The case The numbers The short version

Working model · assumptions are yours to move

What a managed web presence actually costs to run

Every number below is an input you can change. Two of them are measured from real projects and marked as such. Everything else is an assumption, marked as such, and most of them are probably wrong in ways this page is designed to let you find out.

Rates 28 Jul 2026Base currency USDContractor · revenue share

01 Assumptions

Only two inputs here are evidence. The measured tag means it came off a real project; assumed means someone picked it, and that someone was me.

02 One client, one month

The retainer, less what it actually costs to serve. This is the number the whole business rests on, and it is small and boring by design.

03 One operator

An operator owns their clients end to end — intake, build, revisions, ongoing maintenance. This is their whole book, monthly.

04 Where this beats an agency that still writes the code

Cost to serve one client for one month, as an operator's book grows. An agency's cost is labour and stays roughly flat per client. Ours is a fixed operator plus a few dollars of tokens, so it falls with every client added.

Cost to serve per client per month, versus a hand-coding agency
This model Hand-coding agency Retainer price

05 Against the $300 site

The commoditised end of the market sells a build and disappears. That is not a cheaper version of this business — it is a different business, and this is the gap.

Lifetime revenue per client, three models compared

06 The tailwind, and how little it matters

Token cost falls over time — open-weight models, competition, better harnesses. It is a real effect and it is worth modelling honestly, which means showing that it is not what makes this work.