campaign-analyzer

Process

1. Clarify the question — What specific insight is the user trying to extract? ("Why is CPA rising?" vs "What's the LTV by channel?")
2. Select analysis type — Match the question to the right framework (pulse, strategic, dimensional, funnel, cohort, incrementality, cross-channel)
3. Gather required data — Request exports needed for the analysis
4. Run the analysis — Apply the specific method
5. Identify the story — What happened, why it happened, what to do about it
6. Deliver findings — Structured output with insights, not just metrics

Analysis Type Selector

Daily Pulse Monitoring

What to check:

• Cost, impressions, clicks, conversions, spend pacing vs budget
• Sudden metric changes (CPC spikes, conversion drops, budget exhaustion)
• Meta: learning phase status per ad set
• Disapprovals, policy flags, payment issues

Alert thresholds: Automated alerts for >20% deviation from 7-day averages on key metrics.

What NOT to do: Don't optimize during daily pulse. Daily review is for catching emergencies, not making strategic changes. Save optimization decisions for weekly reviews.

Cadence insight (Supermetrics research): Most successful performance marketers prioritize basic metrics (cost, impressions, clicks, conversions) for daily decisions. Deeper analysis (frequency, LTV, attribution) is purposeful, not constant.

Weekly/Monthly Strategic Review

Key metrics by platform:

Review structure:

1. Scaling candidates — Campaigns with strong ROAS + available impression share (Google) or frequency headroom (Meta) → recommend budget increases
2. Pause/pullback candidates — Campaigns with high CPA + declining trend → recommend reductions
3. Creative performance — Which ads/creatives are winning, which are fatiguing
4. Trend direction — Is the account improving, stable, or declining over trailing 4 weeks?

Multi-Dimensional Performance Analysis

Slice performance across dimensions to find hidden drivers.

Dimensions to Analyze

Cross-Tab Analysis

Combine dimensions to find non-obvious patterns:

Red Flags to Watch

• One dimension dominating cost but not conversions → Budget misallocation
• Large variance across segments → Opportunity to isolate winners
• Hidden loss leaders → Specific segments dragging average down

Funnel Analysis

Map every stage from impression to final conversion.

Standard Funnel Stages

Calculate stage-by-stage conversion rates. Identify the biggest drop-off points — that's where to focus optimization.

Drop-Off Diagnostics

Meta-specific (June 2025): Instant Experience is no longer counted as a landing page view. If your Meta funnel metrics suddenly showed a drop in mid-2025, this is likely why. Adjust funnel definitions accordingly.

Cohort & Lifetime Value Analysis

Cohort Windows

Track customer cohorts at standard intervals:

• Day 1, 7, 14, 30 (short-term retention)
• Day 30, 60, 90 (medium-term)
• Day 180, 365 (long-term LTV)

Metrics to Track per Cohort

LTV by Acquisition Source

Compare LTV by channel, campaign, and audience segment:

What this reveals: Channels with higher short-term CPA may still be profitable if LTV is strong. Channels with low CPA but poor LTV are actually expensive.

POAS > ROAS

POAS (Profit On Ad Spend): Gross profit ÷ ad spend. Better than ROAS for ecommerce because it factors in actual product margin.

Example: $500 ROAS looks great until you learn the product has 8% margin → true profit is $40 per $100 spent. Another channel at 300% ROAS with 40% margin returns $120 profit per $100 spent.

Implementation: Feed LTV data back into value-based bidding. Use Google's Maximize Conversion Value + tROAS and Meta's ROAS Goal with dynamic conversion values.

Incrementality & Lift Testing

Platform-reported ROAS often overstates channel impact (credit-grabbing from organic demand).

Methods

Questions Incrementality Answers

• "If I turned off YouTube, what happens to non-brand search?"
• "If I turned off Meta prospecting, does total revenue change?"
• "Is brand search capturing demand created by upper-funnel channels?"
• "Are retargeting conversions incremental or would they have happened anyway?"

MMM + incrementality calibration: Google's Meridian (open-source MMM, 2025) combines granular media signals with incrementality calibration. The right way to do cross-channel budget planning at scale.

Cross-Channel Analysis

Marketing Efficiency Ratio (MER)

MER = Total Revenue ÷ Total Ad Spend (all channels)

• Blended metric. Resilient to attribution changes
• Goal: maintain or improve over time
• Doesn't tell you which channel deserves credit, but tells you overall efficiency

Marginal ROAS

As you increase spend in a channel, each additional dollar produces less return (diminishing returns). The question: at what spend level does marginal ROAS drop below target?

Method:

1. Plot spend vs revenue for a channel over 6-12 months
2. Fit a diminishing returns curve
3. Find the spend level where marginal ROAS = target ROAS
4. That's your efficient frontier for that channel

Channel Role Assignment

Not every channel should optimize to the same metric. Assign roles:

2025 Google feature: Import cost data from Meta, TikTok, Pinterest directly into GA4 for cross-platform comparison in one place.

For multi-channel budget reallocation, use the ad-spend-allocator skill.

