Meta Ads Audit · 11 min read · Published May 26, 2026
Conversions API Setup: Why Your Pixel Alone Is Lying To You
Most agencies tell you they "set up your CAPI" as if installation is the deliverable. Installation is not the deliverable. Properly deduplicated Pixel and Conversions API events with Event Match Quality above 7.0 is. Here is the 5-step verification.
Founder, BTB Audits. $150M+ in ad spend managed across Meta and Google
Most agencies will tell you they "set up your Conversions API" as if installation is the deliverable. It is not. A properly deduplicated Pixel and Conversions API (CAPI, Meta's server-to-server way of sending sales events) setup with Event Match Quality (EMQ, Meta's 1 to 10 score for how well your data matches a real user) above 7.0 is the deliverable. The difference is roughly 30 percent of the conversion signal that should be reaching Meta but is not.
In about 9 of 10 accounts I audit at $20K or more in monthly Meta spend, CAPI is technically live but functionally broken. Events fire twice. Event IDs do not match. Customer data is hashed wrong or missing. EMQ scores sit below 5.0. The agency told the founder "CAPI is set up." Reporting is inflated. The cost gets blamed on iOS 14. The actual cost is the broken setup. The patterns repeat across $150M+ in managed ad spend.
At the Meta Performance Marketing Summit 2026, Meta reported that accounts running both the Pixel and the Conversions API saw 17.8 percent lower cost per result than accounts running the Pixel alone. The number is real. The path to claiming it is not the one-click installer. It is deduplication and signal quality. The 5 sections below walk through what to verify and what to fix.
What CAPI does (and what the Pixel cannot)
The Conversions API is server-to-server data transmission from your website to Meta. The Pixel is browser-side data transmission via JavaScript (the small piece of code that runs in the user's browser). The structural difference matters.
- Pixel runs in the user's browser. It is blocked by iOS 14.5+ ATT (App Tracking Transparency) opt-outs, ad blockers, privacy browsers like Brave and Firefox strict mode, and any client-side issue. Slow page loads, JavaScript errors, and third-party script conflicts all break it.
- CAPI runs from your server to Meta's server. Nothing the browser does can block it. The data sends regardless of what the user's browser allows.
The two are not interchangeable. Each captures something the other misses.
- CAPI alone misses browser-side context: user agent, IP (Internet Protocol) address, browser-set cookies that the Pixel captures.
- Pixel alone misses everything CAPI captures from blocked or restricted browsers.
- Together, with proper deduplication, they cover the full conversion signal.
How deduplication works. Both the Pixel and CAPI send the same event (for example, a Purchase) with the same event_id parameter. Meta sees both signals, matches the event IDs, and counts the conversion once. Without matching event IDs, Meta sees two separate conversions and double-counts. The reported number goes up. The real revenue does not.
This is the upstream version of Stage 2 connections and event firing in the Meta audit method. The audit cannot move past this stage if the signal layer is broken.
Why iOS 14.5+ made CAPI non-optional
Some context. In April 2021, Apple's App Tracking Transparency framework required iOS apps, including Facebook and Safari on iOS, to ask users for permission to track them across other apps and websites. About 3 in 4 users said no. The signal Meta had been working with for years collapsed in a single quarter.
The impact on Meta advertisers showed up fast.
- Pixel events from iOS Safari users dropped 30 to 50 percent overnight.
- Attribution windows on iOS users shrank from 28 days to 7 days.
- AEM (Aggregated Event Measurement, Apple's privacy-preserving replacement) took over for iOS conversions.
- Reporting accuracy on iOS-heavy audiences became structurally lower.
CAPI was Meta's answer. Server-side tracking does not depend on the user's browser or app permissions. The signal sends from your server regardless.
The honest version: CAPI does not restore iOS tracking to its pre-2021 state. It partially compensates by sending stronger signal on the conversions you can still see. Brands without CAPI in 2026 are running on 30 to 50 percent of the conversion signal they had in 2020. Brands with broken CAPI are running on 60 to 70 percent. Properly configured CAPI gets back to roughly 85 to 90 percent.
