Performance Marketing skill 03 of 13
Conversion Tracking, Consent Mode and Conversions API SKILL.md
The Conversion Tracking skill makes Claude, Codex or another AI agent audit and design tracking for Google Ads, GA4, Meta, TikTok, LinkedIn and Microsoft Ads. It removes duplicate conversions, sets up Consent Mode v2 in consent-required markets, adds enhanced conversions and server-side events such as the Meta Conversions API with deduplication, passes revenue, profit or qualified-lead values and imports offline outcomes.
conversion-tracking-capi/SKILL.md, version 1.0.0, 1,612 words.
By Afeez Adebayo, Marketing, SEO & Growth Expert with 10+ years of experience.
Created 10 October 2026. Updated 10 October 2026. Free to use under CC BY 4.0.
Every automated bidding system learns from the conversions you send it. Send duplicates and it overspends; send nothing for consent-denied users and it underbids in the markets that matter; send leads that never become customers and it finds more of them. In 2026, good tracking means consent-aware tags, server-side events, deduplication, offline outcomes and values that reflect profit.
This skill audits and designs conversion tracking for Google Ads, GA4, Meta, TikTok, LinkedIn and Microsoft Ads, including Consent Mode, enhanced conversions, the Meta Conversions API and other server-side APIs, and offline conversion imports.
What has changed recently
| Change | What it means | Label |
|---|---|---|
| Consent Mode v2 required for Google Ads features with EEA and UK users since March 2024 | Without it, measurement and remarketing for those users are limited | Confirmed, Google |
| Google discontinued most Privacy Sandbox APIs (October 2025) and kept third-party cookies in Chrome under existing user choice | No sudden cookie cut-off in Chrome; Safari and Firefox still restrict tracking | Confirmed, Google (as reported) |
| From June 2026, Consent Mode is the control governing Google Ads use of identifiers from the Google tag | Consent Mode setup matters even more | Reported |
| Server-side APIs (Meta CAPI, TikTok Events API, LinkedIn CAPI, Google enhanced conversions) are standard practice | Recover signal lost to browsers and ad blockers, with consent | Consensus |
Operating principles
- One source of truth per conversion. Each platform should count each conversion once.
- Consent first. Server-side tracking doesn't remove the need for consent under GDPR and similar laws.
- Optimise for value. Pass revenue, profit or qualified outcomes, not just counts.
- Primary goals drive bidding. Keep only the conversions you want to optimise for as primary.
- Test after every release. Site changes break tags more than anything else.
When to use, and when not to
Use it when results look too good or too bad, after site or checkout changes, when launching in consent-required markets, when platforms show poor event match quality, and for lead businesses that need offline outcomes.
Don't use it for attribution debates or channel allocation (paid-media-reporting, incrementality-testing).
Inputs to ask for, in one message
- List of conversion actions per platform and which are primary.
- Tag setup: Google Tag Manager, gtag, platform pixels, server-side container.
- Consent management platform and markets served.
- Commerce or CRM platform and how orders or leads are recorded.
- Recent site changes and any known discrepancies.
Raw SKILL.md file
This is the exact file agents read, including the credit header and licence notice. Save it as conversion-tracking-capi/SKILL.md.
