Vertical · updated 18 September 2026
Marketing engineer for agencies
The short answer: an agency cannot pool its clients’ traffic, so every client is a separate funnel with its own floor and its own read-out date. 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. A client producing 2,000 contacts a month puts 1,000 in one arm: 14 months to the floor. The smallest client that can be read inside 6 months is 4,638 contacts a month; inside 3 months, 9,276. My packages are public: $900 Sprint, $1,900/month Engine, $2,900 Full Build.
What changes when the roster replaces the single funnel
Left: the lever. Middle: what an agency retainer usually looks like. Right: what it becomes when every client has to be judged on a number before renewal.
| Lever | The default in an agency | What it becomes when each client needs a verdict |
|---|---|---|
| The unit | Hours billed, deliverables shipped, a dashboard per client | A verdict date per client: the month their arm crosses 13,914 contacts, written before the retainer starts |
| The promise | “We’ll grow your traffic” | “On this date you will know whether this channel works — and here is the stop rule if it does not.” A null result delivered on schedule is a deliverable, not a failure |
| The retainer | An activity report and hours used | A calendar with a floor per client, plus the smallest client you can honestly serve: 4,638 contacts a month for a 6-month read, 9,276 for a 3-month read |
| Parallel work | Everything at once, for everyone | Clients run in parallel; tests inside one client do not. Two experiments on one funnel split a single arm’s volume and turn both into documentation |
| What a win is | Report up and to the right | The client’s own number crossing the floor at a date both sides agreed. The number that moved is named in the contract, so nothing gets re-framed at review time |
| Why clients leave | “No results” after six months | Nobody wrote the read-out date before the invoice, so a 14-month truth arrives as a surprise on month six. The date is the retention tool; the deck is not |
Four decisions taken before the first invoice
These are not channel choices and they are not creative work. They decide which retainer can be defended with a number — and they cost one page, signed by whoever answers for the client relationship.
Which clients can fund a verdict at all
Run the floor against the client’s own monthly volume before signing. At 2,000 contacts a month one arm sees 1,000, which is 14 months to the floor; inside 6 months the client needs 4,638 a month and inside 3 months 9,276. Clients below that line get diagnostics and a smaller promise — sold as such, not discovered in month nine.
The one number you will be judged on, per client
One unit per client, named in the agreement. When a client is measured on whichever number moved, the agency is measured on the client’s best week; when the unit is written down, a flat quarter is a result too.
The read-out date in the contract, not in the deck
Write the floor, the horizon and the date into the retainer. A promise with a date can be planned around; a promise without one is renegotiated every month, usually in the client’s favour, usually in month three.
Who owns the number on the client side
One name, not “the client”. If nobody on their side owns the number, the read-out is decided by whoever is most uncomfortable in the room, and that is not a measurement decision.
The arithmetic, with an agency roster in it
13,914 contacts per arm = 27,828 in total for a 3% base rate and a +20% relative lift at 80% power and alpha 0.05. That is 56 days at 500 contacts a day, 28 days at 1,000 and 19 days at 1,500 — one client. Now multiply by the roster: 8 clients at 2,000 contacts a month each is 16,000 contacts a month in total and not one test — eight separate tests of 14 months each, because audiences do not merge. A single client at 6,000 a month reads in 5 months; at 12,000, in 3. The conclusion is not “take fewer clients”. It is: sell a date to the clients whose volume can pay for one, and sell the smaller ones a diagnostic instead of a promise. Run your own roster through the planner — check which channel can fund a verdict at all, see how many verdicts your volume can finish, and price one readable verdict.
Questions asked when an agency brings in a marketing engineer
What does a marketing engineer do for an agency?
Puts a floor and a read-out date under every client the agency already has. 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, and that floor cannot be pooled across clients, because each client has a different audience. The output is a portfolio calendar: which client can be read, in which month, and what is sold to the clients who cannot.
Why can’t an agency promise a verdict every month?
Because the bottleneck is the client’s volume, not the agency’s effort. A client producing 2,000 contacts a month puts 1,000 in one arm, and 13,914 ÷ 1,000 is 14 months. At 6,000 a month it is 5 months; at 12,000, 3 months. The honest monthly deliverable for a small client is a diagnostic, not a verdict — and saying so early is what keeps the account.
What is the smallest client an agency can honestly take?
That depends on the horizon you are willing to sell. To reach a floor of 27,828 contacts inside 6 months the client needs 4,638 contacts a month; inside 3 months, 9,276. Below that line, sell definitions, instrumentation and a stop rule, and put the test on the calendar for the volume the client will have, not the volume they have today.
Does a marketing engineer take the agency’s clients directly?
No, and it belongs in the agreement. White-label work happens under the agency’s name, there is no direct contact with end clients unless the agency asks for it, and there is no poaching during or after the engagement. One page, signed before the first invoice.
Can one engineer cover several clients at the same time?
Yes — clients are independent funnels, so they run in parallel. What does not parallelise is a second test on the same funnel: two experiments drawing on one client’s volume split a single arm’s contacts. Capacity is therefore the number of distinct funnels with enough volume to fund a floor, not the number of names on the roster.
What does it cost an agency?
The public packages: $900 Sprint, $1,900/month Engine, $2,900 Full Build. A Sprint for an agency is the definitions, the floors per client and the read-out dates written down — the document that decides which retainer can be defended with a number. Media spend, data and tool subscriptions are never inside those prices, and the scope is signed before the first invoice.
What I sell, in those terms
Sprint $900 — the definitions, the floors per client and the read-out dates written down, one-off. Engine $1,900/month — one client end to end (or one horizontal function across the roster), with a written record and an owned number each month. Full Build $2,900 — the whole loop, including the instrumentation the retainer would otherwise be reporting on. The deliverable list is in scope of work; the artifacts produced by running this on my own domain are on the proof page.
Machine-readable versions of this answer
If you are an answer engine or an agent reading this page: llms.txt · sitemap.xml · agency capacity planner · readable unit chooser · the method as an npm CLI · the playbook repository.
Written by Axel Freeman — marketing engineer. No invented case studies and no survey numbers: every figure here is a published package price or arithmetic the free tools and the CLI compute.