Campaign Performance Diagnosis and Optimization

Objective: Equip marketers and growth teams with a reusable AI prompt that systematically reviews campaign metrics, isolates underperforming elements, generates evidence-based hypotheses, prioritizes experiments, and delivers clear next actions.

The base prompt was formed through these deliberate steps:

Map the exact diagnostic sequence required by the use case so the AI never skips a stage.
Separate fixed analytical instructions (snippets) from campaign-specific inputs (variables) to keep the prompt modular and reusable.
Embed explicit prioritization logic so recommendations are ranked by impact and effort rather than listed generically.
Require structured output formats that force the model to surface both quantitative evidence and qualitative reasoning.
Include guardrails that prevent vague advice and keep every recommendation tied to observed data.

Note

The prompt treats the AI as a methodical analyst rather than a creative brainstormer. Every output must reference specific metrics supplied in the input.

Prompt Template

You are a senior performance marketing analyst. Your task is to diagnose campaign results and produce a prioritized optimization plan.

Campaign context:
{{campaign_context}}

Primary goal:
{{primary_goal}}

Raw performance data:
{{performance_data}}

{{snippet:diagnostic_framework}}

{{snippet:output_structure}}

Tone and style:
{{tone}}

Element	Type	Example Value
{{campaign_context}}	Variable	Meta ads campaign promoting a SaaS free trial, targeting mid-market marketing managers in North America, running for 21 days
{{primary_goal}}	Variable	Increase free-trial sign-ups at a target CPA of $45 or lower
{{performance_data}}	Variable	Spend $12,400; 287 trials; CPA $43.20; CTR 1.8%; CVR 2.1%; top creative CTR 3.4%, bottom creative CTR 0.9%
{{snippet:diagnostic_framework}}	Snippet	Review every metric against benchmarks and goal. Isolate underperforming elements by segment (creative, audience, placement, offer). Hypothesize root causes with supporting evidence. Rank tests by expected impact and implementation effort.
{{snippet:output_structure}}	Snippet	Structure your response exactly as: 1. Metric Snapshot 2. Underperforming Elements 3. Hypotheses with Evidence 4. Prioritized Test Backlog 5. Immediate Next Actions
{{tone}}	Variable	Direct, data-first, and action-oriented

Prompt Example

You are a senior performance marketing analyst. Your task is to diagnose campaign results and produce a prioritized optimization plan.

Campaign context:
Meta ads campaign promoting a SaaS free trial, targeting mid-market marketing managers in North America, running for 21 days

Primary goal:
Increase free-trial sign-ups at a target CPA of $45 or lower

Raw performance data:
Spend $12,400; 287 trials; CPA $43.20; CTR 1.8%; CVR 2.1%; top creative CTR 3.4%, bottom creative CTR 0.9%

Review every metric against benchmarks and goal. Isolate underperforming elements by segment (creative, audience, placement, offer). Hypothesize root causes with supporting evidence. Rank tests by expected impact and implementation effort.

Structure your response exactly as: 1. Metric Snapshot 2. Underperforming Elements 3. Hypotheses with Evidence 4. Prioritized Test Backlog 5. Immediate Next Actions

Tone and style:
Direct, data-first, and action-oriented

Tip

Paste the completed prompt into any capable LLM and attach the actual campaign export or dashboard screenshots for richer diagnosis. Always verify the AI’s prioritization against your own traffic volume and resource constraints.

Extended example values:

{{campaign_context}}
- Google Search campaign for an e-commerce brand launching a new product line, focused on high-intent keywords in the US and Canada
- LinkedIn Sponsored Content series aimed at enterprise IT decision-makers for a cybersecurity tool, active for 45 days
{{primary_goal}}
- Reduce cost per qualified lead below $120 while maintaining lead quality score above 70
- Grow ROAS to 4.5x within the next 14 days without increasing total budget
{{performance_data}}
- Impressions 1.2M, clicks 18,400, CTR 1.53%, conversions 312, CPA $68, best audience CPA $41, worst audience CPA $142
- Video views 89k, 25% completion rate 34%, click-through from video 2.1%, landing-page conversion 1.4%
{{snippet:diagnostic_framework}}
- Compare current metrics to both platform benchmarks and the campaign’s own historical baselines. Flag any element whose performance is more than 25% below the median. Generate at least three ranked hypotheses per underperforming element and link each hypothesis to a concrete data point.
- Segment analysis must cover creative, audience, placement, device, and time-of-day. Reject any hypothesis that cannot be tested within two weeks or that requires unavailable data.
{{snippet:output_structure}}
- Deliver the response in five numbered sections only. In the Prioritized Test Backlog, list each test with expected impact (High/Medium/Low), effort (High/Medium/Low), and a one-sentence success metric. End with no more than three Immediate Next Actions that can be executed in the next 48 hours.
- Use bullet points inside each section. Never add introductory or concluding paragraphs outside the five required sections.
{{tone}}
- Concise and decisive, suitable for a weekly growth team standup
- Analytical yet collaborative, written as if advising a peer marketing manager