Paid media · build guide
PPC Intelligence
Classifies every search term by intent, drafts negatives, and shifts budget to what works.
What it does
PPC Intelligence watches what your ad spend actually buys. It classifies every search term that triggered your ads by intent, flags terms burning budget with zero conversions, builds negative keyword lists across match types, and drafts a budget reallocation plan that moves spend from underperforming campaigns to high-intent ones. A weekly digest tells you what changed, what it recommends next, and what needs your sign-off.
Compounding here looks like hygiene: the account gets cleaner every cycle. Negative lists keep bad traffic out permanently, the classification rules sharpen each time you correct an edge case, and the audit log becomes a memory of every decision — so each weekly pass starts from a better baseline than the last.
You run the loop yourself on scheduled exports: pull the reports, classify terms in a spreadsheet, and apply every change by hand through Google Ads Editor. It suits a marketer managing one or a few accounts who wants the discipline of the system working before wiring up any APIs.
The same loop runs as a scheduled agent workflow: read-scoped connections pull the data, the agent classifies terms and drafts changes inside written guardrails, and only small, reversible actions ship without you. It suits teams with clean conversion tracking who are done copy-pasting CSVs every Monday.
Before you start
Prerequisites.
Path A
Build it with manual data extracts.
You run the loop yourself on scheduled exports: pull the reports, classify terms in a spreadsheet, and apply every change by hand through Google Ads Editor. It suits a marketer managing one or a few accounts who wants the discipline of the system working before wiring up any APIs.
Export your reports on a fixed day
Each week, download the Google Ads campaign export (CSV), the search terms report, and GA4 landing page performance. Add Search Console data if you have it. Drop everything into a dated folder so each cycle is comparable to the last.
Classify search terms in a spreadsheet
Pivot the search terms report by spend and conversions. Tag each meaningful term as converting, informational, competitor, or irrelevant. You can paste batches into Claude to pre-classify, but you verify every tag before it drives a change.
Build the negative keyword list
Turn irrelevant and informational terms into exact and phrase negatives on a dedicated tab. Group them into shared campaign-level lists so one decision protects the whole account, not one ad group.
Draft the budget reallocation plan
Rank campaigns by cost per conversion and conversion volume. Propose shifting budget from the worst performers toward high-intent campaigns, and get sign-off from the budget owner before touching anything in the account.
Apply changes with Google Ads Editor
Upload the negative lists and budget edits in one batch. Record every change — date, entity, before and after value — in a change-log tab so any edit can be reversed later.
Set guardrail rules in the platform
Configure Google Ads automated rules as your safety net: pause a keyword when CPA exceeds your threshold for a set window, alert on sharp week-over-week CTR drops, and cap daily spend per campaign.
Cadence: Weekly for the export-classify-apply cycle (an hour or two once the workbook is built); monthly for the deeper budget reallocation pass.
Path B
Integrate it with MCP connections.
The same loop runs as a scheduled agent workflow: read-scoped connections pull the data, the agent classifies terms and drafts changes inside written guardrails, and only small, reversible actions ship without you. It suits teams with clean conversion tracking who are done copy-pasting CSVs every Monday.
Schedule the read-scoped pulls
Wire the agent — Claude with MCP servers, orchestrated by n8n or Make — to pull search terms, campaign performance, and GA4 landing page data on a weekly schedule. Reads and writes use separate credentials, so the write scope stays as narrow as possible.
Write the guardrails file
Codify the rules in a persistent AGENTS.md instruction file: the classification taxonomy, negative match-type policy, spend caps, CPA thresholds, and exactly which actions require human approval.
Run the classification loop
Each week the agent classifies new search terms, scores waste by spend against conversions, drafts negatives, and models budget shifts. It logs its reasoning against the taxonomy, so you can audit why any term got tagged the way it did.
Act within caps, queue the rest
Negative keyword additions under your agreed weekly limit are applied automatically into shared lists and logged. Budget changes, campaign pauses, and anything above threshold are queued as approval requests with the projected impact attached.
Digest, approve, revert
A weekly digest lands in Slack or email: what was applied, what awaits your approval, what was flagged. Every applied change carries before-and-after values in the audit log, so a bad call is one revert away.
Tools & guardrails
Run it safely.
- Negatives auto-apply only up to an agreed weekly count, and only into shared campaign-level lists; larger batches wait for approval
- Every budget reallocation and campaign pause requires explicit human approval — the agent only auto-applies negatives
- Hard daily spend caps run as native Google Ads automated rules, independent of the agent, so a failed workflow can never overspend
- Every action is written to an audit log with before-and-after values, so any change can be reversed from the log
Want new build guides as they ship?
Optional — every guide on this site is free either way. Leave your details and we'll email you when new automations join the library, nothing else.
Prefer it built for you?
We stand this automation up as a governed sprint on the stack you already run — scoped, built, and handed over with the guardrails configured. No new platform, no migration.