Almost every posting for this role is a list: eight channels, eleven tools, one line about being
data-driven. Then the first quarterly review turns into an argument about which number decides, and everybody
discovers that the posting never said.
The fix is not more detail. It is four decisions made before the posting goes live, and six lines written
instead of the tool list.
The four decisions
1. Who signs off on the definition. If marketing and sales both own the word "lead", the first argument the
hire has is the same one the company has been having for a year, only now with a salary attached. One name,
one definition, in the posting.
2. What the hire may switch off. A role with no stop rule gets no verdict. Name the condition in the
posting: no readable return in 48-72 hours, or a cost per order past a stated multiple, and the channel goes
off. Without it, every campaign runs until somebody gets bored, and the hire is judged on the campaigns that
were never allowed to end.
3. Which number they own in the first quarter. Not "pipeline". One number, with the volume behind it, in
writing.
4. Where the artifacts live. Definitions, dashboards and stop rules have to be readable by the next hire.
A posting that never mentions a written record is hiring a contractor with a salary.
The volume that belongs in the posting
Here is the line that quietly breaks most of these roles. A 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. If
the company sends a few hundred a month, no channel can be read inside the first quarter, and the honest
posting says so: it asks for a cheaper readable unit (a reply instead of a sale), a higher base rate (tighter
list, sharper promise), or a smaller promise about what will be known by when.
You can compute that floor before writing the posting
(test planner), write the stop rule before the first send
(kill rule calculator), and generate the posting itself from the level
and surfaces you actually have (job description generator).
If you prefer it local, the same generator ships as the jd command in the CLI:
npx marketing-mindset jd --level lead --surfaces seo,paid
Six lines instead of a tool list
- What the role owns end to end, in one sentence.
- The one channel or loop it owns first, with the measurement around it.
- The event that ends a test - written by the person who will act on it, not by the person hiring.
- The volume actually on hand, next to the floor that volume has to clear.
- The two systems the hire has to write to on day one, and who owns the data model.
- The artifact due in each of the first four weeks, so that week one leaves a document behind.
What to ask in the interview
One artifact produced before you met them: a definition of a lead, a stop rule, a dashboard with a volume floor
on it, or a written record of a channel that was killed and why. Then ask what the number was and what happened
when it was not reached. Candidates who have run a test to a verdict have the answer; candidates who have run
campaigns have a list of channels.
The full version of this, with the editable table and the Q&A,
is here. No invented case studies or salary benchmarks in either place - the arithmetic is all
from published packages you can run yourself.
Top comments (1)
도구 목록보다 누가 정의를 승인하고, 어떤 조건에서 채널을 끌 수 있는지 먼저 적어야 한다는 구분이 좋네요. 특히 기본 전환율 3%, 상대 상승 20%에서 그룹당 13,914명이 필요하다는 계산은 분기 목표를 쓰기 전에 실제 표본 규모부터 확인해야 한다는 점을 선명하게 보여줍니다.