That last gap is the difference between a setup that is installed and a setup that works.
| Aspect | Pixel | Conversions API |
|---|---|---|
| Where it runs | User's browser | Your server |
| What it captures | Browser context (user agent, IP, cookies, click ID) | Server events with full customer parameters |
| What blocks it | iOS ATT, ad blockers, privacy browsers, JS errors | Effectively nothing browser-side can block |
| Setup complexity | One-line script in the site head | Server integration or tag manager |
| EMQ ceiling | Limited by browser-available data | Up to 9.0+ with full parameter coverage |
| Required in 2026? | Yes (covers what CAPI cannot) | Yes (covers what Pixel cannot) |
Event Match Quality: the metric most operators do not audit
Event Match Quality is Meta's 1.0 to 10.0 score for how well the customer data you send with each event can be matched to a Meta user. Higher EMQ means Meta can tie the conversion to a specific person. That means better optimization, better lookalikes, and more accurate reporting.
The customer parameters that move EMQ:
- Email (hashed and normalized)
- Phone (hashed, country-code formatted)
- First name, last name (hashed, lowercase)
- Date of birth (hashed)
- City, state, country, zip (hashed where required)
- Client IP address (raw)
- Client user agent (raw)
- External ID (your own customer ID, hashed)
- Facebook Click ID (fbc) and Facebook Browser ID (fbp)
The thresholds that matter:
- Below 5.0: poor match quality. Optimization suffers a lot.
- 5.0 to 7.0: acceptable. Works but not well.
- 7.0 to 8.5: good. The target for most operators.
- Above 8.5: excellent. Requires sending most or all of the customer parameters.
The common misread: operators assume installing CAPI gives them good EMQ by default. The reality: out-of-the-box Shopify CAPI typically lands at 4.0 to 6.0. Lifting to 7.0 or higher takes explicit work to send the extra customer parameters. The technical layer underneath every Meta attribution gap is here. Why platform-reported numbers diverge from Shopify is the operator-level version of this problem. EMQ is one of the reasons Meta's number drifts from your back-end.
The 5-step setup verification
After CAPI is installed by your dev team, your Shopify integration, or a tag manager, run these 5 checks before you call the setup done.
Step 1: Both Pixel and CAPI are firing. Open Meta Events Manager. On the event source page, each major event row should show events from both "Browser" (the Pixel) and "Server" (CAPI). If only one source is firing, deduplication cannot happen because there is nothing to deduplicate against.
Step 2: Event IDs are matching. Open the Test Events tab. Fire a test purchase from the live site. The Pixel event and the CAPI event for that purchase must share the same event_id parameter. Different event IDs mean Meta will count the same event twice.
Step 3: Deduplication is above 90 percent. In Events Manager, each event row shows a deduplication rate. Below 90 percent means a meaningful share of events are still being counted twice. The closer to 100, the cleaner the setup.
Step 4: Event Match Quality is above 7.0. Read the EMQ score for Purchase, AddToCart, and ViewContent. Below 7.0 means the customer data being sent is too thin. Add email, phone, name, and external ID to the CAPI payload to lift it.
Step 5: The numbers match Shopify within 5 percent. Pull the last 7 days of Purchase events in Events Manager. Compare to the last 7 days of completed orders in Shopify. After deduplication, the gap should be within 5 percent. If Meta shows 15 to 30 percent more than Shopify, deduplication is broken.
A setup that passes all 5 steps is working. A setup that fails 2 or more is technically live but practically broken. That is the state of most accounts at $20K and above in monthly Meta spend that have not been audited.
A worked example. Anika ran a DTC (direct-to-consumer) supplements brand at $80K per month on Meta. CAPI was installed through the native Shopify integration. EMQ for Purchase sat at 4.2, well below the 7.0 threshold. Her dev team added customer email and phone to the CAPI payload through a server-side tag manager. EMQ lifted to 7.8 inside 14 days. Meta's reported CPA (cost per action) dropped 18 percent over the next 30 days. The optimization algorithm had usable signal for the first time in a year. The patterns repeat across $150M+ in managed ad spend.
Most accounts at this spend tier pass 2 or 3 of the 5. The other 2 or 3 quietly cost 15 to 25 percent of the reported conversions. This is the technical layer underneath the broken signal pattern in Leak 1 on broken signal foundation. The leak post names the problem at the account level. The 5 steps above name it at the event level.
Want this run on your account? Get a Free Quick Scan in 48 hours.
How to actually deploy this
Three deployment paths. Ranked by reliability.
Option 1: Native Shopify integration. Lowest friction. Mid-level quality. Shopify's built-in Facebook and Instagram channel sends server-side events. Setup is one click. EMQ typically lands at 5.0 to 6.0 because limited customer parameters are sent by default. Good enough for under $20K in monthly Meta spend. Not enough above it.
