Skip to content

The right email when a customer signs up, stalls or leaves

Canopus is an autonomous growth manager for SaaS that sends lifecycle email sequences triggered by what happens to each customer, from your own domain, once you have approved each sequence.

A campaign is one send to a list. A sequence is a standing instruction: when someone signs up, reaches the last two days of a trial, starts paying, leaves or goes quiet, they enter it, and they leave it the moment it stops making sense for them. You approve the wording once. The mail your product sends on its own, a receipt or an alert, goes out from the same domain through an API.

How does a lifecycle email sequence work?

  1. Canopus notices what happened to each contact

    Every day, each contact in the CRM is compared with what Canopus already knows about them: a new sign-up, a first activation, a trial with two days left, a trial that ended without a payment, a first payment, a churn, three days without activity. Each moment comes from data the CRM holds, filled by your product events and Stripe.

  2. An agent writes the sequence, and you approve it once

    Canopus proposes one of five shapes known to work: a five-email onboarding, a reminder two days before a trial ends, a nudge for a trial that never started, a win-back after churn, and an upgrade note for an active customer still on the small plan. It is written in your voice, for everyone who will ever enter it. You read it, edit it if you want, approve it, and it runs.

  3. People enter continuously and leave on an exit condition

    Somebody who signs up overnight gets the first email on the next run. Each sequence declares what makes it pointless: the trial reminder stops for anyone who pays or leaves, the nudge stops when the person activates, the win-back stops when they come back. The onboarding keeps running after an upgrade, because a new customer still has the product to learn.

  4. One email a day per person, across every sequence

    When two sequences want the same person on the same day, your ranking decides which goes first, and the other waits a day without losing its place. An address that bounced hard or complained leaves every sequence, whatever else was set.

  5. Your product mail goes out from the same domain

    Receipts, alerts, reports: your code calls POST /api/v1/emails and the message leaves from a tx subdomain of yours, on a stream whose reputation stays apart from your campaigns. The API is documented at /docs/email-api.

What does this replace?

  • A sequencing tool with its own copy of your contacts to keep in sync

    Sequences read the CRM Canopus already keeps, so there is no second list to reconcile.

    Replaces: A sequencing tool
  • Automations that fire an email when a Stripe event arrives

    Trials, payments and churn reach the CRM from Stripe, and the sequences read them there, with nothing to wire.

    Replaces: An automation tool
  • Remembering to take a customer out of the trial reminder after they paid

    An exit condition does it, checked once more right before each email leaves.

    Replaces: Manual list hygiene
  • A second email service, billed apart, for receipts and alerts

    POST /api/v1/emails sends your product mail from your own domain, inside the same monthly quota.

    Replaces: A transactional email service
  • A new customer receiving three emails in one day from three automations

    One sequence email per contact per day, whichever sequence it comes from.

    Replaces: Overlapping automations

What decides who enters, who leaves, and what is sent?

The moments a sequence can react to are a closed list of eight, and every one is backed by data the CRM already holds: the contact appeared, the contact started a trial, the contact activated, the trial ends in two days, the trial ended without a payment, the contact started paying, the contact churned, the contact has been silent for three days. There is no free-form trigger on purpose. A condition no event can ever satisfy fails in silence: the sequence looks active and never sends, which is the hardest failure there is to notice. Two of the eight are windows rather than instants. A trial in its last two days, or a trial silent for three, is true across several days, and Canopus records it once per day so the same condition cannot enrol somebody twice. The three days are the same three days the CRM uses for its stalled-trial segment, so a view and a sequence never disagree about who counts as stalled.

A sequence is approved as a template, not send by send. It writes to somebody at three in the morning when they sign up, so nobody can be there to approve each email; what you approve is the wording everybody who enters will read. The agent leaves a sequence proposed, and only a person switches it on. Any edit after that sends it back to proposed and suspends sending, without taking anybody out: the people already part way through keep their place and resume when you approve the new wording. There is no A/B test in a sequence. People arrive in a trickle rather than all at once, and splitting a trickle is a different kind of test from the one a campaign runs.

Exits are declared per sequence because the same trigger wants different exits. Two sequences can both start at sign-up: the onboarding should keep teaching somebody who pays on day two, while a reminder about the end of a trial becomes absurd the moment they pay. So each shape carries its own list, and you can change it. The exits are applied before the run sends anything, and then once more immediately before each individual email, because the whole point of an exit is the person who upgraded two minutes after their step came due. An email cancelled in error is invisible. An absurd email that was sent is read.

Frequency is decided per person rather than per sequence. Canopus sends at most one sequence email to a contact in any day, and the question is asked about the contact, because the collision worth preventing is two different sequences reaching the same person on the same morning. Sequences are walked in the order you ranked them, so your ranking decides which one gets the day, and the other is pushed back without losing its position. Sequence sends count in the monthly email quota of your plan, alongside campaigns. An address that bounced hard or complained, on any stream, is taken out of every sequence whatever exits were set.

The mail your product sends itself, a password reset, a receipt, a finished export, goes through POST /api/v1/emails. It leaves from a tx subdomain of your sending domain, with a few more DNS records to publish, so its reputation lives apart from your campaigns and a bad campaign cannot hurt your receipts. A key belongs to one project, is shown once and stored hashed. Every request can carry an Idempotency-Key, and a retry with the same key returns the original send instead of sending twice. The limits are 120 requests a minute per key for sending and 300 for reading. Transactional sends count in the same monthly quota, with one difference at the line: campaigns stop at the quota, product mail keeps going in a tolerated overage at no charge, with a notification, up to twice the quota. The full reference, with every error code, is at /docs/email-api.

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.