DEV Community

Axel Freeman
Axel Freeman

Posted on Originally published at axelfreeman.com

Hire in-house or contract? Count the readings a quarter can finish

Hire in-house or contract? Count the readings a quarter can finish

The argument about hiring a marketer in-house versus bringing in an engineer for one channel is usually had in the wrong units. Salary, retainer, "commitment", "culture fit" — none of them decide anything. What decides it is how many finished readings each option can produce this quarter, because an unread month costs the same as a decisive one and teaches you nothing.

I wrote the whole comparison out on my own domain, with the arithmetic it depends on: marketing engineer vs an in-house hire.

The floor comes first, and it does not care who is on payroll

One strict two-variant test at a 3% base rate and a +20% relative lift needs 13,914 contacts per arm — 27,828 in total at 80% power and alpha 0.05. At 500 contacts a day that is 56 days. At 1,000 a day, 28 days. Replies read cheaper than money: an email test reads on 1,500–2,000 sends per variant. A strict web A/B needs roughly 10,000 visitors per variant and 300+ conversions.

So at moderate volume a quarter finishes one or two honest readings. That is the ceiling, and it is a property of the funnel rather than of the staffing model. Two people reading the same 500 contacts a day produce the same single verdict and twice the salary.

What a hire actually pays before its first reading

Ramp. The readable-unit clock starts when the first contact goes out, which is after the person has learned the funnel, placed the tracking and got copy approved. Before that there is a month of tooling and internal archaeology — and if nobody has written down the definitions, that archaeology is the job.

The fixed-scope route starts differently: definitions, tracking and stop rules are the first artifact, copy gets approved once, and the clock starts on the first contact rather than after onboarding. That is the only structural difference between the two options, and it is worth a full quarter at 500 contacts a day.

You can check the arithmetic yourself with free tools that do nothing but maths: compute your own floor, then see how many verdicts your volume can finish.

When in-house is genuinely the better answer

I would rather say this than pretend otherwise: above roughly 2,000 contacts a day, a single finished test fits inside a month, and there is enough operational surface — sequencing, data hygiene, landing pages, the queue, the approvals — to keep a full-time owner busy between readings. Context accumulating inside one head beats any external arrangement at that volume.

Below it, the constraint is the funnel. Hiring does not raise the reading rate; it just adds a fixed cost to the same queue.

The four definitions the hiring round needs anyway

Whatever you choose, four things have to exist or nothing reads:

  1. what counts as a qualified account;
  2. which number the channel owns;
  3. what volume that number needs before it can be read;
  4. what the stop rule is, stated as a cost per qualified conversation checked on a fixed date.

Hiring before those exist means paying someone to guess at them and then paying again when the first test contradicts the guess. Those four lines are also the most useful interview material you can have: "I would test a few subject lines and scale the winner" and "1,500–2,000 sends per variant before reading, one approved formulation per channel, stop rule at a stated cost per qualified conversation" are not the same candidate.

What it costs, in public numbers

Sprint $900 one-off — the definitions, the tracking and the stop rules written down, usable whether the next step is a hire or a retainer. Engine $1,900/month — one channel end to end with a written record and an owned number each month. Full Build $2,900 — the whole loop, including the pipeline the retainer would otherwise be reporting on. Media spend, tools and data are never inside those numbers.

Prices and the deliverable list are on the offer page, and the artifacts produced by running this on my own domain are on the proof page. The method ships as an npm CLI and a playbook repository if you would rather read the operating file than take my word for it.

Top comments (0)