How to Analyse Google Ads Performance with Claude and Ooty Falcon

Google Ads produces more data than most advertisers know what to do with. Campaigns, ad groups, keywords, search terms, audiences, devices, locations, time of day -- every dimension is available, spread across dozens of report views in an interface designed for granularity, not synthesis.

The problem is not access. It is synthesis. When something goes wrong with your account, figuring out why typically means opening six different reports, exporting spreadsheets, cross-referencing numbers, and spending 45 minutes on analysis that should take 10.

This tutorial shows how to use Claude with Ooty Falcon to run your entire Google Ads analysis in a conversation. Falcon pulls the data. Claude does the thinking. You make the decisions.

The AI Ads Analysis Workflow

Nine steps from account overview to prioritised action list -- in one session

1

Account Overview

Spend, CTR, conversions, ROAS across all campaigns

2

Keyword Performance

Best and worst keywords by ROAS and spend

3

Search Term Audit

Find irrelevant queries burning budget

4

Budget Pacing

Identify campaigns hitting limits or under-spending

5

Period Comparison

What changed and why it changed

6

Ad Copy Review

Which variants win and which drag CTR

7

Diagnostic Deep-Dive

Trace a symptom to its root cause

8

Report Summary

Plain-language summary for stakeholders

9

Action List

Prioritised changes with expected impact

What You Need

• An Ooty account with Falcon activated -- ooty.io
• Claude Desktop (or any MCP-compatible client) with Falcon connected via MCP endpoint and license key
• Google Ads connected via OAuth in your Ooty dashboard (Falcon > Connect > Google Ads)
• Your Google Ads customer ID (the 10-digit number in the top right of Google Ads)

Falcon connects to Google Ads through OAuth -- you authorise Ooty to read your account data through Google's official Ads API. No credentials are shared with Claude directly.

What Falcon Gives Claude

When Falcon is connected, Claude can access:

• getCampaignPerformance -- campaign-level metrics with anomaly detection
• getKeywordPerformance -- keyword-level clicks, cost, conversions, ROAS
• getSearchTerms -- actual search queries triggering your ads
• getBudgetPacing -- spend vs budget tracking across campaigns
• getAccountHealth -- account-level health signals and alerts
• getAdCopyPerformance -- ad variation performance comparison
• getPerformanceComparison -- period-over-period comparisons

You ask questions in plain language. Claude decides which tools to call.

C

Prompt to Claude

“My overall ROAS has dropped from about 5x last month to 3.2x this month. Walk me through a diagnostic -- what are the most likely causes and how do we find which one is actually happening?”

Falcon runs:getCampaignPerformancegetKeywordPerformancegetSearchTermsgetBudgetPacing

Step 1: Get Your Account Overview

Start any analysis session with a broad picture. This gives Claude context before you drill into specifics.

Prompt:

"Give me an overview of my Google Ads account for the last 30 days. I want total spend, impressions, clicks, CTR, conversions, CPA, and ROAS. Flag anything that looks off -- campaigns with unusual spend, very low CTR, or zero conversions."

Claude calls getCampaignPerformance and getAccountHealth, then synthesises the results. Instead of dumping raw data, it highlights what matters: "Your Search campaigns are spending $1,240 with a 4.2x ROAS, which is strong. One campaign -- 'Brand Protection' -- has consumed 18% of your budget with zero conversions this period. That is worth investigating."

That immediate triage takes multiple report views and manual cross-referencing in the standard Google Ads interface.

Step 2: Identify Your Best and Worst Keywords

Keywords drive search campaign economics. Understanding which ones create value and which waste budget is the foundation of optimisation.

Prompt:

"Pull my keyword performance for the last 30 days. Show me which keywords have the best ROAS, which have the worst, and which are spending the most with no conversions. Filter to keywords with at least $20 in spend so we skip statistical noise."

Claude calls getKeywordPerformance with your filters and returns a prioritised view. Two patterns usually emerge: keywords spending significantly with zero conversions (candidates for pausing or bid reduction), and high-performing keywords that may be under-bid relative to their value.