--- name: conversion-tracking-capi description: "Audit and design conversion tracking for Google Ads, GA4, Meta, TikTok, LinkedIn and Microsoft Ads: one primary conversion per action, Consent Mode v2 in consent-required markets, enhanced conversions, Meta Conversions API and other server-side events with deduplication, revenue, profit or qualified-lead values, offline conversion imports, testing and daily monitoring. Use whenever someone mentions conversion tracking, duplicate conversions, Consent Mode v2, Meta CAPI, server-side tracking, enhanced conversions, offline conversion imports, event match quality or ad platforms reporting different numbers." license: CC-BY-4.0 metadata: author: Afeez Adebayo author_url: https://afeezadebayo.com author_linkedin: https://www.linkedin.com/in/afeez-adebayo/ author_x: https://x.com/za_icon source: https://afeezadebayo.com/skills-md/marketing/performance-marketing/conversion-tracking-capi/ category: marketing subcategory: performance-marketing version: "1.0.0" created: "2026-10-10" updated: "2026-10-10" --- # Conversion Tracking, Consent Mode and Conversions API Every automated bidding system learns from the conversions you send it. Send duplicates and it overspends; send nothing for consent-denied users and it underbids in the markets that matter; send leads that never become customers and it finds more of them. In 2026, good tracking means consent-aware tags, server-side events, deduplication, offline outcomes and values that reflect profit. This skill audits and designs conversion tracking for Google Ads, GA4, Meta, TikTok, LinkedIn and Microsoft Ads, including Consent Mode, enhanced conversions, the Meta Conversions API and other server-side APIs, and offline conversion imports. ## What has changed recently | Change | What it means | Label | |---|---|---| | Consent Mode v2 required for Google Ads features with EEA and UK users since March 2024 | Without it, measurement and remarketing for those users are limited | Confirmed, Google | | Google discontinued most Privacy Sandbox APIs (October 2025) and kept third-party cookies in Chrome under existing user choice | No sudden cookie cut-off in Chrome; Safari and Firefox still restrict tracking | Confirmed, Google (as reported) | | From June 2026, Consent Mode is the control governing Google Ads use of identifiers from the Google tag | Consent Mode setup matters even more | Reported | | Server-side APIs (Meta CAPI, TikTok Events API, LinkedIn CAPI, Google enhanced conversions) are standard practice | Recover signal lost to browsers and ad blockers, with consent | Consensus | ## Operating principles 1. **One source of truth per conversion.** Each platform should count each conversion once. 2. **Consent first.** Server-side tracking doesn't remove the need for consent under GDPR and similar laws. 3. **Optimise for value.** Pass revenue, profit or qualified outcomes, not just counts. 4. **Primary goals drive bidding.** Keep only the conversions you want to optimise for as primary. 5. **Test after every release.** Site changes break tags more than anything else. ## When to use, and when not to Use it when results look too good or too bad, after site or checkout changes, when launching in consent-required markets, when platforms show poor event match quality, and for lead businesses that need offline outcomes. Don't use it for attribution debates or channel allocation (`paid-media-reporting`, `incrementality-testing`). ## Inputs to ask for, in one message 1. List of conversion actions per platform and which are primary. 2. Tag setup: Google Tag Manager, gtag, platform pixels, server-side container. 3. Consent management platform and markets served. 4. Commerce or CRM platform and how orders or leads are recorded. 5. Recent site changes and any known discrepancies. ## Workflow ### Step 1. Inventory and map For each platform: conversion actions, source (tag, GA4 import, API, offline import), primary or secondary, value setting, counting (one or every). ### Step 2. Remove duplicates One primary purchase conversion per platform. If both a native tag and a GA4 import exist, choose one for bidding and set the other as secondary. ### Step 3. Set up Consent Mode Default consent states set to denied before the banner where law requires; update on choice; all four signals (ad_storage, analytics_storage, ad_user_data, ad_personalization). Use advanced mode where your legal basis allows, so Google can model conversions for consent-denied users. ### Step 4. Add server-side events Meta Conversions API, TikTok Events API, LinkedIn Conversions API and Google enhanced conversions, sending hashed first-party data with consent. Deduplicate browser and server events with a shared event ID. ### Step 5. Pass values Revenue for every purchase; consider profit-based values or new-customer values. For leads, assign values by stage and import offline outcomes. ### Step 6. Import offline conversions For lead businesses, send qualified lead, opportunity and closed-won events back to Google, Meta, LinkedIn and Microsoft with click IDs or hashed contact data. ### Step 7. Test Tag Assistant, GA4 DebugView, Meta Events Manager test events and event match quality, platform diagnostics. Make a test purchase or lead and trace it through every system. ### Step 8. Monitor Daily conversion counts by platform compared with orders or leads in your own system; alerts for drops or spikes. ## Tracking plan template ``` | Event | Trigger | Platforms | Method (tag/server/import) | Value | Primary? | Dedup key | Consent required | | purchase | Order confirmed | Google, Meta, TikTok | Tag + server | Revenue (or profit) | Yes | order_id | Yes | | qualified_lead | CRM stage = SQL | Google, Meta, LinkedIn | Offline import | Stage value | Yes (lead gen) | crm_id | n/a | ``` ## Consent Mode signals | Signal | Controls | Default before choice (consent-required markets) | |---|---|---| | ad_storage | Advertising cookies | denied | | analytics_storage | Analytics cookies | denied | | ad_user_data | Sending user data to Google for advertising | denied | | ad_personalization | Personalised advertising (remarketing) | denied | Basic mode loads Google tags only after consent; advanced mode sends cookieless pings before consent so Google can model conversions. Choose with your legal team. ## Deduplicating browser and server events 1. Generate one event ID per action (for example the order ID for a purchase). 2. Send the same event ID with the browser pixel event and the server event. 3. Use the same event name on both. 4. Check the platform's deduplication diagnostics (Meta Events Manager shows deduplicated events). 5. Monitor event match quality and improve it with more hashed identifiers (email, phone) where consented. ## Lead value ladder example | Stage | Example value | Sent to | |---|---|---| | Form submitted | Low | All platforms (secondary) | | Qualified lead (SQL) | Medium | Google, Meta, LinkedIn (primary for bidding) | | Opportunity created | Higher | Google, LinkedIn | | Closed won | Actual contract value | Google, Microsoft, LinkedIn | Calibrate values from historical conversion rates between stages and average deal value. ## Release checklist (run after every site change) - Test purchase or lead in each key template. - Confirm one event per action in each platform's debug tool. - Check consent banner still blocks and releases tags correctly. - Compare the next 48 hours of conversions with orders or leads in your own system. - Check that values and currencies are passed. ## Edge cases and judgement calls | Situation | What to do | |---|---| | Purchase happens on a third-party checkout | Use server-side events from the checkout platform or post-purchase webhooks | | Phone sales | Call tracking and offline imports with call IDs | | App plus web | Align events across Firebase or MMPs and web tags; avoid double counting | | Low volume of final conversions | Optimise on a well-correlated earlier event (qualified lead) with care | | Strict consent markets with low opt-in | Advanced Consent Mode and modelled conversions where lawful; set expectations | ## Handoffs to related skills - Overall account review: `paid-media-audit` - First-party audiences: `audience-first-party-data` - Reporting: `paid-media-reporting` - Channel bidding: `google-search-ads`, `performance-max-shopping`, `meta-ads` ## Worked example A lead-generation firm in financial services optimised Google and Meta for form fills. Sales said most leads were unqualified. Tracking showed Google Ads counting form fills twice (two tags), Meta pixel only (no CAPI, event match quality poor) and no offline imports. The fix: one form conversion per platform, Meta CAPI with deduplication, Consent Mode v2 with advanced mode for UK users, and weekly offline imports of qualified and closed outcomes with stage values. Bidding switched to qualified-lead value. Form volume fell, but qualified leads and closed revenue per pound rose within two months. ## Output format ``` # Tracking review: <business>, <date> ## Conversion inventory and issues by platform ## Consent Mode status by market ## Server-side and deduplication status ## Values and offline imports ## Tracking plan (table) ## Test results and monitoring setup ``` ## Common mistakes - Two primary conversions for the same action. - Server-side events without deduplication. - Consent defaults set to granted before the banner. - Optimising lead campaigns on raw form fills. - No monitoring after site releases. ## Questions to ask before you recommend changes - Which conversions drive bidding today on each platform? - Which markets require consent, and what does the CMP send? - Where is the source of truth for orders and leads? ## Self-check before you answer - Is each conversion counted once per platform? - Is Consent Mode correct for consent-required markets? - Are server events deduplicated and consented? - Are values and offline outcomes passed for bidding? ## Official documentation - Google consent mode: https://support.google.com/google-ads/answer/10000067 - Google enhanced conversions: https://support.google.com/google-ads/answer/9888656 - Meta Conversions API: https://developers.facebook.com/docs/marketing-api/conversions-api - TikTok Events API: https://business-api.tiktok.com/portal/docs - LinkedIn Conversions API: https://learn.microsoft.com/en-us/linkedin/marketing/conversions/conversions-overview --- Created by Afeez Adebayo, Marketing, SEO & Growth Expert Website: https://afeezadebayo.com LinkedIn: https://www.linkedin.com/in/afeez-adebayo/ X: https://x.com/za_icon Source: https://afeezadebayo.com/skills-md/marketing/performance-marketing/conversion-tracking-capi/ Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). You may copy, share and adapt this skill, including for commercial use, as long as you credit Afeez Adebayo and link to the source URL above. Keep this notice in any copy or adapted version.
