SealMetrics + Meta Ads
Meta's reported numbers are an estimate, especially post-iOS 14.5. SealMetrics gives you the deterministic baseline. These prompts compare the two.
MCPs required: SealMetrics MCP + Meta Ads MCP
SEAL-106 — Meta ROAS vs real ROAS
You have access to SealMetrics MCP for site {site_id} and Meta Ads MCP for account {meta_account}.
Last 30 days per ad set:
- Spend (Meta Ads)
- Conversions and revenue reported by Meta
- Conversions and revenue captured by SealMetrics (matched on utm_campaign + utm_source = facebook|instagram + utm_medium = paidsocial|cpc)
- ROAS_meta and ROAS_real
- Delta
Highlight ad sets where Meta ROAS overstates real ROAS by more than 0.5x. Those are likely benefiting from modeling on iOS users.
Output: 1 ranking table + 3-bullet bidding recommendation.
SEAL-107 — Conversion gap post-iOS 14.5
Using SealMetrics MCP for site {site_id} and Meta Ads MCP for account {meta_account}:
Per ad set, last 30 days:
- Conversions reported by Meta (split by iOS / Android / Web if available)
- Conversions in SealMetrics by source = facebook|instagram, broken down by device platform
Compute % under-reporting on iOS vs Android. Highlight ad sets where the iOS gap is larger than 30% — those are where Meta's modeling matters most.
End with a recommendation on which ad sets to optimize using SealMetrics conversions instead of Meta's reported.
SEAL-108 — High CPM, low quality traffic
For site {site_id} (SealMetrics MCP) and Meta account {meta_account} (Meta Ads MCP):
Per campaign, last 30 days:
- CPM (Meta)
- Bounce rate of incoming SealMetrics traffic
- % bot/agent_suspected from that source
List campaigns with above-account-median CPM AND above-site-median bounce rate AND above 5% bot share. Those are paying premium for low quality.
Output: ranked table by spend + a 3-bullet recommendation: pause, change audience, change placement.
SEAL-109 — Audience ROAS ranking
Using SealMetrics MCP for site {site_id} and Meta Ads MCP for account {meta_account}:
Per audience (saved or custom), last 60 days:
- Spend (Meta)
- Real conversions and revenue (SealMetrics)
- Real ROAS = revenue_sealmetrics / spend
- Revenue per session
Rank audiences by real ROAS desc. Highlight top 3 to scale and bottom 3 to retire.
End with a 3-bullet recommendation including 2 lookalike candidates to test based on the top performers.
SEAL-110 — Creative fatigue detector
For site {site_id} (SealMetrics MCP) and Meta account {meta_account} (Meta Ads MCP):
Per ad creative, last 14 days:
- Frequency (Meta)
- CTR trend (Meta)
- Real CR trend on the destination landing (SealMetrics)
Identify creatives with frequency above 3 AND declining real CR over the last 2 weeks (more than 15% drop). Those are fatigued.
Output: list of fatigued creatives with absolute spend and lost potential revenue. Recommend swap or refresh per creative.
SEAL-111 — Best creative + landing combination
Using SealMetrics MCP for site {site_id} and Meta Ads MCP for account {meta_account}:
Cross every active creative with the landing page it sends to. Last 30 days:
- Spend
- Real conversions (SealMetrics)
- Real ROAS
Output the top 10 creative-landing combinations by ROAS and the bottom 10. End with a "test plan": 3 new combinations to launch based on what's working.
SEAL-112 — Adsets with high frequency and high bounce
For site {site_id} (SealMetrics MCP) and Meta account {meta_account} (Meta Ads MCP):
Per ad set, last 14 days:
- Frequency (Meta)
- Bounce rate of incoming SealMetrics traffic from that ad set
- Real CR
Find ad sets with frequency greater than 4 AND bounce rate above 65% AND real CR below site median. Those are over-saturated and the audience is no longer responsive.
Output: ranked table + a 3-bullet rotation plan.
SEAL-113 — Placement performance
Using SealMetrics MCP for site {site_id} and Meta Ads MCP for account {meta_account}:
Split last 30 days by placement (Reels, Feed, Stories, Marketplace, Audience Network, etc.):
- Spend
- Real conversions and revenue from SealMetrics, mapped via utm_content or campaign naming convention I provide
- Real ROAS per placement
Output: ranked table. End with a 3-bullet placement reallocation suggestion.
SEAL-114 — Cold vs retargeting cannibalization
For site {site_id} (SealMetrics MCP) and Meta account {meta_account} (Meta Ads MCP):
Split campaigns into prospecting (cold) and retargeting based on naming convention or audience type.
Per group: spend (Meta), real conversions (SealMetrics), real revenue, share of total social conversions.
Estimate cannibalization risk: % of retargeting conversions that are also from users who arrived initially via prospecting in the same lookback window.
Output: a verdict (high / medium / low cannibalization) + a 3-bullet budget reallocation suggestion.