Follow-up:

"For the keywords with high spend and no conversions -- what search terms are actually triggering them? Are the landing pages matched to the intent?"

Step 3: Audit Your Search Terms

The gap between keywords you bid on and the queries actually triggering your ads is where wasted spend hides. Broad match and phrase match keywords can trigger searches with entirely different intent.

Prompt:

"Show me my top search terms by spend for the last 30 days. Flag any that look irrelevant to my business -- I sell [your product/service]."

Claude calls getSearchTerms and you will typically find a mix: highly relevant queries performing well, and irrelevant queries burning money. A company bidding on "project management software" with broad match might find ads triggering for "free project management templates" or "project management certification courses" -- queries from people who are not buying software.

According to WordStream's analysis of Google Ads accounts, the average account wastes 76% of its budget on search terms that never convert. Even well-managed accounts typically find 10-15% of spend going to irrelevant queries each quarter.

Follow-up:

"From the search terms, which ones should I add as negative keywords? Give me a list I can paste directly into Google Ads."

That is a 30-minute manual task reduced to a 2-minute conversation.

Step 4: Check Budget Pacing

If campaigns exhaust their daily budget by midday, you miss afternoon conversions. If they consistently under-spend, targeting may be too restrictive.

Prompt:

"Check budget pacing across all active campaigns. Are any hitting their daily budget limit before end of day? Are any significantly under-spending? Tell me if I should reallocate."

Claude calls getBudgetPacing and returns pacing status across campaigns. Budget-limited campaigns get recommendations to increase budget (if ROAS justifies it) or adjust bidding to spread spend more evenly. Under-spending campaigns might signal targeting issues or overly restrictive negative keyword lists.

Google's own documentation notes that budget-limited campaigns can miss 15-30% of eligible impressions depending on competition and time-of-day demand patterns.

Step 5: Compare Period Performance

Performance changes need context. A 20% drop in conversions means something different if it follows a 30% increase last month versus a flat trend.

Prompt:

"Compare my campaign performance this month vs last month. What changed, which campaigns improved, which declined, and what is the most likely explanation for the biggest changes?"

Claude calls getPerformanceComparison and reasons about causes: "Your Non-Brand Search campaign saw a 34% drop in conversions. Impression share held steady but CTR dropped from 4.8% to 3.1%. This suggests ad copy fatigue -- the ads are showing but people are clicking less. Conversion rate when they do click is stable, so the landing page is fine."

That kind of diagnostic reasoning separates useful analysis from staring at a chart that went down.

Diagnostic Reasoning

How Claude traces a ROAS drop through multiple data points to find the root cause

Symptom

ROAS dropped

Which campaigns dropped?

Shopping campaigns, not Search

Did CPAs increase?

Yes, +45% in Shopping

Did click volume change?

No -- similar clicks, higher cost

Search term quality?

Stable -- not the cause

Root Cause

Bid strategy or product feed issue in Shopping campaigns

Step 6: Review Ad Copy Performance

Ad variants that do not compete effectively drag overall CTR and Quality Score. Regularly reviewing which copy works -- and pausing what does not -- compounds performance over time.

Prompt:

"Show me ad performance by copy variation for my top 3 campaigns by spend. I want CTR, conversion rate, and ROAS for each variant. Flag anything significantly underperforming the top variant."

Claude calls getAdCopyPerformance and presents a comparison. Common finding: one or two ad variants carry most of the performance while others consume impression share with lower CTRs. Google's own responsive search ad documentation recommends removing consistently underperforming assets to improve overall ad strength.

Step 7: Investigate a Specific Problem

Analysis often starts with a symptom: "My ROAS dropped" or "Why is this campaign spending so much?" Conversational analysis handles diagnostic workflows well because Claude can run multiple tool calls in sequence without you directing each one.

Prompt:

"My overall ROAS dropped from about 5x last month to 3.2x this month. Walk me through a diagnostic -- what are the most likely causes and how do we find which one is actually happening?"