Workflow
Step 1. Inventory and map
For each platform: conversion actions, source (tag, GA4 import, API, offline import), primary or secondary, value setting, counting (one or every).
Step 2. Remove duplicates
One primary purchase conversion per platform. If both a native tag and a GA4 import exist, choose one for bidding and set the other as secondary.
Step 3. Set up Consent Mode
Default consent states set to denied before the banner where law requires; update on choice; all four signals (ad_storage, analytics_storage, ad_user_data, ad_personalization). Use advanced mode where your legal basis allows, so Google can model conversions for consent-denied users.
Step 4. Add server-side events
Meta Conversions API, TikTok Events API, LinkedIn Conversions API and Google enhanced conversions, sending hashed first-party data with consent. Deduplicate browser and server events with a shared event ID.
Step 5. Pass values
Revenue for every purchase; consider profit-based values or new-customer values. For leads, assign values by stage and import offline outcomes.
Step 6. Import offline conversions
For lead businesses, send qualified lead, opportunity and closed-won events back to Google, Meta, LinkedIn and Microsoft with click IDs or hashed contact data.
Step 7. Test
Tag Assistant, GA4 DebugView, Meta Events Manager test events and event match quality, platform diagnostics. Make a test purchase or lead and trace it through every system.
Step 8. Monitor
Daily conversion counts by platform compared with orders or leads in your own system; alerts for drops or spikes.
Tracking plan template
| Event | Trigger | Platforms | Method (tag/server/import) | Value | Primary? | Dedup key | Consent required |
| purchase | Order confirmed | Google, Meta, TikTok | Tag + server | Revenue (or profit) | Yes | order_id | Yes |
| qualified_lead | CRM stage = SQL | Google, Meta, LinkedIn | Offline import | Stage value | Yes (lead gen) | crm_id | n/a |Consent Mode signals
| Signal | Controls | Default before choice (consent-required markets) |
|---|---|---|
| ad_storage | Advertising cookies | denied |
| analytics_storage | Analytics cookies | denied |
| ad_user_data | Sending user data to Google for advertising | denied |
| ad_personalization | Personalised advertising (remarketing) | denied |
Basic mode loads Google tags only after consent; advanced mode sends cookieless pings before consent so Google can model conversions. Choose with your legal team.
Deduplicating browser and server events
- Generate one event ID per action (for example the order ID for a purchase).
- Send the same event ID with the browser pixel event and the server event.
- Use the same event name on both.
- Check the platform's deduplication diagnostics (Meta Events Manager shows deduplicated events).
- Monitor event match quality and improve it with more hashed identifiers (email, phone) where consented.
Lead value ladder example
| Stage | Example value | Sent to |
|---|---|---|
| Form submitted | Low | All platforms (secondary) |
| Qualified lead (SQL) | Medium | Google, Meta, LinkedIn (primary for bidding) |
| Opportunity created | Higher | Google, LinkedIn |
| Closed won | Actual contract value | Google, Microsoft, LinkedIn |
Calibrate values from historical conversion rates between stages and average deal value.
Release checklist (run after every site change)
- Test purchase or lead in each key template.
- Confirm one event per action in each platform's debug tool.
- Check consent banner still blocks and releases tags correctly.
- Compare the next 48 hours of conversions with orders or leads in your own system.
- Check that values and currencies are passed.
Edge cases and judgement calls
| Situation | What to do |
|---|---|
| Purchase happens on a third-party checkout | Use server-side events from the checkout platform or post-purchase webhooks |
| Phone sales | Call tracking and offline imports with call IDs |
| App plus web | Align events across Firebase or MMPs and web tags; avoid double counting |
| Low volume of final conversions | Optimise on a well-correlated earlier event (qualified lead) with care |
| Strict consent markets with low opt-in | Advanced Consent Mode and modelled conversions where lawful; set expectations |
Handoffs to related skills
- Overall account review:
paid-media-audit - First-party audiences:
audience-first-party-data - Reporting:
paid-media-reporting - Channel bidding:
google-search-ads,performance-max-shopping,meta-ads
Worked example
A lead-generation firm in financial services optimised Google and Meta for form fills. Sales said most leads were unqualified. Tracking showed Google Ads counting form fills twice (two tags), Meta pixel only (no CAPI, event match quality poor) and no offline imports. The fix: one form conversion per platform, Meta CAPI with deduplication, Consent Mode v2 with advanced mode for UK users, and weekly offline imports of qualified and closed outcomes with stage values. Bidding switched to qualified-lead value. Form volume fell, but qualified leads and closed revenue per pound rose within two months.
