Strategy, Goals, and economics
How a Startup turns one number into dated Goals, chooses Plays to reach them, and judges each choice by the unit economics it produces.
Strategy is the set of choices a Startup makes about whom to serve, what to offer, and where to spend. Goals state what those choices must produce; Metrics report whether they did.
One number at the root
Every Startup starts from one number: its minimum success criteria. It is a Goal with a single Key Result, outside revenue by a date, and its default comes from a Studio Setting.
parsepaper.dev sells an API that turns scanned invoices into structured records, priced per page to developers. Its Founder sets $5,000 a month in outside revenue within six months.
The number back-solves into a plan before anything is built. Each step divides by a rate, so a weak rate shows up as an impossible volume.
| Step | Rate | Needed |
|---|---|---|
| Outside revenue | The minimum success criteria | $5,000 |
| Paying developers | $40 average spend each | 125 |
| Trials | 4% of trials convert to paid | 3,125 |
| Visitors to the docs | 10% start a trial | 31,250 |
If no affordable mix of Plays brings 31,250 developers a month to the docs, the model fails on paper, for free. If one does, each row becomes a dated Key Result, and together they chart the path to the number.
Four records kept apart
A finished task list is not a Goal met. Four distinct records keep the gap visible.
| Record | What it holds | parsepaper.dev |
|---|---|---|
| Goal | An objective with one or more Key Results | Make self-serve signup repeatable |
| Key Result | A Metric, a baseline, a target, and a deadline | Paying developers, from 40 to 125, by the end of the quarter |
| Metric | A measure computed from Events or Entries | Paying developers in the trailing 30 days |
| Task | One piece of work | Publish the quickstart for the SDK |
Publishing a quickstart is work; more paying developers is the result. A Metric is always derived from recorded Events, never typed in, so nobody meets a Key Result by editing a number.
Choices become Decisions and Plays
parsepaper.dev chooses developers at small accounting software companies over enterprise finance teams. That one choice sets which demand to pursue, how much to invest in support, and which Evidence matters.
Each choice is a Decision: the options, the choice, the reason, and who decided. It stands until a later Decision supersedes it, so strategy changes by record, never by drift.
A Play turns a choice into repeatable work. It names what to do, the Metric it moves, and its cost, and the Startup runs it as a Process. parsepaper.dev runs three demand Plays that cost no cash: Developer relations, Open source parsers for two common invoice formats, and Agent registries.
A long-tail Startup spends $0 on demand by default. A founder can fund a demand budget at any amount. That spend is an investment Decision with a target and a stop rule, and it runs under Limits.
Economics of one unit
Every Startup knows what one unit of its work earns. For parsepaper.dev, the unit is 1,000 pages parsed.
| Line | Amount |
|---|---|
| Fee | $12.00 |
| Model inference | −$3.90 |
| Storage and compute | −$0.50 |
| Retries on unreadable scans | −$0.70 |
| Payment fees | −$0.65 |
| Support | −$0.35 |
| Contribution | $5.90 |
Each line can sink the business on its own. A cheap model that misreads totals turns every saving into retries and refunds.
The payback test joins acquisition to contribution. A developer spending $40 a month contributes about $19.70. If a paid probe wins developers at $60 each, payback takes three months; Repeatable requires one month or less.
Estimates and readings
Before launch, every number above is an estimate inside a Hypothesis. After launch, readings replace estimates line by line.
Each figure keeps its provenance: estimate or reading, with its source and window. A margin that blends the two hides which half is wrong.
A Metric names its definition, population, window, and source. Without them it answers the wrong question with false precision.
Experiments decide
Every Change to an Offer, Price, Page, message, or Play can run as an Experiment. It varies copy or behavior through Flags and judges the variants by one primary Metric, with guardrail Metrics that stop it at once.
Price is the exception: the same Offer never shows two Prices in one market. parsepaper.dev tests $15 against $12 by geography, charging $15 to new developers in one region for four weeks.
Conversion holds within two points and revenue per developer rises 22%, so $15 becomes the list Price everywhere. The losing variant concludes with its Evidence kept, and the winner becomes the definition.
Allocation and holding
The Studio compares Startups by these results and allocates attention and money across its Portfolio. A small profitable Startup and a fast-growing one earn different allocations, each on its own numbers.
Holding is a valid choice. When no available action has a justified return, a Startup enters Hold: it keeps serving Customers while its owner adds no new investment.
Next
Which Role owns each Goal, and with what authority: Roles and responsibility.