Acentecom · Google team

Junior media buyer: role brief

Draft for Gianfranco and Uzi · from Carlos · v3, Oct 11, 2026 (v1: Sep 30) · Status: waiting on Uzi's OK to open the search

Review notes for Carlos: what changed in v3

Why a second person Updated

The job: find where we can spend more at or above the NC ROAS floor, prepare the change with its evidence, and make sure what goes live works.
156
campaigns live before the Sep 24 pause (136 Search, 13 PMax, 7 Demand Gen)
34
Google Ads accounts, 8 brands
$6.40M
Q4 new-customer revenue target (Sep 15 to Dec 31)
up to 16
test builds a week (one landing and one ad test per brand)

Since Aaron left on Oct 1, one person has covered strategy, builds, feeds, tracking QA and daily operations. The Q4 plan expects growth to come from scaling more often, across more campaigns, brands and channels, while each step stays the same size. That only works if every live campaign is read every day. One person can't do that and also run the weekly tests on 8 brands.

Gian put it plainly on Oct 7: the strategy keeps moving, but the numbers fall behind. The junior owns the numbers. Every day they read each campaign against its floor, propose what to scale or fix, and apply what Carlos approves.

Which target this moves: Q4 NC revenue and new orders, by covering every campaign every day instead of the largest ones. No figure is put on it yet. The first 60 days of logged scaling changes will give one.

The core of the role: the scaling loop New

Every morning every live campaign falls into one of four boxes. Each box has one move. This is how Aaron scaled, and it's the logic our scaling engine already runs.

At or above floor · budget used up
Raise the budget
The budget is the ceiling. Step up to +20%, then wait 7 days.
At or above floor · budget not used
Wait, or loosen the target
The target is the ceiling. Lower tROAS only when the return is well above floor.
Below floor · budget not used
Raise the tROAS
A budget cut would do nothing here. A target step moves it.
Below floor · budget used up
Cut the budget, then raise tROAS
A cut alone doesn't fix ROAS. The campaign just fills the smaller budget.

What the junior does each day

  1. Reads NC ROAS (Triple Whale, Triple Attribution 14d) against the floor for each campaign, plus budget use and Google's own status reasons: budget constrained, bid strategy constrained, learning, limited by policy.
  2. Writes one proposal per campaign that needs a move. Each one has the campaign ID, current and proposed value, the evidence, and the Q4 target it moves (NC revenue, orders or subscriptions).
  3. Applies what Carlos approves through the guarded scripts, with a dry run first and every change logged.
  4. Reads each change again 7 days later: did spend grow, and did NC ROAS stay at or above the floor?

Beyond bids and budgets: expansion, every week

The caps stay: +20% / -33% budget per step, 7 days between steps, and one change per campaign per cooldown. Carlos is the one who decides to go past a cap. The NC ROAS floor is never traded for volume.

Split by function, not by channel New

A common setup is one person on Search and another on PMax and Shopping. It works when two seniors each own a channel of similar size. Here it doesn't, for four reasons:

Split by channelSplit by function (proposed)
WorkloadSearch was 136 of 156 live campaigns (87%). One person would carry almost everything.Work splits by type of task, so the load is even.
TargetsFloors and Q4 targets are set per brand. A brand's Search and PMax reach the same buyers, and someone has to split the budget between them.One person sees the whole brand and decides its mix.
Shared setupFeeds, Voluum links and Netz pages serve every channel. A page change hits both people's campaigns.One QA pass covers every link.
Junior levelThe junior would own budget and bid decisions on a whole channel from day one.Decisions stay with the lead until the junior has a track record.

How it changes over time: by function for the first 60 days. From day 60 the junior takes a lane by brand, not by channel: two or three of the smaller brands, end to end across all their channels, with Carlos approving. Brand is the right unit because the targets are set per brand.

Who owns what Updated

AreaCarlos (lead)Junior
Strategy and targetsQ4 plan, floors, budget split between brands and channelsWeekly pacing read: actual against month-to-date target by brand, with a recovery list when behind
Scaling and optimizationApproves every change; decides when to go past a capDaily scaling read and proposals; applies the approved ones; reads each change again after 7 days
ExpansionPicks which openings to pursueWeekly list of new keywords, products, placements and markets, each with its evidence
TestingPicks the tests, writes the hypothesis, calls the resultBuilds the ads and campaigns for each test, keeps the test log current, flags when a test reaches its volume floor
Advertorials and landing pagesAngle, outline and copy. Publishes in KittyBuilder, Netz and Voluum.Prepares the page and link spec; after Carlos publishes, runs the QA visit (tracking, CTA, checkout, California version)
Netz migrationDecides which campaigns go back live and any URL changeKeeps the migration sheet current, checks that each link loads, lists the doubts for the end-of-day review
Feeds and Merchant CenterFeed strategy, policy decisionsDaily disapproval check, title and image variants, feed edits
Policy and approvalsAppeals, landing-page fixesDaily ad approval check, resubmits disapproved ads after a fix
Demand Gen and videoCreative direction, audience planUploads assets, builds ad groups, rotates creatives
Reporting and tasksReads results, presents in the weekly, sets prioritiesChecks the daily report ran, answers data requests, keeps ClickUp and the link sheets current

How the work gets done: AI-first New

The junior works the way Carlos does, and Carlos trains them in it.