Output format
# Tracking review: <business>, <date>
## Conversion inventory and issues by platform
## Consent Mode status by market
## Server-side and deduplication status
## Values and offline imports
## Tracking plan (table)
## Test results and monitoring setupCommon mistakes
- Two primary conversions for the same action.
- Server-side events without deduplication.
- Consent defaults set to granted before the banner.
- Optimising lead campaigns on raw form fills.
- No monitoring after site releases.
Questions to ask before you recommend changes
- Which conversions drive bidding today on each platform?
- Which markets require consent, and what does the CMP send?
- Where is the source of truth for orders and leads?
Self-check before you answer
- Is each conversion counted once per platform?
- Is Consent Mode correct for consent-required markets?
- Are server events deduplicated and consented?
- Are values and offline outcomes passed for bidding?
Official documentation
- Google consent mode: https://support.google.com/google-ads/answer/10000067
- Google enhanced conversions: https://support.google.com/google-ads/answer/9888656
- Meta Conversions API: https://developers.facebook.com/docs/marketing-api/conversions-api
- TikTok Events API: https://business-api.tiktok.com/portal/docs
- LinkedIn Conversions API: https://learn.microsoft.com/en-us/linkedin/marketing/conversions/conversions-overview
On this page
- When to use this skill, and when not to
- Example prompts
- Example output
- What the agent needs from you
- How this skill hands off to related skills
- How to run Conversion Tracking and Conversions API step by step
- Situations this skill handles
- What makes this skill different
- Practical tips from real projects
- Frequently asked questions
- Changelog
When to use this skill, and when not to
Use it when platform results look too good or too bad, after site or checkout changes, when launching in consent-required markets, when platforms show poor event match quality and for lead businesses that need offline outcomes.
Every automated bidding system learns from the conversions you send it. Duplicates make it overspend, missing consent-denied signals make it underbid, and leads that never become customers teach it to find more of them. Good tracking in 2026 means consent-aware tags, server-side events, deduplication, offline outcomes and values that reflect profit.
Don’t use it for attribution debates or allocation (use Paid Media Reporting or Incrementality Testing).
Example prompts
- “Google Ads shows twice as many purchases as Shopify. Fix our tracking.”
- “Set up Consent Mode v2 for our UK and EU traffic.”
- “Add the Meta Conversions API with deduplication.”
- “Import qualified leads and closed deals back into Google and LinkedIn.”
- “Write a tracking plan for our new checkout.”
Example output
The skill returns a fixed structure. A shortened example for an imaginary site:
# Tracking review: example financial services lead gen, 10 October 2026 ## Issues | Platform | Issue | Fix | | Google Ads | Form fill counted twice (two tags) | One primary conversion | | Meta | Pixel only, poor event match quality | CAPI with event_id dedup, hashed email and phone | | All | Bidding on form fills | Offline imports of SQL and closed-won with stage values | | UK traffic | Consent defaults granted before banner | Defaults denied; advanced mode where lawful |
Server-side tracking never replaces consent; the skill keeps legal basis and consent states at the centre of every setup.
What the agent needs from you
The skill asks for the conversion actions per platform and which are primary, tag setup (Tag Manager, gtag, pixels, server container), your consent management platform and markets, your commerce or CRM platform and how orders or leads are recorded, and recent site changes or known discrepancies.
How this skill hands off to related skills
Tracking underpins every other skill. Account review comes from the Paid Media Audit, audiences from First-Party Audiences, reporting from Paid Media Reporting, and bidding on the values you send from Google Search, Performance Max and Meta.
How to run Conversion Tracking and Conversions API step by step
- Install the skill as
conversion-tracking-capi/SKILL.md. - List conversion actions per platform, with source, value and primary status.
- Share your tag and consent setup and the markets you serve.
- Name your source of truth for orders and leads.
- Review the tracking plan: deduplication, Consent Mode, server events, values and imports.
