Every test you run should pay for the next one.
Ideas are cheap. Learnings compound.
Oddyssey takes a hypothesis, builds the creative and funnel around it, runs the test on the platform that fits, and ties the result back to the one thing you changed. When you plan the next test, what the last one taught you is still there.
- One workspace per brand
- Claims kept with their evidence
- One experiment, one declared variable
- Read out on margin, not clicks
- Declared variable
- Angle: cost of inaction
- Held constant
- Offer, audience, funnel
- Decision rule
- Set before launch
- Evidence
- CL-032 approved, NAM
- CL-032
- Approved, NAM, expires 2027-01
- CL-088
- Conditional, disclosure required
- Unregistered
- Held back from publish
- Contribution margin
- Versioned assumptions
- Allowable CAC
- By offer and cohort
- Payback
- Net of refunds
- Campaign
- Created paused, read back
- Budget ceiling
- Set per policy
- Naming
- Consistent, ID-linked
Your tests are running. Your learnings are not.
More tests every quarter, less to show for each one. Not because the tests are bad, but because nothing carries from one to the next.
- The research behind the last campaign sits in a doc nobody opens before writing the next brief.
- Creative gets judged on click-through in the ad account. Whether it made margin gets worked out in a spreadsheet, weeks later.
- Six months of tests and nobody can say why the winner won. So the next test starts from zero.
Go wild on the idea. Stay strict on everything after it.
One chain from the question to the decision. Each step hands the next one the brief, the claims, the budget and the variable it needs.
- 01
Hypothesis
A bet you can state in one sentence, with the reasoning and the evidence written down.
- 02
Grounded experiment
One variable changes, everything else stays fixed, and the decision rule is set before you spend a dollar.
- 03
Creative and funnel
Ads and pages built from the approved brief, inside your brand rules and your approved claims.
- 04
Tested where it belongs
Live on the platform you pick, or the one the numbers point to, inside the budget ceilings you set.
- 05
Attributed to the variable
The result ties back to the one thing you changed, read on margin after refunds, not on clicks.
- 06
Decision and next iteration
What you learned is written down with its limits, and it becomes the starting point of the next test.
Step 06 feeds step 01. Your next test starts from what the last one proved, not from a blank page.
You already pay for parts of this. Oddyssey is the part that connects them.
No knock on these tools. A writing tool can't check a claim, a dashboard can't change a funnel, and an agency takes its learnings with it when the contract ends. That's the gap this fills.
Claims stay attached to their evidence, market and expiry
- AI copy tools
- Writes any claim you prompt for
- Analytics dashboards
- Does not model claims at all
- Creative agencies
- Kept in a shared doc, if at all
Offers priced with costs, fees and refunds beside the copy
- AI copy tools
- No cost model
- Analytics dashboards
- Reports revenue, not contribution
- Creative agencies
- Priced in a deck, reconciled later
Briefs start from the strategy you approved
- AI copy tools
- Starts from a blank prompt
- Analytics dashboards
- Not in scope
- Creative agencies
- Written per project, format varies
Variants recombined inside rules you set
- AI copy tools
- Free-form output, no constraints
- Analytics dashboards
- Not in scope
- Creative agencies
- Versioned by hand
Pages, checkout and lifecycle versioned together
- AI copy tools
- Not in scope
- Analytics dashboards
- Reads the funnel, cannot change it
- Creative agencies
- Built in a separate page tool
Campaigns created paused, read back, inside ceilings
- AI copy tools
- Not in scope
- Analytics dashboards
- Read-only on ad objects
- Creative agencies
- Built by hand in the ad account
Spend reconciled against orders, refunds and disputes
- AI copy tools
- No measurement layer
- Analytics dashboards
- Platform-reported metrics
- Creative agencies
- Monthly reporting cycle
Decision rules set before launch; more data needed is a valid read
- AI copy tools
- No decision model
- Analytics dashboards
- Dashboards, you decide
- Creative agencies
- Judgement call, undocumented
Learnings recorded with their limits and scope
- AI copy tools
- Nothing persists between prompts
- Analytics dashboards
- Charts, not learnings
- Creative agencies
- Leaves when the contract ends
The same chain, down where the work gets done.
Eight stages, one record. The IDs you set in the strategy are the IDs you report on, so nothing gets lost moving between tools.
- 01
Research
Sources, observations and inferences kept apart, each cited.
- 02
Offer
Price, terms, costs, proof and eligibility as one versioned object.
- 03
Strategy
The bet, the one variable and the limits, written down before spend.
- 04
Creative
Briefs carrying the objective, the angle, the claims and what stays fixed.
- 05
Funnel
Pages, forms, checkout and lifecycle versioned together.
- 06
Campaign
Created paused, read back, inside budget and bid ceilings.
- 07
Measurement
One event vocabulary, identity carried through, margin reconciled.
- 08
Learning
A record with its evidence, its uncertainty and where it applies.
Stage 08 feeds stage 01. The next idea has to beat what this test proved.
AI proposes. Plain checks verify. People approve what matters.
The agents take over the work that's approved, repeatable and low risk. New offers, new claims and spend above your limits stay with a person.
Brand truth and claims
Positioning, voice, visual rules and assets with owners and approval status. Approved, conditional and prohibited claims carry their evidence, markets, disclosures and expiry. Unregistered claims are held back from publish.
