Intern
What you actually do
Run the QA on the last thing shipped, staff or agency. Failures are tickets. You are the same test either way.
Library / Compare / Agency vs in-house production: the spec is the SOW in both cases
This is not fractional-vs-agency. This is who ships pages, ads, and emails once a spec exists — staff or a shop. In-house is right for ongoing cadence and context. Agency is right for burst capacity and skills you will not hire. The leak is the same: production without a handoff, extra CTAs, campaigns that impersonate journeys, nobody running Monday. Choose by cadence vs burst, and wire tickets to the spec either way.
This is not fractional-vs-agency. This is who ships pages, ads, and emails once a spec exists — staff or a shop. In-house is right for ongoing cadence and context. Agency is right for burst capacity and skills you will not hire. The leak is the same: production without a handoff, extra CTAs, campaigns that impersonate journeys, nobody running Monday. Choose by cadence vs burst, and wire tickets to the spec either way.
Agency production is right for a launch, a redesign burst, ads ops you cannot staff, with assets landing in your accounts. It is wrong as the owner of the CRM spec.
In-house production is right for weekly emails, small page changes, location support, with the same tests. It is wrong when 'they are busy' means they skip write-in because they can nag each other in Slack.
Shared leak: shipping without tests, CTA soup, two senders, scoreboard unowned.
How to choose: if the work is a pulse, burst is fine. If the work is a clock (Monday, recare, flows), in-house (or a fractional operator) owns the clock. Do not give an agency the clock without a named internal owner.
Wiring implication: same QA script for both. Agency access is least privilege. Files in your sub-account. Kickoff is the spec page, not a mood board.
Intern
Run the QA on the last thing shipped, staff or agency. Failures are tickets. You are the same test either way.
Operator
Bind production to the spec. Clocks in-house (or owned). Pulses can burst. Monday is internal.
CEO
Where the humans sit is secondary. Whether tickets can ship a leak is primary. Spec as SOW for both. Until that is true, headcount and retainers are interchangeable fog.
Spec, ticket, write-in test, cadence vs burst, brand context, QA owner, scoreboard, extra-CTA ban.
Ticket without a spec reference → rejected, in-house or agency
New URL → production path UTM + thank-you test before spend
Burst agency → dated end; assets in your GHL/Klaviyo not only their drive
In-house → cadence calendar; they do not invent a second sale map
What must happen every week vs what is a project.
Why: Agencies on clocks without an internal owner drift. In-house on rare bursts may be slower — that can be OK.
Done when: Two lists.
Form, UTM, book, email exclusion. Used by staff and vendors.
Why: Two quality standards is how vendors look worse or staff skip.
Done when: A checklist in the sub-account description or a pinned note.
Job, writer, stage. No 'make a funnel' without the conversion named.
Why: Vague tickets are how extra CTAs appear.
Done when: A rejected vague ticket.
GHL/Klaviyo/Google. Agency drive is a copy, not the record.
Why: Vendor-only assets are hostage plus a second database.
Done when: A test: you can edit the live form without the agency.
Even if an agency reports. The number is yours.
Why: Outsourced scoreboards are how you stop believing them.
Done when: A person who will defend write-in and show rate.
Burst, scarce skills, dated pulse, assets in your logins, spec as SOW. Wiring: least privilege, QA script, end date.
Clocks, context, location support, weekly ship. Wiring: same QA, they do not own a shadow sale map, Monday person is them or they report to one.
Production without tests. Extra CTAs. Vendor or staff drive as record. Unowned Monday. Building vs vendor is not the leak. Unbound tickets are.
Clock vs pulse. Same QA. Spec-cited tickets. Internal Monday owner. Assets in your accounts. Do not use 'we have people' as a reason to skip write-in.
Until they skip tests and you buy ads into a dead thank-you. Fully loaded plus leak is the cost. QA is not optional because they sit next to you.
Their process may be campaigns. Your process is the spec. If they will not take the spec as SOW, they are the wrong shop for this layer.
They can run the meeting. You still need a company owner or you cannot fire them. Ownership that cannot be fired is a hostage.
This week
This week: clock vs pulse lists, one QA script, ticket template with spec fields, confirm assets live in your logins. No new vendor without that.
Both can be a filing cabinet. Neither is a system until write-in and stages are jobs.
Read ↗Pretty vs native does not matter if two links both take Discovery.
Read ↗DTC lifecycle and coaching spines are different jobs. Two senders is a tax.
Read ↗Rebuilding from memory every client is how you never productize. Blasting a zip is how you never train.
Read ↗Working notes
Pick the one sequence that deserves the next 90 days. Request the file. It arrives by email — not a public dump, not a drip of slogans.
Email and WhatsApp stay open. A call is for installing the system — not for a tour of tools.