Inteligência semanal de tendências e públicos

INPUT

• Campaign-level performance data, two queries: one for the last 7 days, one for the prior 7 days (used as baseline)
• Audience targeting details

PROCESS

SECTION 1 — Trend Analysis

Step 1.1 — Baseline Calculation

Using the two insight calls above, calculate the 7-day rolling average for each campaign. Use the prior 7-day period as the baseline to compare against the most recent 7 days.

Step 1.2 — Flag Declining Campaigns

Flag campaigns where ANY of the following is true for 3+ consecutive days:

• CPA rose more than {{{CPA_RISE_PCT}}}% vs the prior 7-day rolling average AND current CPA exceeds 1.5× TARGET_CPA (absolute floor safeguard)
• CVR dropped more than {{{CVR_DROP_PCT}}}% week-over-week
• ROAS declined more than {{{ROAS_DECLINE_PCT}}}% vs prior period (only if revenue data available)

(Do not verify whether creative or audience was changed — apply the metric condition and flag it.)

Step 1.3 — Root Cause Classification

For each flagged campaign, classify the most likely root cause based on available metrics:

Observed pattern	Suspected cause
Frequency rising + CTR declining	creative_fatigue
Reach plateaued + frequency high	audience_saturation
CPM rising + no CTR improvement	auction_competition
Cannot explain with above patterns	external_factor

When classifying, state the evidence clearly (e.g., "CTR dropped from X% to Y% over 7 days while frequency climbed from Z to W, suggesting creative_fatigue"). Do not present the classification as a definitive fact — frame it as the most likely explanation given the data.

Step 1.4 — Output Trend Report

For each flagged campaign output:

• Campaign ID
• Metric trend line (daily values for CPA, CVR over the 7-day window; ROAS included if revenue data is available)
• Days in decline
• Suspected root cause with supporting evidence
• Recommended early intervention

SECTION 2 — Audience Overlap Analysis

Step 2.1 — Overlap Risk Assessment

Meta does not expose a direct audience overlap percentage via API. Use targeting data (age/gender/location) to identify overlap risk by comparing targeting parameters across active ad sets. Flag pairs where targeting parameters indicate likely significant overlap and no detected exclusion logic exists.

This is an approximate assessment — exact overlap % is not available via current tools.

Step 2.2 — Flag and Quantify

For each flagged pair output:

• Ad set pair IDs
• Overlap risk level (High / Medium / Low, based on targeting similarity)
• Estimated budget waste from internal competition:
  • Low — less than 5% of combined spend
  • Medium — 5–15% of combined spend
  • High — more than 15% of combined spend
• Recommended fix: add_exclusion / consolidate / adjust_targeting

OUTPUT

📈 SECTION 1: TREND ANALYSIS

Campaign ID	Metric	Daily Values (7d)	Days in Decline	Root Cause	Recommended Intervention

If no campaigns flagged: ✅ No campaigns showing significant decline trends this week.

🔍 SECTION 2: AUDIENCE OVERLAP

Ad Set A	Ad Set B	Overlap Risk	Est. Budget Waste	Recommended Fix

If only one campaign with one ad set active: ✅ No overlap risk — single ad set structure. If targeting data is insufficient to assess overlap: ⚠️ Cannot assess overlap — targeting data not sufficient via current tools. Manual review recommended.

GUARD

• If fewer than 7 days of historical data available: output ⚠️ Insufficient historical data for trend analysis (need 14 days minimum). Output available data only.
• Do not execute any changes automatically regardless of findings. Output recommendations only.

CONFIG (user-configurable)

• CPA_RISE_PCT: 10
• CVR_DROP_PCT: 15
• ROAS_DECLINE_PCT: 10 (only applicable if revenue data available)
• TARGET_CPA (required): user-defined target CPA used for absolute floor safeguard