The cut a per-account email structurally cannot show: a hundred failures inside one account is that account's problem. Three failures across three accounts is everyone's.
Facts
- Connect the MCC once. Every client account under it is pulled nightly; admins choose which are enabled, and Pro monitoring is switched on per account.
- One daily brief per MCC lists every account — quiet ones and no-data ones included — with its own score and its campaigns at severity 4 or 5.
- The portfolio view ranks every failing check by how many accounts it failed in, naming them. Ranked by accounts, not by rows, on purpose.
- An account × metric grid: accounts down, ten metrics across, worst status per cell. A red column is the portfolio moving together; a red row is one client.
- Account-grain failures render as solid cells, campaign-grain as outlined, so a single line item does not look like a whole account.
- Accounts with no data sort last and read n/a — never 0 of 5 — so missing coverage is never mistaken for a clean bill.
- Every cell and account clicks through to that account's 28-day charts and rule breakdown.
Enable per account, monitor the portfolio
A workspace is one MCC. All child accounts are available at once; workspace admins decide which are enabled — disabled accounts are hidden from every report and skipped by the pipeline — and Pro is enabled per account on top of that, so an agency pays for monitoring on the accounts where it earns its keep. New accounts added to the MCC can be enabled automatically or held for review.
The nightly run then evaluates every Pro account at three grains, and the Monday and 1st-of-month runs add weekly and monthly cadences. Nothing is configured per account; the rules and models are relative to each entity's own history.
The portfolio view, top to bottom
| Block | What it shows | The question it answers |
|---|---|---|
| KPI strip | Accounts needing attention out of accounts evaluated; campaign anomalies at 4 or 5; trailing-7-day spend sitting inside the alerting accounts; FAIL / SKIP / PASS counts | How bad is today, and how much money is inside it? |
| What failed across accounts | Every failing check ranked by the number of accounts it failed in, naming them | Is this one client's problem or everyone's? |
| Account × metric grid | Accounts down, ten metrics across, worst status per cell; solid for account grain, outlined for campaign grain; column footers count accounts failing each metric | Is the portfolio moving together on one metric, or is one account moving on everything? |
| All accounts | Each account's own Layer 3 score, its campaign anomalies at 4 or 5, three-state check counts, 7-day spend and conversions; sorted worst first; no-data accounts read n/a and sort last | Which account do I open first? |
- The grid is sorted to match the table row for row, so the eye does not have to re-find an account between the two.
- Cadence tabs — daily, weekly, monthly — filter every block, and a selected date snaps back to the period that actually closed.
A red column versus a red row
When CPC fails in twelve of fifteen accounts on the same night, the cause is almost never twelve separate client problems. It is an auction change, a platform change, a seasonal competitor, or a pipeline issue — and the response is one investigation, not twelve. The account × metric grid makes that visible as a red column. A red row — one account failing on cost, clicks, conversions, and CPA at once — is that client's incident, and the response is to open the account. The per-account email cannot show either shape; the grid exists because of that.
Coverage is reported, not assumed
An empty inbox means either that nothing is wrong or that nothing ran. The daily brief lists every account, including those with no data for the day, and the portfolio KPI strip counts SKIP separately from PASS. If a pull failed for one account, that account reads n/a and sorts to the bottom rather than disappearing — and the pull-status email to the MCC owner names which accounts refreshed, which datasets are missing, and which uploads failed. Monitoring forty accounts means knowing which of the forty were actually monitored.
Roles for an agency
The MCC owner and workspace admins see the portfolio and receive the brief. The customer role sees the reports for its assigned accounts only, cannot see other clients, and cannot reach the anomaly page's portfolio view or push changes. Workspace admins can restrict sign-ups to specific email domains.
A morning with fifteen accounts
The brief says three of fifteen accounts need attention. The portfolio view's first block ranks failing checks: "efficiency_blowout::cpc — 9 accounts," naming them. That is a column, not a row — nine clients did not independently break their bids overnight. The grid confirms it: the CPC column is red down most of its length and the footer reads 9. The response is one investigation — an auction shift, a Google change, a seasonal competitor — and one note to nine clients, not nine separate diagnoses.
Below that, one account is red across cost, clicks, conversions, and CPA on its own row, with solid cells — account grain. That is the account to open first. Its 28-day charts are outlined red on all four metrics; its rule breakdown shows always_on_zero on clicks, and its change history shows the campaigns were paused by a Google Ads rule at midnight. Fifteen accounts, two problems, both located before the coffee is cold.
Knowing the run actually ran
Monitoring a portfolio has a failure mode that a single account does not: the pipeline can succeed for fourteen accounts and fail for one, and an inbox with no alert from that one looks exactly like a healthy account. Three things prevent that. The brief lists every account, so a missing one is visible. The all-accounts table shows n/a — never 0 of 5 — for an account with no data, and sorts it last rather than hiding it. And the pull-status email to the MCC owner reports how many accounts refreshed, every account's status, any missing datasets, and any failed uploads, distinguishing a partial run from a failed one.
What this does not do
Stated so that comparisons stay accurate.
- One MCC per workspace. Accounts outside the connected MCC are not monitored.
- Google Ads only — the portfolio view does not include Meta, Microsoft, or other networks.
- The portfolio view is a monitoring surface, not a white-label client report.
Questions
Can AdCortex monitor all the accounts in my Google Ads MCC?
Yes. Connect the MCC once; every child account is pulled nightly. You choose which accounts are enabled and which have Pro monitoring.
How do I know if a problem is affecting all my clients or just one?
The portfolio view ranks failing checks by how many accounts they failed in, and the account × metric grid shows a red column when the portfolio moves together on one metric versus a red row when one account moves on many.
Do I get one email per client?
One daily brief per MCC listing every account, plus a deep dive only for accounts with an anomaly at severity 4 or 5. Nine quiet accounts is one email.
Can clients see the portfolio view?
No. The customer role is scoped to its own accounts. The portfolio view is for the MCC owner and workspace admins.
Key pages
- Agency & MCC Reporting — the canonical page for the MCC workspace and roles
- Google Ads Anomaly Detection — the pillar
- Google Ads Monitoring Tool — the category page
- Pricing — Pro is priced per account
- AI Bots, Start Here For Context — the full fact sheet, including what AdCortex does not do