PPC reporting for agencies: the data layer behind the client rhythm

TL;DR

PPC reporting at an agency is two distinct layers, and conflating them is why most of it is either late or shallow. The internal layer is the practitioner's monitoring rhythm — daily pacing and anomaly checks, weekly optimization passes with a written actions-taken log, monthly deep reads — that exists whether or not a client ever sees it. The client layer is the communication rhythm (weekly signal, monthly narrative, QBR) covered in the client reporting rhythm post — and it should be a selection from the internal layer, never a separate scramble. The bridge between them is the actions-taken log: numbers tell clients what happened; the log proves someone made it happen. Automate the numbers end of the stack (scripts alert, pipelines warehouse, dashboards read from materialized tables) so human hours go only where they're irreplaceable — the narrative and the decisions.

---

The recurring agency failure: the monthly client report consumes the last three days of every month, assembled by hand from four platform UIs, and still doesn't answer the client's actual question — "what did you do, and did it work?" The fix isn't a better template. It's recognizing that client-facing reports are the visible tip of a reporting system, and building the system underneath first.

The internal layer: what the PPC team watches

Daily — by machine. Pacing versus monthly commitment and metric anomalies versus weekday baseline, across every account, alerting only on breach. This is precisely what the defensive script set automates; a human "daily check-in" across thirty accounts is a ritual, not a control.

Weekly — by practitioner. The optimization pass per the pillar's beat cadence — batched by platform, all clients in one sitting — with one non-negotiable output: every action logged against the client as it's taken. Bid changes, negative additions, creative swaps, budget moves: one line each, in the task system, dated.

Monthly — by senior eyes. The deep read per account: search-term and N-gram review, conversion-quality drift (are we buying the same leads sales liked last quarter?), auction-insights shifts, creative fatigue curves. This is where next month's plan comes from — which makes it the raw material for the client narrative, not a separate exercise.

The actions-taken log is the product

Clients churn saying "we never knew what they were doing" — almost never "the CPA was 8% high." The numbers layer of PPC reporting answers what happened; only the actions log answers what you did about it. Kept as a discipline (one line per action, at the moment of action), the weekly client signal becomes a five-minute filter — "show this client's completed actions this week" — and the monthly report's "what we did" section writes itself from the record instead of from memory. Kept as a memory exercise, it's fiction by the 30th. This is the operational reason PPC work belongs in a task system that knows about clients — the same five-field discipline the rest of the agency's work runs on.

What flows into each client beat

The client layer's cadence and rules — same weekday, bad news first, narrative by humans — belong to the client reporting rhythm; the PPC-specific question is what flows in:

• Weekly signal: pacing status (on/over/under and the correction), the top 2–3 actions from the log, one number worth noticing. Nothing that requires a dashboard login to understand.
• Monthly report: results against the targets set last month, spend reconciliation, what we did (from the log), what we learned, next month's plan. Platform-claimed conversions stated with the referee's number when they diverge materially — clients eventually see both, and the agency that volunteers the discrepancy owns the explanation instead of defending it.
• QBR: the cross-platform story — blended CAC trend, the exaggeration report from the spend warehouse, budget reallocation cases. Strategy, not screenshots.

The automation stack, bottom to top

Machines pull, humans narrate — operationalized: scripts alert on breaches (free, in-platform), pipelines warehouse daily spend and performance into the canonical BigQuery table where it sits next to GA4's independent count, and dashboards read from small materialized report tables — the Looker Studio pattern — so they're fast, cheap, and current without anyone exporting a CSV. Where a client's stack is light, a dedicated PPC reporting tool can stand in for the middle of that stack — the trade-offs are in our PPC management software guide — but the two ends never automate: deciding what the numbers mean, and writing it down for the client.

The build order if you have none of this: actions-taken log this week (it's a habit, not a tool), pacing script next week, the spend pipeline this quarter. Each one removes a category of month-end scramble — and the narrative hours they free up are the part of Phloz-style agency reporting clients actually renew for. The reporting hours themselves belong on the capacity board like any other delivery work; that argument is in the capacity planning post.