration. Audiences are built over these fields, your Stripe cohorts and your product activity, saved once and offered to every campaign and sequence.
How does the CRM fill itself?
Your product sends events
A signed POST endpoint per organization takes ten event types: a ping, and nine about a person (created, updated, trial started, plan changed, payment failed, payment recovered, churned, reactivated, activity). Signature is HMAC-SHA256 over the raw body, compared in constant time.
Connections3 ready- StripeCustomers, subscriptions, revenueSyncedsynced 2 minutes ago
- PostHogSign-ups and product usageSyncedsynced 5 minutes ago
- Sending domainmail.yoursaas.com · DKIM, SPF, DMARCVerifiedchecked 11 minutes ago
Your landing page posts leads
A public endpoint keyed by a form key, with a honeypot field a person never sees and a bot always fills. UTM parameters are captured and stored on the contact. A new prospect lands on the first stage of the pipeline.
A CSV brings the list you already have
Headers are mapped, types are inferred from the values rather than the header, and the columns the file needs are created before the first row lands. Fill mode is the default: the import writes only where Canopus is silent.
Stripe writes what only Stripe knows
Five fields, matched on the Stripe customer id first and the lower-cased email second: MRR, plan, billing period, status and sign-up date. The first pass is a button you press; every pass after it rides the ordinary sync.
RevenueLast 30 daysMonthly revenue€4,280+8.2% vs last monthActive subscribers146New customers19Attributed this month€612from campaignsLast six monthsYou read it as views, edit it in place
Saved views are tabs with their own filters, columns and sort. Cells are edited in the grid. Prospects also render as a kanban with a column per stage, and moving a card to a stage marked as won turns that prospect into a customer.
Prospect pipeline7 prospectsNew2Sofia MarchettiNorthboundBen OkonkwoLatchkeyContacted3Hannah VogtPagewiseYuki TanakaKettlePierre LambertRivageDemo2Marta SilvaFieldnoteOwen ClarkeSundial
What does this replace?
A spreadsheet base with a column per idea and no link to the product
Two tables fed by the product itself, where a new field is a row rather than a schema change.
Replaces: A spreadsheet baseTyping MRR and plan into cells from Stripe open in another tab
Stripe owns five fields outright and rewrites them on every pass.
Replaces: Copy and pasteA form that emails you and a rule that files the email somewhere
A public lead endpoint that creates the prospect on the first stage directly.
Replaces: An automation toolFiltering a spreadsheet by hand every time you want a segment
A saved view is a question, and the same question becomes a campaign audience without being rewritten.
Replaces: A filtered sheetDiscovering that a customer changed their email and now exists twice
Matching goes by product identifier, then Stripe customer id, then email, in that order.
Replaces: Manual de-duplication
What are the columns, and who is allowed to write them?
Six keys are real columns of the contact table: email, name, company, source, stage and last seen. A seventh joined them later, the identifier your own product knows a person by, which is visible and filterable and can never be retyped, because changing it would break the match between a contact and every event that keeps it current. Everything else a founder can see (first name, plan, MRR, status, sign-up date, tags, notes, and every column a spreadsheet export ever carried) is a key of a JSON bag called custom, declared by a field row that gives it a label, a type and, for a list, its options. The split is a test rather than a convention: a key earns a column when the database needs it as one, which means when it is joined on, made unique, or used to order a page. That is why the Stripe customer id is a column and the MRR it governs is not. One registry knows the split and one SQL builder turns a field into an expression cast by its type, so a filter that works in the grid works identically in a campaign audience. The thing this buys is plain: adding a field is a row, never a migration, and there is no deploy between a founder deciding to track something and tracking it.
