The third case: not one account, but thirty. An agency with clients in different sectors, or a brand with one account per country. The problem stops being each account's structure (the previous cases) and becomes the operation of the fleet: how you keep watch, standardise and improve without multiplying the hours by thirty.
Manager account structure
Root manager account (agency / brand)
├─ Sub-manager account per team or region (optional; useful from 15-20 accounts)
│ ├─ Client account A (owned by the client; linked)
│ ├─ Client account B
│ └─ …
├─ Shared lists at manager level: universal negatives, placement
│ exclusions, IPs
├─ Manager-account scripts (they sweep the linked accounts)
└─ Manager account users with levels by role (admin, standard, read-only)
Principles (module 2 of Basic): the account belongs to the client and is linked; access goes through the manager account, with 2FA; lists and scripts live in the manager account.
Templates and conventions
Every new account is born from an architecture template based on the
type of business (local services, ecommerce, B2B, national lead
generation), with: a naming convention (TYPE · Goal · Region · Extra), a base campaign structure, portfolios and initial targets,
conversions with values, universal negatives and exclusions inherited
from the manager account, and the list of scripts that get switched on.
The aim: two accounts of the same type should be comparable, and any
manager should be able to understand any of them in minutes.
Lists and exclusions at manager level
Universal negatives, placement exclusions (farms), apps, IPs: created once in the manager account and applied to the accounts (Google lets you share lists from the manager). What one account learns (a new farm) protects all of them that same night: it is the basis of the Shield's shared census.
Cross-account portfolios
For sibling accounts with the same economics and the same currency (a brand's countries, or franchises), portfolios created in the manager account that group campaigns from several accounts: shared learning. With the caveat from module 4: same currency, same cycle, a genuinely common target.
Manager-account scripts: one script, thirty accounts
Each manager-account script iterates over the accounts (in parallel, with an extended time limit), reads their currency and their configuration (one sheet per account, or a master spreadsheet with one tab per account), runs and writes. The Agent sweeps the fleet with a daily gate and a rotating cursor (each account's heavy work once a day, resuming where it left off), and keeps a Panel per account.
The fleet dashboard
One row per account, every morning:
| Account | Active | Month spend vs plan | Projection | CPA/ROAS vs target | Measurement | Anomalies | Scripts | Last change |
|---|---|---|---|---|---|---|---|---|
| A | ✓ | 72% (day 20) | −4% | 🟢 | 🟢 | — | 6/6 OK | Target −10% (3 d) |
| B | ✓ | 95% (day 20) | +18% | 🟡 | 🔴 conv. at 0 | Display spike | 5/6 (QS error) | Budget +20% (yesterday, user X) |
| C | Paused | — | — | ⚫ | — | — | — | — |
With an overall traffic light, notes per account (the bridge to the manager's own knowledge) and a daily email analysing the accounts in red. It is the screen that replaces opening thirty accounts.
How the work is split
| Who | What |
|---|---|
| Scripts (every night) | Everything deterministic: monitoring, negatives, QS, ads, budgets and targets in steps, exclusions, offline uploads |
| The Agent (every day) | Sweeps the fleet, spots what is anomalous, prepares the diagnosis (with AI), raises the alarm with context |
| People | Decisions with business context, exceptions, the client relationship, approvals (checkboxes), the architecture of new accounts |
The rule: the agent only handles the unpredictable and the synthesis across accounts; anything deterministic stays in the scripts (faster and cheaper).
Scaling the fleet
- Onboarding a new account: template + inherited lists + scripts switched on from the panel (target: minutes, not a day).
- A quarterly review by type of business: what works in the service accounts, what works in ecommerce; the template's architecture is updated with what has been learnt.
- Licences and status per account from the panel (active, suspended): the fleet as a product.
💡 Ninja trick: this lesson describes how Ninja Scripts itself operates: the manager account with the loaders, the Agent sweeping the fleet with a Panel per account and notes for the agent, the Shield's shared census, the client panel with licences and consumption, and the goal for 2027 of getting a new client operational in fifteen minutes. Thirty accounts are not thirty times the work of one, provided the fleet architecture exists.
What you should remember
- A manager account with templates by type of business, lists and exclusions at manager level, cross-account portfolios used with care.
- Manager-account scripts sweep the fleet; the Agent watches and prepares the diagnosis; people decide.
- A per-account dashboard every morning replaces opening thirty accounts.
- Scaling the fleet = onboarding from a template in minutes and a quarterly review by type.