- Test end to end and set daily monitoring.
Situations this skill handles
Tracking breaks in predictable ways. The skill handles these situations:
| Situation | What the skill does |
|---|---|
| Duplicate purchases in Google Ads | Keeps one primary conversion and demotes the other |
| Poor Meta event match quality | Adds CAPI with hashed identifiers and deduplication |
| EU and UK traffic | Configures Consent Mode v2 defaults and updates |
| Third-party checkout | Uses server events or webhooks from the checkout platform |
| Phone sales | Adds call tracking and offline imports |
| Low final conversion volume | Optimises on a well-correlated earlier event with care |
What makes this skill different
Most tracking guides cover installing a pixel. This skill designs the whole signal chain: which conversions drive bidding, how consent is honoured, how browser and server events are deduplicated, what values tell bidding about profit and how offline outcomes teach platforms which leads matter.
It finishes with testing and monitoring, because the most common tracking failure is a site release that quietly breaks a tag for weeks.
Practical tips from real projects
- One primary conversion per action, per platform.
- Defaults denied before the banner where consent is required.
- Shared event IDs for browser and server events.
- Values beat counts for bidding.
A worked example: a financial services lead-generation firm optimised Google and Meta for form fills, and sales said most leads were unqualified. Google Ads counted form fills twice, Meta ran pixel-only with poor match quality and there were no offline imports. One conversion per platform, Meta CAPI with deduplication, Consent Mode v2 with advanced mode for UK users and weekly offline imports of qualified and closed outcomes with stage values let bidding switch to qualified-lead value. Form volume fell, but qualified leads and closed revenue per pound rose within two months.
Recent changes shape the setup. Consent Mode v2 has been required for Google Ads features with EEA and UK users since March 2024. Google discontinued most Privacy Sandbox APIs in October 2025 and kept third-party cookies in Chrome under existing user choice, though Safari and Firefox still restrict tracking. Server-side APIs from Meta, TikTok, LinkedIn and Google are now standard practice for recovering lost signal with consent.
Lead businesses benefit most from a value ladder: low values for form submissions as secondary goals, higher values for qualified leads and opportunities, and real contract values for closed deals, so bidding learns the difference between a curious visitor and a buyer.
The release checklist keeps tracking healthy: a test purchase or lead on each key template, one event per action in each debug tool, consent behaviour checked and 48 hours of conversions compared with your own records.
Basic and advanced Consent Mode differ in one important way: basic mode loads Google tags only after consent, while advanced mode sends cookieless pings beforehand so Google can model conversions. Choose with your legal team.
Frequently asked questions
Why do ad platforms show different numbers?
They use different attribution windows, count view-throughs differently, model consent-denied conversions and sometimes count duplicates. Fixing duplicates and labelling windows explains most gaps.
What is Consent Mode v2?
Google’s system for adjusting tags to users’ consent choices with four signals: ad_storage, analytics_storage, ad_user_data and ad_personalization. It is required for Google Ads features with EEA and UK users.
Does server-side tracking avoid consent rules?
No. Server-side events recover signal lost to browsers and ad blockers, but you still need a legal basis and consent where the law requires it.
How do I stop double counting?
Keep one primary conversion per action on each platform, and use a shared event ID to deduplicate browser and server events.
What are offline conversion imports?
Sending outcomes such as qualified leads, opportunities and closed deals from your CRM back to ad platforms, so bidding learns which leads matter.
How often should tracking be tested?
After every site release that touches pages, checkout or forms, plus daily monitoring comparing platform conversions with your own orders or leads.
Changelog
- 1.0.0, 10 October 2026: first release.
Use, adapt and share this skill under CC BY 4.0: credit Afeez Adebayo and link to this page.
All Performance Marketing skills
- Paid Media Audit
- Media Planning and Budget Allocation
- Incrementality Testing: Lift and Geo Experiments
- Marketing Mix Modelling
- Google Search Ads and AI Max
- Performance Max and Shopping
- Meta Ads (Facebook and Instagram)
- TikTok, LinkedIn and Other Paid Social
- Ad Creative Strategy and Testing
- Landing Pages and Conversion Rate Optimisation for Paid Traffic
- Audiences and First-Party Data for Paid Media
- Paid Media Reporting and Attribution