Stripe is the one connector allowed to decide who is a customer. PostHog seeing a session proves somebody used the product, which is a hint; a Stripe customer object with a live subscription is the payment relationship itself, recorded by the system that collects the money. So Stripe owns five fields outright, on any contact it bills: MRR, plan, billing period, status and sign-up date. It matches on the Stripe customer id first and the lower-cased email second, writing the id on the pass that first matches by email, which is what makes every later pass follow the person rather than the string when they change their address. It creates the payers Stripe knows and the CRM does not, and it touches nothing else: company, tags and the columns that came from your export are never read there. Every pass rewrites those five, so a figure corrected by hand on a Stripe-linked contact is overwritten. The correction belongs in Stripe, where it will also be true on the invoice. They stay editable on a customer who pays by invoice, and the columns say whose they are. The first pass is a button rather than a background job, because it is the one that can create dozens of contacts and rewrite freshly imported columns, and it is worth watching.
Three doors let data in, and each has a rule that cannot be argued with. The product webhook is signed, rate-limited per IP and per organization, capped at 64 KB and 50 custom keys, deduplicated on your event id for good, and it honours the occurred-at timestamp rather than arrival order, so a retried plan change that lands after a cancellation changes nothing. An unknown key inside the payload becomes a text column on sight, if automatic columns are on. A refused value (prose sent to a number field, a full option list) is stored on the event and named in the response instead of being swallowed, because an empty column is the one failure nobody can debug from the outside. The lead endpoint is public by nature, so it is bounded to 16 KB and 30 custom keys, and a lead never demotes somebody who is already a customer. The CSV import fills blanks by default and never overwrites silently: replace mode exists behind a warning, tag lists union rather than replace under fill, and neither mode may turn a customer back into a prospect. Up to 5,000 rows, measured at nine seconds for a file of that size across twelve columns.
Filters are a flat list of conditions that all have to hold, which keeps a view readable at a glance, and the one thing that flatness could not express has its own operator. "Plan is any of Max Annual, Max Monthly" is a single condition carrying several values, offered on text, select and multi-select. It is not offered on numbers or dates, which have ranges that say it better. Its negative is not a plain negation: "plan is none of pro" includes the contacts with no plan at all, because in SQL a comparison against null is null and a naive negation would hide exactly the people the words include. A saved view stores its filters, its visible columns, its sort and whether it draws as a table or a kanban, and the same stored question resolves as a campaign audience through the same SQL builder. Audiences go further on Growth and Scale: one builder over the CRM fields, the Stripe cohorts (plan, MRR, sign-up date, when a customer started paying, when they churned) and product activity (activation, last activity, last product event), with a live count, saved once and offered to every campaign and sequence. A "like my best customers" audience is computed from the traits your best paying customers share, on your own data rather than on third-party data, and Canopus can draw an audience from each segment of your strategy for you to accept. On Scale, an ads proposal can target a saved audience too.
Who moves a card is a rule, and the rule is that you do. A prospect changes stage because a person drags it or uses the keyboard menu, or because your product sent an event saying the person converted; nothing moves one on a timer, and there are no tasks, deals or stage automations to second-guess you. There is one prospect pipeline with configurable stages (New, Contacted, Demo, Proposal, Won, Lost by default, between 2 and 30 of them, each with a colour from a closed palette), and the customer table is a grid rather than a second board. Duplicate contacts are never merged by a guess: a conflicting email is counted and reported so you decide. CSV export out of Canopus is listed as out of scope; the data is yours through the views and the signed webhook you already own.
Emptying the CRM is a danger-zone action. It erases contacts, their timelines, the columns, the views and the stages, then seeds the defaults back inside the same transaction, so a failure partway leaves the organization exactly as it was rather than erased with nothing put back. An administrator confirms it by typing the organization slug, and the audit log records the counts, so "I emptied the CRM" is answerable later with how much was in it. Two things survive deliberately: the public form key, because regenerating it would silently break a landing page nobody touched, and the log of webhook events already received, which is what makes a replay work. Admins rather than owners, because admins are the people who run an import, and a bad import is the commonest reason to want to start over.
Questions founders ask
See your own numbers first
Connect Stripe and your real revenue is on screen in five minutes, before an agent has said a word. Create your account, answer the questions, and your strategy gets written.