Product kinds
A Product is what a Customer gets from a Startup, with a promise; every Product is one of four kinds, supplied, bought, and earned in declared ways.
- Read as
- GuideSpec
A Product is what a Customer gets from a Startup, with a promise the Customer can rely on. Every Product is one of four kinds, and declares who supplies the work, who buys it, and how it earns.
The promise comes first
Three Customers can send the same request to the same API and expect different things. One buys facts, one buys a tool, and one buys a finished job. A Product's kind names which promise the Startup makes.
| Kind | The promise | Built by | Example |
|---|---|---|---|
| Data, sold as a Dataset | The records are true. | SoftwareBuilder | Verified building permits for roofing contractors. |
| Software | The tool works. | SoftwareBuilder | Lesson booking for independent music teachers. |
| Agent | The task passes, or it is redone or free. | SoftwareBuilder | Month-end close for design studios, in their own accounting tool. |
| Service | The outcome arrives, or the remedy applies. | ServiceBuilder | Certified legal translation, redone free if review fails. |
The line between an Agent and a Service is where done lives. An Agent is graded on the buyer's task, in the buyer's tools. A Service holds the work until it is done, under an Outcome Contract that names acceptance and the remedy for a miss.
A Product can be several kinds at once. A Dataset served through an API is also Software, and it keeps the promises of both.
Who supplies the work
Supply is digital or human. Digital supply is infinite: code, models, and Agents do the work, so a digital Product needs only demand, while human supply is finite.
Each Human Supplier is recruited, checked, credentialed, and paid out, and the Startup that sells their work runs supply acquisition beside demand. A notary marketplace cannot sell more sessions than it has commissioned notaries in a state.
Where the law reserves an act to a license holder, the step is a Reserved Act and its Function is always human. A Product that needs one goes Live only once the Credential class is staffed.
How a Product earns
How a Product earns is a second axis, set on its Offer. Any kind pairs with any earning model.
| Earning model | The Customer pays | Example |
|---|---|---|
subscription | Each period, for access. | $29 a month per music teacher. |
usage | Per call, record, or task, counted by a Meter. | Two cents per verified permit. |
commerce | Once per item, at checkout; an upstream supplier holds the stock and ships it. | A split ergonomic keyboard, dropshipped. |
marketplace | A Supplier directly, while the Startup takes a share. | Nonprofits hire grant writers; the Startup keeps 12 percent. |
directory | Nothing; listed Suppliers pay to appear. | Transcription vendors pay $99 a month for a verified listing. |
Every Offer declares the Event that earns its revenue: a Dataset or Software earns on Action.done. An Agent earns on the Evaluator verdict that its task passed. A Service earns when the Customer accepts the work, or when acceptance is deemed after its window.
The warranty test separates a Service from a marketplace. If the Startup warrants the outcome, it sells a Service; if the Customer contracts with the Supplier directly, it runs a marketplace.
Who buys
The buyer is a person, a business, a developer, or an Agent buying for one of them. A Motion names the pair on each Offer: b2b, b2c, b2d, b2a, and the chains such as b2a2c, where an Agent buys for a consumer.
The Motion changes who the Customer is, which Interfaces lead, and when a purchase needs an Approval or a Mandate. An Agent buyer reads the Offer's machine contract, asks for a quote, and checks out inside its own spending Limits. It never needs the Site.
Product Templates
A Product Template is a reusable definition of one kind of Product. It fixes the kind, the supply, who does the work, the Interfaces, the buyer, the earning model, and the assurance.
| Product Template | Kind | Supply | Buyer | Earning model |
|---|---|---|---|---|
| Dataset API | data | digital | human-or-agent | usage |
| Headless SaaS | software | digital | human-or-agent | usage |
| SaaS | software | digital | human | subscription |
| Directory | data | digital | human-or-agent | directory |
| Marketplace | software | human | human-or-agent | marketplace |
| Licensed-human marketplace | service | human | human-or-agent | marketplace |
| Agent as a Service | agent | digital | human-or-agent | usage |
| Services-as-Software | service | digital | human-or-agent | usage |
| E2E managed service | service | human | human | subscription |
| Dropship store | software | digital | human-or-agent | commerce |
A Startup Template combines Product Templates with an operating model and the Motions they sell through. The catalog lists every combination, with what each adds to its base.
Offers make Products available
A Product is never sold bare. An Offer makes it available under a Brand, at a Price, on terms, and one Product can carry several Offers for different buyers. Free is a Price: a $0 Offer still has terms.
Accepting an Offer grants Entitlements: a Verb on a Noun in the Customer's Tenant, bounded by a Limit. An Order keeps the Offer Version it bought, so a later price change never rewrites a past sale.
Products compose
A Service uses an Agent, the Agent uses Software, and the Software reads a Dataset. Each link is a builtFrom Relationship, shown on both Products. What a Product draws on but did not build is a Source: owned, wrapped, or recruited.
Coverage, cost, and assurance roll up the chain. The translation Service's assurance is no stronger than its weakest step, and a remedy runs down the chain, so the upstream Agent's remedy settles first.
Next
Operating models: who runs the Startup, and what changes as humans join.