Subscription: Claude Pro at US$20/month is the minimum. Who pays for it is in the decisions list below.

A day in the role Updated

US Eastern time, because the stores and the budgets run on it (some budgets run out by 2 to 4 pm).

A week in the role Updated

DayMain work
MondayPacing read before noon, with a recovery list for any brand that's behind. Expansion review: search terms, Keyword Planner, products and brands without coverage.
TuesdayGoogle Weekly Tactical: Carlos sets the landing test and ad test per brand. The junior turns them into a build list in ClickUp the same day.
WednesdayBuilds. Page and link specs go to Carlos by end of day.
ThursdayCarlos publishes the pages and links, the junior runs QA, and the tests go live. What's asked on Tuesday ships Thursday or Friday, as agreed with Gian.
FridayScaling review: last week's changes against their results. Test log update. Weekly feed and Merchant Center check.

Ground rules Updated

Never

  • Publish or change anything in KittyBuilder, Netz or Voluum. For now that's Carlos's job. The junior prepares the spec and runs the QA.
  • Apply a change that Carlos hasn't approved, or trade the NC ROAS floor for volume.
  • Step a budget past the caps (+20% up, -33% down, 7 days between steps).
  • Select what to change by campaign or ad group name. Use the ID, because names repeat across accounts and brands.
  • Point an ad or a feed item at a bare store URL. Every link goes through Voluum.
  • Use someone else's credentials.

Always

  • Log every change: what, where, before and after, who approved it.
  • Quote NC ROAS from Triple Whale with its model (Triple Attribution 14d or Last Click 7d). Google Ads conversions aren't the ROAS source.
  • Say which Q4 target a proposal moves, and by roughly how much.
  • Check a health landing page for drug names and claims before sending traffic to it. Google reads the page, not only the ad.

Access to set up before day one Updated

ToolUsed forLevelWho grants it
Claude CodeThe daily workPro (US$20/month) minimumOpen: who pays
Team repo (skills and scripts)Shared skills, scripts, test logAcentecom part only, not Carlos's full workspaceCarlos
Google Ads (MCC, 34 accounts)Reads, builds, approved changesStandard, not admin; own API credentialsCarlos
Merchant Center (per brand)Disapprovals, feed checksStandardCarlos
Feed Google SheetsTitles, images, variantsEditorCarlos
Triple WhaleNC ROAS, new-customer revenueReadGian
Voluum, Netz, KittyBuilderQA and specs onlyView only if the tool allows it; no publishingSaro or Gian, Pawel, Gian
Daily Google reportDaily numbers, pacing, testsLinkCarlos
ClickUp and Slack (#google channels)Tasks, requests, daily notesMemberGian

How we know it's working Updated

Growth

Reliability

First 90 days Updated

Days 1 to 30: set up, learn, check

  • Week 1: Claude Code, the team repo, API credentials and access. Account map (8 brands, 34 accounts, naming conventions) and the floors. Shadows the scaling read.
  • Week 2: runs the daily checks and the scaling read, with Carlos reviewing both. Joins the Tuesday tactical.
  • Week 3: runs both alone before the sync. Builds the first test ads from a brief.

Days 30 to 60: propose and build

  • Scaling proposals need fewer changes each week. Owns the pacing read and the expansion review.
  • Builds test campaigns and ads from a brief without rework.

Days 60 to 90: own a lane

  • Takes two or three of the smaller brands end to end across all channels, with Carlos approving.
  • Proposes tests and expansion ideas of their own, each with a hypothesis and a metric.
  • Routine changes within the caps no longer need pre-review. Publishing in KittyBuilder, Netz and Voluum stays with Carlos.

Profile and hiring Updated

Hiring steps

  1. Screening call with Carlos (30 minutes).
  2. Paid practical task, about 2 hours, any AI tool allowed: from a sample campaign export, sort the campaigns into the four scaling boxes and write the proposals; build one Search ad group with two responsive search ads from a one-page brief; find the errors planted in a sample link sheet. They send their AI chat with the result. What we're judging is how they checked the AI's output.
  3. Final call with Gian.

Media buyer or native copywriter

On Oct 7 we talked about a junior media buyer, a copywriter for native, or both.

Recommendation: hire the media buyer first. The gap Gian pointed to is operational: numbers, scaling and builds. Ad and advertorial copy is covered today by Carlos, the copy framework from Olga, and Agus's reviews.

A native copywriter makes sense once the advertorial tests (decision on Oct 21) show which angles win and we need more than one new advertorial per brand each week. They would draft the advertorial and native copy. Carlos would keep the angle and the final approval.

Decisions needed from Gian and Uzi Updated

  1. OK to open the search (Uzi).
  2. Budget and contract type: employee or contractor.
  3. Who pays for the Claude subscription (Pro, US$20/month minimum): the company or the hire.
  4. Location and time zone, given the 4-hour overlap with US Eastern.
  5. Target start date. If they start in early November, they'll only be running the daily read alone by late November, during BFCM.
  6. Media buyer only, or media buyer now and a native copywriter later.

How we work together