Output Format

## Campaign Performance Analysis

### Question Analyzed
[Specific question the analysis addresses]

### Data Sources
- [Platforms]
- [Date range]
- [Segments included]

### Key Findings
1. **[Finding 1]** — [Data point] — [Why this matters]
2. **[Finding 2]** — [Data point] — [Why this matters]
3. **[Finding 3]** — [Data point] — [Why this matters]

### Supporting Data
[Tables and metrics supporting findings]

### Narrative
[Story: what happened, why it happened, what to do]

### Recommended Actions
| Priority | Action | Impact | Owner | Timeline |
|----------|--------|--------|-------|----------|
| 1 | [action] | [expected impact] | [who] | [when] |

### Data Confidence & Caveats
- [What the data supports reliably]
- [What needs more information or isn't certain]

Example

Input: "Our ecommerce account's blended ROAS dropped from 3.8x to 2.9x over the last quarter. Google is steady, Meta declined hard. I have 90 days of data by channel, campaign, and creative. Help me figure out what's happening."

Output:

Campaign Performance Analysis

Question Analyzed

Why did blended ROAS decline from 3.8x to 2.9x in Q1 2026, and which channel/campaign/creative drove the decline?

Data Sources

• Google Ads + Meta Ads exports
• 90-day window (January–March 2026)
• Segmented by channel, campaign, creative

Key Findings

1. Meta drove 85% of the blended decline — Google ROAS held at 4.1x (down from 4.2x — flat). Meta ROAS dropped from 3.5x to 2.1x (-40%). Blended decline is almost entirely Meta
2. Meta decline is creative fatigue, not audience or tracking — Top 3 creatives running 50+ days each. Average creative frequency climbed from 3.1 to 5.8 over 90 days. Meta research: 45% conversion drop at 4+ exposures
3. CPM inflation contributed ~15% of Meta decline — Meta average ad price up 9% in 2025. For this account, CPMs rose from $12.40 to $14.20 (+15%). This is real cost pressure, not just creative failure
4. Google is stable because Smart Bidding is managing efficiency — PMax and Search are both on tROAS with regular n-gram review. No similar discipline on Meta creative

Supporting Data

Channel-level ROAS trend:

Meta creative frequency trend:

Narrative

Your blended ROAS decline is a Meta creative fatigue story with a CPM inflation chaser. Google is fine — Smart Bidding + disciplined n-gram review keeps it healthy. On Meta, you've run the same 3 creatives for 50+ days, frequency is at 5.8 across top creatives, and Meta's research shows 45% conversion rate drop at 4+ exposures. Meta's CPM increase (+9% industry-wide in 2025, +15% for your account) adds background pressure but is secondary — even at old CPMs, fatigue would still be dragging ROAS.

The silver lining: your one recent creative addition (D — UGC video, frequency 3.1) is performing well. This is the signal that creative rotation works for your audience; you just haven't been doing it.

Recommended Actions

Data Confidence & Caveats

• High confidence: Creative fatigue is the primary driver. Frequency data, ROAS decline timing, and the Creative D performance all point to the same conclusion
• Medium confidence: CPM inflation contribution (~15%) is a rough estimate based on industry averages; your specific CPM trend confirms it's real but the exact split with creative fatigue can't be precisely calculated without an A/B test
• Untested: Incrementality — platform-reported ROAS may overstate true channel impact. Consider running a Meta Conversion Lift test in Q2 to validate blended ROAS represents actual business lift
• Not included: LTV analysis — if Meta retargeting has higher LTV, the CPA increase may be tolerable. Recommend 90-day LTV analysis as follow-up
• Use the meta-ads-creative-engine skill for creative testing framework implementation
• Use the ad-spend-allocator skill for cross-channel budget reallocation after creative rotation stabilizes

Guidelines

• Don't confuse correlation with causation. Two trends moving together (ROAS down, frequency up) correlate. To confirm causation, test the intervention (refresh creative, see if ROAS recovers).
• Don't analyze without a question. "Tell me about the account" produces vanity dashboards. "Why did CPA rise in March?" produces insight. Always start with the specific question.
• Don't ignore base rates. An account with 10 conversions/day has huge daily variance. A change that looks significant may be noise. Require statistical meaningfulness (50+ conversions per segment for reliable segment analysis).
• Don't overweight platform-reported ROAS. Both Google and Meta take credit for conversions that may have happened anyway. For strategic decisions, triangulate platform ROAS with MER, incrementality, and LTV.
• Don't report metrics without narrative. "CPA is $62" is data. "CPA rose from $48 to $62 because creative fatigue hit in mid-February, same period frequency exceeded 4, confirmed by Creative D's outperformance proving rotation works" is insight. Insight drives action.
• Don't skip segmentation. Account averages hide winning and losing segments. Always break down by device, geo, audience, creative, placement. Hidden patterns drive hidden wins.
• Meta Instant Experience change (June 2025): No longer counts as a landing page view. If Meta funnel metrics look broken mid-2025, this is why.
• Meta attribution API change (October 2025): 7d_view and 28d_view windows removed from Insights API. Reports pulling these need updates.
• Don't overreact to a single week. Weekly variance is normal. Look for 2+ week trends before declaring changes.
• Cross-references: For Google-specific optimization actions, use the google-ads-optimizer skill. For Meta-specific optimization, use the meta-ads-optimizer skill. For creative testing frameworks, use the meta-ads-creative-engine skill. For cross-channel budget reallocation, use the ad-spend-allocator skill. For reporting the analysis to stakeholders, use the paid-media-reporter skill. For full audits, use google-ads-audit or meta-ads-audit.
• Confidence: Channel attribution is inherently imperfect. All multi-touch attribution models make assumptions. State what the data supports clearly and what requires incrementality testing to confirm.