Your campaign is underperforming. ROAS is below your break-even, CPA is higher than your target, or results have dropped off after a period of solid performance. You know something needs to change but you're not sure what.
The wrong move is to start changing things randomly. The right move is to diagnose the problem first, then fix the specific cause.
Claude is very good at this if you give it the right data and ask the right questions. Here's the exact workflow.
Before you touch anything
If a campaign has been running for fewer than 7 days or has fewer than 50 conversions, it may still be in the learning phase. Underperformance during this window is normal. Making changes resets the learning phase and delays optimization further. Wait for the data before diagnosing.
Step 1: Get Claude to diagnose the problem
Start with a diagnosis prompt before trying to fix anything. Changing the wrong thing wastes time and budget.
Diagnosis prompt (with live AdAdvisor connection): 'My campaign [name] has been underperforming for the last 7 days. ROAS is below my break-even. Run a diagnosis. Check creative frequency, CTR trends, audience overlap, budget pacing, and CPM trends. Tell me the most likely cause and rank any other possible causes.'
Diagnosis prompt (with exported data): 'I'm attaching 30-day performance data for a Meta campaign. It's been underperforming for 7 days. ROAS is [X] against a break-even of [Y]. Diagnose the most likely cause. Check for: creative fatigue signals, audience issues, bidding problems, landing page or offer mismatch, and budget pacing. Rank your findings by most to least likely.'
The key is asking Claude to rank findings. You want to know which problem to fix first, not a list of everything that could theoretically be wrong.
The four most common causes and how Claude diagnoses each
The four most common causes and how Claude diagnoses each
Creative fatigue
Frequency above 2.5, CTR declining week over week, CPM rising without a change in targeting. Claude catches this quickly with a frequency and CTR trend check. Prompt: 'Check the frequency and CTR trend for each active ad in this campaign over the last 14 days. Flag any ads where frequency is above 2.5 or CTR has dropped more than 20% week over week.' Fix: Pause the fatigued creatives and introduce fresh variations. Do not pause the entire campaign. Do not change targeting. Change the creative only.
Audience saturation or overlap
Delivery slowing down, CPM rising, frequency climbing even on newer creatives. Audience overlap between ad sets drives up costs because you're bidding against yourself. Prompt: 'Are any of my active ad sets targeting audiences that significantly overlap? Flag any where overlap is likely above 20% and estimate the CPM impact.' Fix: Add the smaller, more specific audience as an exclusion in the broader ad set. Do not delete any audiences. Just add exclusions.
Budget pacing issues
Spend spiking early in the day and running out by afternoon. Or budget barely spending at all, which usually means CPM bids are too low to win auctions. Prompt: 'Check the hourly spend distribution for the last 7 days. Is the budget pacing evenly across the day or is it front-loaded? Is there any day where budget ran out before midnight?' Fix: For front-loading, try campaign budget scheduling to cap spend during low-performance hours. For underspending, increase your bid cap or switch from manual to cost cap bidding.
Offer or landing page mismatch
Strong CTR but low conversion rate. People are clicking and then not converting, which means the problem is after the ad, not the ad itself. Prompt: 'This campaign has a CTR of [X] but a conversion rate of [Y]. The drop-off happens after the click. What are the most likely explanations? Check if there's a mismatch between ad messaging and landing page, and whether the landing page load time or mobile experience could be a factor.' Fix: This one is usually outside Ads Manager. Check that the landing page matches the ad's promise, loads in under 3 seconds, and is fully functional on mobile.
Step 2: Fix one thing at a time
The most common mistake when fixing an underperforming campaign is changing multiple things at once. When you change the creative and the audience and the bid strategy in the same week, you don't know which change made the difference.
Fix the highest-ranked cause from your diagnosis first. Give it 5 to 7 days of data. Then assess and fix the next one if needed.
Ask Claude to prioritize
After your diagnosis, ask: 'If I can only make one change this week, which would have the highest impact on performance based on what you've found?' This forces a prioritization rather than a list of suggestions that all feel equally urgent.
Step 3: Monitor the fix
After making a change, use Claude to monitor whether it's working.
Monitoring prompt: 'It's been 5 days since I refreshed the creatives on this campaign. Compare ROAS, CTR, and frequency to the 5 days before the change. Is the fix working?'
This closes the loop. You're not just making changes and hoping. You're checking whether each change moved the metrics in the right direction before making the next one.
When a campaign is genuinely past saving
Sometimes the diagnosis is that the offer, the pricing, or the audience doesn't work. No amount of creative or targeting optimization fixes a product-market fit problem.
Signal to watch for: If you've refreshed creative, fixed audience overlap, and checked the landing page, and ROAS is still below break-even after 30 days of meaningful spend, the problem is upstream of the campaign. Ask Claude: 'Based on 30 days of this campaign data, is there evidence of a fundamental offer or audience mismatch rather than a technical campaign issue?'
A clear-eyed answer to that question is more valuable than continuing to optimize a campaign that isn't fixable.
If you're running this diagnosis with AdAdvisor's MCP connected, Claude already has your break-even ROAS and campaign history in context. No data export needed.