Offer and pricing design
One-time, service, subscription, membership, bundle, tier pricing, trial, lead offer, marketplace and hybrid. Costs, fees, refunds and fulfillment sit next to the customer-facing copy.
Creative production
Briefs built from the strategy you approved, message matrices across awareness stages, modular hooks and bodies recombined inside your rules, and review against brief, brand and claims.
Review and history
Schema checks, plain business rules, editorial review, technical QA and human approval. A durable history of AI output, human edits, approvals and configuration changes.
Funnel and checkout
Building blocks for landing pages, quizzes, applications, booking, carts, bumps, upsells, downsells and post-purchase, with market and language variants.
Campaign deployment
Approved experiments become campaigns, audiences, creatives and tracking with consistent naming. Created paused, read back and compared. Budget ceilings and approval gates respected.
Measurement and economics
One event vocabulary, platform events mapped and unknown events set aside, identifiers carried through, spend reconciled against orders, refunds, disputes and CRM outcomes.
Decisions
Sufficiency, winner, loss and guardrail rules set before launch. More data needed is a valid read. Learnings state their uncertainty and where they do not apply.
The rules hold whether you're watching or not.
These aren't values on a wall. They're checks in the product, and they're why a result from six months ago is still usable.
AI proposes, checks verify
Model output is a proposal. Schema checks and plain business rules decide whether it can exist, and neither one reads model confidence.
Evidence before assertion
A public claim resolves to an approved claim record with a source, a market and an expiry, or it waits in review.
One experiment, one declared variable
What stays fixed is settled before launch, and so is the decision rule. More data needed is a result you are allowed to record.
Read on margin, not clicks
Spend is reconciled against orders, refunds and disputes, so a promising result is judged after the money settles.
A full marketing department, run by agents that hand work to each other.
Each agent does one job and passes a complete package to the next. They propose, plain checks verify, and people approve new offers, new claims and any spend outside your limits.
Marketing strategist
Turns what you already know into a hypothesis with one declared variable and stated limits.
Hands over an approved experiment definition.
Marketing research
Gathers sources, keeps observation apart from inference, and cites everything it uses.
Hands over evidence the hypothesis can stand on.
Content editor
Produces and edits image, video and animation variants inside your brand rules and approved claims.
Hands over assets tied to the brief they came from.
Funnel builder
Assembles pages, forms, checkout and post-purchase steps as one versioned funnel.
Hands over a funnel that matches the creative promise.
Media buyer
Deploys to the platform you pick, or the one the workspace recommends, inside budget and bid ceilings.
Hands over live campaigns with consistent identifiers.
Attribution analyst
Reconciles spend against orders, refunds and disputes, and ties the result to the variable you changed.
Hands the learning back to the strategist.
Content creator
Coming soonAI models and influencer-style presenters generating on-brand video and social content.
Not available yet. Nothing on this page depends on it.
Swap a vendor by remapping fields, not rebuilding your stack.
Read and write scopes stay separate, mappings can be previewed, and sync health and last success are visible. Card data stays with compliant providers.
Commerce
Catalog, inventory, orders, customers, discounts, subscriptions, fulfillment
Payments
Checkout sessions, payment state, refunds, disputes, payouts
Advertising
Campaign objects, creative, audiences, spend, delivery metrics
CRM and sales
Leads, contacts, opportunities, qualification, calls, appointments
Analytics and warehouse
Events, identities, attribution inputs, cohorts, reporting
Communication
Email, SMS, messaging, push, call tasks, support
Creative tools
Asset storage, templates, video, image generation, DAM
Identity and security
SSO, SCIM, role management, secrets, audit export
Start free. Upgrade when the workload does.
Pricing is confirmed with you before you commit. No usage surprises mid-month.
Core
One brand, one market, the full loop end to end.
- Brand truth and claims
- Offer design and unit economics
- Creative briefs and copy variants
- Funnel builder and checkout blocks
- Experiment records and learnings
Plus
Multi-market work with reviewed deployment.
- Everything in Core
- Market and language variants
- Ad account deployment with budget ceilings
- One event vocabulary and reconciliation
- Approval workflows and change history
Growth
Portfolio-scale testing with automation limits.
- Everything in Plus
- Multi-brand workspaces
- Decision rules and fatigue alerts
- Automation limits with spend ceilings
- Portfolio reporting
Enterprise
Controls, residency and support for larger orgs.
- Everything in Growth
- SSO, SCIM and role management
- Data residency options
- Custom connectors and API access
- Named support
Straight answers.
What does Oddyssey replace?
The research doc, the brief template, the creative tracker, the page builder, the test log and the margin spreadsheet. It does not replace your ad accounts, your store or your CRM. It connects to them.
What will it not do?
It is not a merchant of record, a payment processor or an ad network. It will not let software publish claims, change prices or spend budget outside the limits you set, and it will not call a result certain when the data cannot support it.
How is my data separated?
Each workspace is kept separate. Reusable frameworks can move between them, confidential evidence cannot. Sensitive fields can be excluded per destination and history has retention controls.
How do automation limits work?
You set limits for pausing, budget reallocation, variant generation and reusing proven patterns, each with spend and scope ceilings. New offers and new claims stay with a person, and stop is always available.
Can AI invent a claim?
No. A public claim has to resolve to an approved claim record or it goes to the review queue. Unknown claim IDs fail the check, and approval is never inferred from model confidence.
Know why your next winner won.
Start with one brand and one offer. The chain works the same when you're running fifty.