Meta Ads Manager

Your always-on optimization system for Meta Ads.

The Meta Ads canon is the input. What Cortex does with it is the output. Sentries watch every account every day for drift against documented best practice. Optimizers act on the findings. Every change is logged, measured, and traceable back to the docs and learnings that produced it.

4,979
Documents indexed
28
Active learnings
6
Source properties
01
Retrieve
Marketing API · CAPI · Pixel · Catalog · GA4 · CRM
02
Analyze
Match retrieved data against the corpus + learnings layer
03
Decide
Apply structured decision protocol with guardrails
04
Act
Execute via API or surface for human review
05
Measure
Track outcome vs intent · feed back into learnings

Meta Ads AI

Meta Ads questions, answered.

Hybrid retrieval · Opus 4.7
Advantage+ Shopping setup CAPI server-side tagging Audience saturation Creative fatigue signals iOS attribution gaps Lookalike audience sizing
Sample answer
Cortex · answering with 7 sources
How should I set up CAPI deduplication so I don't double-count Purchase events alongside the Meta Pixel?

Meta deduplicates redundant events when the browser and server send identical event_name values plus either the same event_id or the same external_id + fbp pair, all to the same Pixel ID inside a 48-hour window[1]. Outside that window, both events are kept, so a delayed server-side fire after a connectivity issue will still be counted twice[2].

The recommended setup is to mint a single unique event_id per user action server-side, then pass it to the browser via the eventID field on fbq('track', 'Purchase', {...}, {eventID}) and to the server via event_id on the Conversions API request[3]. Both events must also share an identical event_name; spelling drift between Purchase and purchase defeats the matcher[4].

Verify the result in Events Manager under the Event Deduplication tab - the Rate of Events Deduplicated should sit between 50% and 80% for a healthy redundant setup, with Overlap > 60% indicating the dedupe key is reaching both sources[5]. Lower numbers mean a server-only or browser-only segment is leaking into the dataset.

For app events, the same event_id + event_name logic deduplicates Conversions API events against SDK/MMP events[6]. Cortex's CAPI Sentry checks the Dataset Quality API daily and flags any Pixel where deduplication rate falls below 50% or Event Match Quality score drops more than 0.5 week-over-week[7].

Answered in 1.42 seconds · 7 most relevant of 318 sources considered

Citations

[1]Handling Duplicate Pixel and Conversions API Events
developers.facebook.com / documentation / ads-commerce / conversions-api / deduplicate-pixel-and-server-events
officialmarketing-api
[2]Best Practices - Conversions API
developers.facebook.com / documentation / ads-commerce / conversions-api / best-practices
officialmarketing-api
[3]Conversions API End-to-End Implementation
developers.facebook.com / documentation / ads-commerce / conversions-api / guides / end-to-end-implementation
officialmarketing-api
[4]Omni Optimal Technical Setup Guide
developers.facebook.com / documentation / ads-commerce / marketing-api / best-practices / omni-optimal-setup-guide
officialmarketing-api
[5]Verifying Your Setup
developers.facebook.com / documentation / ads-commerce / conversions-api / verifying-setup
officialmarketing-api
[6]Conversions API for App Events
developers.facebook.com / documentation / ads-commerce / conversions-api / app-events
officialmarketing-api
[7]Cortex learning L-0312 · CAPI deduplication health guardrail
cortex_learnings · evidence: 11 chunks
learningconditional_rule

Meta Ads FAQ

What people actually ask.

Last 30 days

The Cortex Meta Ads Management Playbook

Rules, patterns, and antipatterns - applied automatically.

28 active Meta Ads learnings
conditional ruleL-0312
Don't trust deduplicated conversion counts when the Event Deduplication tab shows a rate below 50% or Overlap below 60%. Fix dedupe key coverage first, attribution second.
Sources 11Confidence 94%
patternL-0418
Lookalikes built from a 1k-5k high-value customer seed consistently outperform 10k+ newsletter-subscriber seeds. Seed quality dominates seed size above the 100-person minimum.
Sources 9Confidence 87%
antipatternL-0153
Never push more than 4 ad set budget changes per hour or more than 10 account spend limit changes per day. Both throttles return a hard block, not a soft warning.
Sources 8Confidence 96%
Capconvert / Cortex · Meta Ads4,979 docs · 28 learnings · last refresh 5.21.2026