Fractional marketing engineer vs fractional CMO: pick by the bottleneck

Most companies hire the wrong fractional seat, pay for three months of it, and conclude that
"fractional" does not work. The mistake is almost never the person. It is that the seat and the bottleneck
were never matched.

Two different jobs wearing one word

A fractional CMO decides and delegates. Strategy, channel choice, hiring, vendor management, reporting.
A fractional marketing engineer builds and hands over. The page that carries the offer, the source pages
that AI answers quote, the sending setup that passes a deliverability check, the tracking event that makes a
click measurable. Same calendar shape, completely different output.

Diagnostic, one question:

• If you cannot yet say which customer, which offer, which channel — the missing thing is a decision. Buy judgement. Hiring an executor here just produces fast, well-measured work in the wrong direction.
• If the decision exists and is not shipping — the missing thing is throughput with senior judgement attached. That is the engineering seat, and it closes in weeks instead of a hiring cycle.

What "fractional" should mean on the invoice

The word does not describe hours. It describes what stays when the engagement ends. Two contracts can both say
"fractional, $1,900 a month" and leave you in opposite positions:

1. Assets on the vendor's side. The ad accounts, the landing pages, the sending domains, the analytics property are all theirs. The retainer stops, the machine stops, the history stops. You were renting attention, not building an asset.
2. Assets on your side. Pages on your domain, tracking inside your own analytics property, sending on your own domains. The contract can end on a Tuesday with a 14-day notice and nothing breaks.

Rule of thumb: if a line in the proposal cannot be reduced to a URL or a login you control, it is not an asset.

The parts that actually decide whether it worked

• The offer before the channel. No amount of execution rescues a proposal nobody wants. Write the sentence that says who it is for and what changes for them, then build around it.
• A test floor declared in advance. For cold email reply-rate tests, the honest floor is roughly 1,500–2,000 sends per variant. Below that, any winner you see is a coin flip with a nicer chart. Agree the floor before the test starts, not after it loses.
• Source pages written to be quoted. Definition and query-intent pages with schema and real internal links get cited by answer engines far more often than product pages do. Boring to write, cheap to keep, and they keep working while you sleep.
• One event that matters. Usually clicks to a conversation. Everything else on the dashboard is a supporting actor.

What I publish instead of case-study theatre

The method above is the one I sell, so the honest way to show it is to leave the artifacts up:
a fractional engagement page with the scope and
the fixed packages in public, a proof page listing every shipped artifact
with an openable link, and a free tool
that computes the sample size a test actually needs before you spend the sends.

If you are choosing a seat right now, run the diagnostic in the first section, then ask the two questions that
must be asked before any retainer: what does the scope page say is out of scope, and where do the assets live
on the day after the contract ends.