Two channels, and the reason the other eight were refused
Canopus is an autonomous growth manager that scores ten acquisition channels against a SaaS product’s validated ideal customer and recommends the two or three worth starting.
The two or three it recommends become the channels of your strategy. Every channel is also labelled with what Canopus does about it. Email and the newsletter are run from here. LinkedIn, X and SEO and GEO content are written here and published by you, and on Scale ads are proposed here and launched by you; the approach to a creator and the brief that follows are written here and sent by you. Communities, YouTube Shorts and TikTok are scored and argued for, and you decide.
How does it pick a channel?
It waits for a validated customer
The strategist refuses to run without an ideal customer profile a human has validated, not merely one that was proposed. A channel strategy is the first thing that costs you real time to act on, and basing it on an audience nobody agreed to would be the machine choosing the customer and then choosing where to find it.
It reads your product and your numbers
The product brief, the validated segments with their pains and triggers, and aggregate evidence from your own data over the last 84 days: how many customers are paying, trialing or churned, the median and top-decile revenue, the countries, the referring domains. Percentages are only reported when the cohort is at least ten people.
It scores all ten and recommends two or three
Each channel gets a score out of 100, a rationale, a cost note, an effort level of low, medium or high, and a first step. The recommended ones are grouped under "Start here" and the rest under "Not now", so the ranking reads as a decision. Recommending five is the same as recommending none: a founder’s attention is finite, and a channel worked half-heartedly returns nothing. The summary says what was ruled out and why.
It attaches formats you can start with
A recommended channel carries up to three ids from a library of 32 formats, so the advice is actionable on the day it arrives. "LinkedIn" is not advice. "LinkedIn, starting with a before-and-after post, and here is the shape of one" is.
What does this replace?
A growth consultant’s channel workshop, billed by the day
Ten channels scored against your validated customer, with the rejections written down.
Trying four channels at once and abandoning all of them by March
Two or three, chosen with a stated reason, and a first step for each.
A page of marketing ideas nobody turns into a task
A first step that has to be doable now, naming the thing rather than the discipline.
Replaces: An ideas pageCollecting swipe files of posts and emails that seemed to work
A library of 32 formats, each with the condition under which it is the right choice.
Finding out far too late that the recommended channel needs an agency
Every channel labelled run, written or advised before you commit to it.
What does Canopus run, what does it write, and what does it only recommend?
Three levels, and the difference between them is operational rather than cosmetic. Email is run by Canopus: the producer writes the sequence, you approve it, and it is sent from your own domain, with tracked clicks, one-click unsubscribe and a sending cap. The newsletter is at that level too: Canopus drafts each edition, you approve it, and it leaves from the same domain. LinkedIn, X and search content are written here: the content producer writes posts under your name, Canopus drafts the SEO and GEO pieces of an editorial plan drawn from your latest audit, and you publish both yourself, on your networks and on your own site. Creators sit at the same level: Canopus writes the approach message and, once the creator has said yes, the brief with the code and the exit, and you send both and type back what came of it. Ads sit at that level too, on Scale: Canopus proposes the campaign for Meta, Google Search or LinkedIn with the targeting, the copy, a hypothesis and a budget under the monthly cap you set, and you launch it yourself on the platform with a tracked link, so no spend happens without your explicit approval. YouTube Shorts sit at the second level too: Canopus writes the script, with the hook, each beat timed with what is shown, written on screen and said, and the end shot, and you film and post it. One post’s idea can be turned into an X thread and a script in one click. Communities and TikTok are advised: the strategist scores them, argues for them and names the first step, and you carry them. The strategist is given these labels with the scores and is told to say in the rationale which level a recommendation belongs to. A strategy that recommends Meta Ads without saying so would be a promise the product cannot keep.
A channel is scored against your customer, not against SaaS in general. The question is never whether LinkedIn works but whether this person is on LinkedIn, in a frame of mind where this matters to them. If the ideal customer work found a specific community, that outranks any general truth about channel performance. Price and motion have to agree: a product at 19 euros a month cannot support paid acquisition at B2B click prices, and a product at 2000 euros a month is not bought from a TikTok, and saying so plainly is often the most useful sentence in the strategy. What you already have counts too: an existing audience, a written backlog, an active presence somewhere changes the ranking more than any channel’s theoretical ceiling, because cold-starting a channel is most of its cost.
This score is a judgement and the product says so. It is produced by a model, unlike the three audit scores, which come from deterministic rules precisely so their history means something. The difference is what is being measured. An audit asks whether a title tag exists, which is true or false independently of anyone’s opinion. A channel score answers whether an audience is on TikTok in a buying frame of mind, and there is nothing to measure. Inventing weights over invented sub-scores would produce a number with no meaning and the appearance of one. So the score is rendered small, the rationale large, and the recommended channels are separated visually rather than merely ranked, because the whole output is "these two, not those eight".
The list is closed at ten, on purpose. Every entry is a place a B2B SaaS plausibly finds customers, and a longer list would produce a longer scoring table and no more decisions, because a founder is going to pick two. Both catalogues, the ten channels and the 32 formats, live in code rather than in database tables. The reason is ordinary for formats: the library is static, has no per-organization state, and is read by agents and by the interface, so a table would mean a seed migration plus a synchronisation problem every time it changed. Adding a channel or a format is a pull request rather than a row, which is the right friction, because each one is a claim about how a SaaS funnel works.
A format is a shape, not a template: the producer writes the copy, the format says what shape it takes and when that shape is the right one. Thirty-two of them, spread across the channels. Nine for email, including a five-day onboarding sequence, a trial-ending message two days out, a win-back after churn that only fires if something has actually changed, a plain-text question from the founder with no signature block, and a referral ask to a customer who has stayed, with a tracked code and no invented reward. Six for LinkedIn, including the mistake you made, a teardown of something public, and building in public with real numbers. Three for X, six for search content, including a comparison page against a named competitor, an alternatives-to-the-incumbent page and your landing page rewritten to be pasted, each change tied to the audit finding it answers. Two for creators, the approach message and the brief. Two for communities, and four for short video and paid. Each one carries the condition under which it is the right choice and the beats in order, which is what the campaign and content producers are told to follow. Each format’s prose is in English because its reader is a model; only the name is translated, because the name is the only part you see.
Two things you decide. The channel strategy needs a validated ideal customer and a product brief, so it will refuse to run before those exist, which is deliberate rather than a missing feature: you validate the customer, then the strategy is written. And a written channel stays yours to publish: a post goes out under your name, you type the results back, and attribution joins on the contacts you already hold, and the strangers a post or an ad brings are counted by the source they arrive with. The strategy says which of the two or three channels runs on its own, which waits for your approval, and which is a recommendation you carry, so the decision you make on reading it is an informed one.
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.