Glossary

Average revenue per user — total revenue divided by the number of customers over some period. Useful for spotting whether growth is coming from more customers or from each one paying more.

Customer acquisition cost — what it takes to land one paying customer. The standard formula divides marketing spend by new customers, and where there is no marketing spend it reports a CAC of zero — flattering, and not the truth. State it in the currency you actually pay it in — hours spent acquiring, divided by customers acquired — and keep any real dollars spent in a separate column, rather than converting the hours to dollars at a rate you invented.

The rate at which paying customers stop paying. A subscription with no churn number attached is a subscription nobody has measured yet.

Distribution effort that is still producing something a month after you stop doing it — a ranked page, a referral loop. The opposite is a channel that evaporates the moment you stop posting, which looks identical to a compounding one for the first few weeks.

Monthly revenue, minus monthly direct costs, divided by the hours the project takes to run each month — maintenance plus whatever acquisition work still feeds it. This course’s own construct, not a term with a textbook definition: it is the one figure in the ledger with your own hours in the denominator, which is what makes it comparable to a contracting rate, a raise, or another project on the shelf, in a way a dollars-only metric is not.

The free slice of a product, judged as a marketing line item rather than a cost centre: it earns its keep only if it converts some of its own users to paying, or brings in paying customers who never used it themselves. A free tier doing neither is not a funnel with a slow conversion rate — it is an audience you are supporting for free, on the same finite hours as every other maintenance cost.

A built pipeline — landing page, email sequence, ad spend — that only earns its own cost once enough volume arrives for a conversion rate to mean something. Below that volume, at the scale of finding your first ten customers, every hour spent building one is an hour not spent finding an actual person who might actually buy.

A written rule with three parts — a measurable named precisely enough to know where to read it, a threshold that separates continue from stop, and a date the reading gets taken. This course’s own construct: kill criteria are not a standardized concept, and the three parts exist because each is a different way a criterion gets argued out of existence later by the very hours it was meant to bound.

Lifetime value — the profit one customer is expected to return, over a horizon you choose and write down. The textbook formula (revenue times margin, divided by churn) can run to infinity as churn approaches zero, so do not leave it open-ended: cut it off at a stated stretch of time you have some reason to believe. At solo scale a handful of customers do not produce a statistical churn rate, so treat the number as an order of magnitude rather than a figure to compare against another figure of the same size.

The hours per month you are willing to spend holding already-shipped things steady, at zero growth, before any hour goes into building or selling anything. A fact about your month, not about any one project, so it is chosen by subtraction from a month that actually exists — never defined as “whatever’s left,” which is an outcome rather than a commitment and can never be exceeded. Set the summed maintenance load of everything still reachable against it, and what comes out is a cap on how many things you can keep alive at once — a different constraint from the WIP limit, which caps candidates rather than hours.

Hours per month plus dollars per month to keep one product alive doing zero growth work — no new features, no marketing, just what it costs to not let it break. A property of a single project, measured or estimated per project. Summed across everything still reachable, it is what a maintenance budget gets set against. The number most builders have never actually measured for their own projects.

The legal entity selling the product to the customer — as distinct from a payment processor, which only moves the money. Whoever is merchant of record is what makes the tax obligation theirs rather than yours: their name is on the card statement, they register for and remit VAT, GST and sales tax, and they hold the customer relationship for refunds and chargebacks. Your own income, self-employment or corporation tax on what they pay out stays yours regardless — no merchant-of-record service touches that.

Monthly recurring revenue — the subscription revenue you can count on next month if nothing changes. Excludes one-time charges, which is exactly why it is worth tracking separately from total revenue.

Judging a project by how similar projects actually went, rather than reasoning through this project’s own specific circumstances (the inside view). Bent Flyvbjerg’s term, built on a reference class of comparable projects; at solo scale the comparable projects are your own past ones, and the point of asking is the same — whether you have a structural reason to expect this attempt to go differently, not just a feeling that it will.

Not killed and not pursued: no work, no spend beyond what keeps the lights on, and a dated revisit written down. Distinct from abandoned, which is a kill nobody wrote down.

A set of past attempts, judged as a group, that a new one is compared against instead of being reasoned about on its own — the mechanism behind the outside view. Bent Flyvbjerg’s own version is built from other people’s completed projects, large enough to be statistically meaningful; at solo scale, the only available class is your own past attempts, which is this course’s extension by analogy rather than a finding, and a handful of rows is a prior to argue away from, not a statistically meaningful distribution.

Revenue per thousand — ad revenue divided by pageviews or sessions and multiplied by 1,000. The unit ad networks quote, and the two variants (per pageview, per session) are not interchangeable.

What a searcher actually wants, not the string they typed. Google’s own taxonomy is Know, Know Simple, Do, Website and Visit-in-person — not the informational/navigational/transactional split you will find on SEO blogs, which is a third-party invention presented as if it came from the search engine itself.

What actually has to happen on the day a kill criterion fires: notice with a date on it, no further charges after that date, a refund for time paid and not delivered, an itemised data export with a deletion date, and a decision about the domain — renew and redirect, transfer, or let it lapse. The ledger’s Sunset plan field is where this gets written down, in advance and beside the criterion that triggers it.

What one customer costs against what one customer is worth, at the scale you actually run at. The venture-scale ratio between the two is a test of whether it is safe to spend faster on acquisition — it assumes a capital budget to deploy and treats cash as the scarce resource. At solo scale there is no acquisition budget to deploy faster, the cost side is paid in hours rather than dollars, and the ratio has nothing to say about what else those hours could be doing — importing it wholesale answers a question you were never asking.

A hard cap on how many candidates you are actively pushing at once — at most two in pursue. It is a claim about where new hours go, not about the shelf of already-shipped things you are not actively pushing; what those cost in total is a different constraint, capped by a maintenance budget rather than by this limit.