A strategy you can check, line by line
Canopus is an autonomous growth manager for SaaS that writes your marketing strategy from Stripe, PostHog, your site and your own answers, shows the evidence under every claim, and turns it into a month of planned work once you validate it.
You answer twelve questions. Canopus reads your revenue, your product events, your site and up to five competitors, drafts the strategy, has a second agent attack the draft, then rewrites it. You validate it block by block. A calendar for the month comes out of it, and when the results say the plan is wrong, you get a revision to accept or decline.
How is the strategy written?
You answer twelve questions
Your site address comes first, so the scan reads the site while you answer the rest: the name, the price, the stage, what the product does, the problem it solves, what it replaces, who it is for, their pains, their objections, the channels you already use and the objective you want. The scan pre-fills what a site can say, and you correct it. Each answer is saved as you move to the next screen.
Canopus reads your numbers and your market
Stripe gives revenue, plans, trials and churn, and from them the revenue per customer, the lifetime value and the retention. PostHog gives sign-ups, activation and the events people actually trigger. The crawl gives the product brief and the three audit scores. Up to five competitors are read on their own public pages, each claim kept with the address it was read on.
A second agent attacks the draft
A critic reads the first draft and quotes the sentences that are generic, unsupported, unfeasible or invented, each with what the rewrite has to do about it. The strategy is rewritten once against that list, and you can read the critique next to the corrected version.
You validate it block by block
Four blocks: positioning, target, channels, objective. Under each one, a panel shows what it was written from. You tick each block or edit it first, and the adopt button waits for all four. The adoption is what is recorded, with your name, in the audit log.
Ideal customerVersion 2Proposed from your product and your existing customers- Solo founder, first paying usersShipped the product, now has to find the people who want it.PainObjectionValue
- Two-person team, flat revenueRevenue stopped moving and nobody owns distribution.ValidatedPainObjectionValue
- Agency owner, productisingTurning a service into a subscription and starting from zero.PainObjectionValue
A month of work comes out of it
The calendar projects the strategy over the coming month. Each slot names a channel, a format and a segment, at the cadence the strategy set. Campaigns and posts are written for those slots, and each one still waits for your approval before anything leaves.
Channel scores9 scoredChannelScoreEmailStart hereRuns it92Search contentStart hereWrites it84LinkedInWrites it71CommunitiesAdvises64XWrites it48Meta AdsWrites it31The results can reopen it
Three rules, not a model, decide whether a revision is due: the objective behind for three weeks running, an experiment on one of its channels dropped in the last thirty days, or one channel’s slots skipped three times in four weeks. You get at most one proposal a week, and you accept or decline it.
ReportCampaign closedObservationsTwo campaigns finished and one is still running. The shorter subject line did better; the pricing page post brought the most trials.
Signups38Trials started14New monthly revenue€410Proposed revision
Less LinkedIn, a monthly newsletter
- LinkedIn2× a week1× a week
- Newsletteradded1× a month
LinkedIn posts have not moved trials in six weeks; the newsletter list is growing on its own.
AcceptedDecline
What does this replace?
Paying a consultant for a strategy deck that is out of date by the next quarter
A versioned strategy, reopened when your numbers say so, with every earlier version kept.
Replaces: A strategy consultantWriting a marketing plan from what you hope is true about your customers
Every block sits on a panel that shows the data and the answers it was written from.
Replaces: A plan from memoryReading competitor sites over a weekend and forgetting what each one charged
Up to five competitors read on their own pages, each claim with its source and its date.
Replaces: A folder of open tabsAsking a chatbot for a growth plan and getting the one everybody gets
A critic quotes the generic lines, and the strategy is rewritten against them before you see it.
Replaces: A generic promptDeciding every few days what to post and what to send, from nothing
A month of slots, each with its channel, its format and its segment.
Replaces: A content calendar sheet
What is in a Canopus strategy, and where does each part come from?
A strategy is one document with a fixed shape, and its bounds are tight on purpose. A positioning of a few sentences. One to three segments, ranked, each with its pains and its objections. One to three channels, each with a cadence, the formats it starts with, the reason it was chosen and a signal to watch. One objective: a metric (sign-ups, trials, activated users, MRR or visitors), a target, the baseline it is measured against and a horizon of four to twenty-six weeks. Up to five hypotheses, each with the evidence behind it. And, once competitors have been read, the landscape and what your product holds against them. Three segments and three channels is the most one founder can hold in mind and still disagree with; six of each would be a shopping list, not a decision.
The objective is yours, not the model’s. The metric, the target and the horizon you chose during onboarding are written into the strategy by the code after the agent returns, so no draft can quietly lower the bar it will be judged against. The baseline is measured on your own snapshots, because fifteen more sign-ups means nothing without the number it is fifteen more than, and a later revision needs that number to tell progress from noise. The signal attached to a channel is a threshold to watch rather than a promise: Canopus measures what it sends, but a LinkedIn post is published by you and its results are typed in by hand, and a number cannot commit somebody else’s keyboard.
The quality of a strategy comes from what it is given, so the inputs are a floor rather than a nice-to-have. The onboarding answers are read by every agent as your own words, ahead of anything the crawl found: a site says what the product claims, you say what it is for. With fewer than three answers the strategy refuses to start, because a strategy written on nothing is generic by construction. Stripe and PostHog are read with keys that cannot write. The crawl gives the product brief and the audits. A library of expertise travels with the prompt: go-to-market archetypes, thresholds that separate a healthy figure from a worrying one, and annotated examples of strategies that hold and strategies that do not.
Competitors are read, not imagined. You name up to five, or keep the ones Canopus suggests, and each one’s public site is crawled. Their positioning, their promise and their pricing are stored with the addresses they were read on, and anything the crawl did not show is listed as a gap instead of being guessed. The analysis carries its date, because a price read in September is a fact about September. What comes out lands in the strategy as the landscape, with the alternatives named, and up to five differentiators, which are what the campaign and content agents lean on when they write against an alternative.
Two mechanisms keep the draft honest, and you can see both. The critic lists at most eight flaws, worst first, and each one quotes the draft word for word, says why it fails and states what the rewrite must do; an empty list is a real result and means the draft stood. Then the evidence panel: under each block, a fold-out lists the exact version of the ideal customer, the channel scores and the audits that this version of the strategy was written from, by identifier, never whatever is newest today. A version you corrected by hand inherits its parent’s evidence, and the panel says so rather than passing an older reasoning off as the current one.
Once adopted, the strategy is the only thing the producing agents read, so a campaign, a post and the calendar cannot each assemble their own idea of your customer. The calendar projects the strategy over a month at each channel’s cadence and rotates the formats it named. Revision is decided by rules, because a model asked whether a plan should change says yes often enough to stop being read. The three triggers are checked in a fixed order: the objective first, since it is the only thing the strategy promises, then an experiment that was tried and dropped, then slots that were never done. Signals on channels outside the strategy are ignored, and a revision always reaches you as a proposal you accept or decline.
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.