Claude structures a diagnostic without you directing each step: check campaign-level ROAS changes, look for budget pacing shifts, review search term quality, check for bid strategy changes. It runs the tool calls and synthesises findings: "The drop is concentrated in your Shopping campaigns. CPA increased 45% while click volume stayed similar. This points to either bid strategy changes or product feed quality issues."

Step 8: Generate a Report Summary

If you report to clients or stakeholders, formatting performance data clearly is a recurring task.

Prompt:

"Generate a weekly performance summary for my Google Ads account, this week vs last week. Format as bullet points for a client: overall spend and ROAS, top performing campaign, biggest opportunity, one thing we are fixing this week."

Claude pulls the data, runs the comparison, and writes a summary in plain language -- not the data-dense format Google Ads exports, but a clear narrative a non-technical stakeholder can read in 60 seconds.

Step 9: Build Your Action List

Analysis without action is just reporting. End every session with a prioritised list of changes.

Prompt:

"Based on everything we have reviewed -- campaign performance, keywords, search terms, budget pacing, ad copy -- give me a prioritised action list. What are the three most impactful changes I should make this week?"

Claude synthesises the full conversation: "1. Add 23 negative keywords to Non-Brand Search (stops roughly $200/week in wasted spend). 2. Pause the 3 ad variants with CTR below 2% in Brand campaign and direct budget to the top performer. 3. Increase daily budget for Top Products Shopping by 30% -- it hits its limit by 2pm and leaves conversions on the table."

Analysis Time Comparison

Minutes per task: manual Google Ads workflow vs Claude + Falcon

145m

Manual

14m

With Falcon

90%

Time saved

Account overview15m 1m

Keyword analysis25m 2m

Search term audit30m 3m

Budget pacing check10m 1m

Period comparison20m 2m

Ad copy review15m 2m

Report writing30m 3m

Manual workflowClaude + Falcon

Tips for Better Ads Analysis

Start with what is broken. The most time-valuable analysis is diagnostic. "What looks wrong?" before "How do I optimise what already works?"

Give Claude your goals. "I prioritise ROAS over volume" or "I am in growth mode and care about conversion volume more than CPA" changes the analysis significantly. Claude needs your objective to give relevant recommendations.

Do not analyse in isolation. Google Ads performance connects to landing pages, offer quality, seasonality, and competitor activity. Tell Claude if anything changed recently -- it changes how to interpret the numbers.

Ask for explanations. "Why might this be happening?" produces better output than "What is the ROAS for campaign X?" The synthesis is where the value sits.

Verify before acting. Falcon pulls accurate data from the Google Ads API. But recommendations about what to do with that data should be reviewed before execution. Claude can make a case for pausing a keyword that you know has strategic value for reasons not visible in the data.

Common Mistakes

Optimising with insufficient data. Week-over-week comparisons for small accounts with low click volumes can mislead. Google recommends at least 30 conversions per month before drawing conclusions about conversion-based bidding strategies.

Ignoring impression share. High spend plus high conversions looks good. But if impression share is low, you are leaving performance on the table by under-bidding. Always ask about the full picture.

Treating the search terms report as a one-time task. Search terms drift as match types broaden over time. Adding negatives is an ongoing process -- quarterly at minimum, monthly for active accounts.

Not asking about the landing page. If CTR is fine but conversion rate is low, the problem is often on the landing page, not in Google Ads. Falcon can tell you about ad performance; the landing page requires a different investigation.

What This Replaces

A typical manual Google Ads analysis session involves logging into Google Ads, opening the Campaigns report, opening Keywords in another tab, opening Search Terms in a third, downloading data for comparison, cross-referencing in spreadsheets, and writing up findings. Total: 2-3 hours for a thorough review.

With Falcon and Claude, data retrieval and synthesis happen together. The analyst's time shifts from "fetching and formatting data" to "deciding what the data means and what to do about it." That is a qualitative change in how PPC analysis works.

Get started with Ooty Falcon