Option 2: Tag manager solution. Mid-level friction. Mid to high quality. A server-side tag manager handles the CAPI payload. Stape, Elevar, and GTM (Google Tag Manager) server-side container are common examples. Requires technical setup but no full custom build. EMQ can reach 7.0 to 8.0 with proper configuration. For most DTC brands at $20K to $100K in monthly Meta spend, this is the right balance of effort and quality.
Option 3: Direct API integration via your dev team. Highest friction. Highest quality. Your developers send events directly from your server to Meta's API. You control every parameter and how it is sent. EMQ can reach 8.5+ with full parameter coverage. Worth it above $100K in monthly Meta spend, where the per-conversion value justifies the engineering hours.
| Option | Friction | EMQ Ceiling | Best For |
|---|---|---|---|
| 1. Native Shopify integration | Low (one-click) | 5.0 to 6.0 | Under $20K monthly Meta spend |
| 2. Tag manager (Stape, Elevar, GTM server) | Mid | 7.0 to 8.0 | $20K to $100K monthly spend |
| 3. Direct API via your dev team | High | 8.5+ | Over $100K monthly spend |
The closing position is simple. The question is not "do I have CAPI." Most accounts at $20K and above answer yes. The question is "is my CAPI deduplicating cleanly and firing at EMQ above 7.0." Most of those same accounts answer no.
The fix is configuration, not new infrastructure. The 5-step verification names the gap in 60 minutes. The remediation usually takes 7 to 14 days of dev time. The lift shows up inside 30 days on the CPA line in Ads Manager. The ROAS (return on ad spend) implied by your gross margin gets easier to clear once the algorithm has usable signal. The gross-margin math behind a good ROAS works only when the inputs to the algorithm are clean.
Frequently asked questions
Common questions
About CAPI
Do I need the Conversions API if I have the Pixel?
Yes. The Pixel runs in the browser and is blocked by iOS ATT opt-outs, ad blockers, and privacy browsers. In 2026, that is 30 to 50 percent of your iOS traffic invisible to Meta. CAPI runs from your server and is not blocked by any of that. The two together, with matching event IDs for deduplication, cover the full conversion signal. The Pixel alone is leaking signal that the algorithm needs to optimize. CAPI alone misses browser context the Pixel can still capture. Both are required.
What is Event Match Quality and why does it matter?
Event Match Quality (EMQ) is Meta's 1.0 to 10.0 score for how well the customer data you send with each event can be matched to a Meta user. Higher EMQ means Meta can attribute the conversion to a specific person, which improves optimization, lookalikes, and reporting accuracy. The threshold most operators should target is 7.0 or above. Below 5.0, optimization suffers meaningfully. Out-of-the-box Shopify CAPI typically lands at 4.0 to 6.0 because limited customer parameters are sent. Lifting to 7.0 requires explicit work to send email, phone, name, and external ID.
How do I set up the Conversions API on Shopify?
Three options. The simplest is the native Shopify Facebook channel, which is one click and sends server-side events. EMQ usually lands at 5.0 to 6.0. Better is a tag manager like Stape, Elevar, or a server-side GTM container, which can reach EMQ 7.0 to 8.0 with proper configuration. The highest-quality path is a direct integration written by your dev team, which can reach EMQ 8.5 or higher. Pick the option that matches your monthly Meta spend. The native integration is fine below $20K. Above $20K, the extra signal quality is worth the additional setup work.
Why are my Pixel and CAPI conversions different?
Two common reasons. First, the event IDs do not match. If the Pixel and CAPI send the same event with different event_id parameters, Meta sees two separate conversions and counts them both. The reported number is inflated. Second, only one of the two is firing. If only the Pixel is sending an event, you are losing all the iOS and ad-blocker traffic. If only CAPI is sending, you are losing browser context that improves match quality. Open Events Manager and check the deduplication rate per event. Below 90 percent is the diagnostic that something is misaligned.
About BTB Audits
What does a CAPI audit actually cover?
Every audit I run reconciles the Pixel and CAPI configuration against the 5-step verification in this post. The Free Quick Scan flags the size of the gap from public signals. The paid Forensic Report opens Events Manager, runs the test events, reads the deduplication and EMQ scores per event, and writes the exact configuration changes needed. The report sizes the lift in dollars based on your monthly Meta spend. The remediation is your dev team's work, but the spec is in the report.
Will this work for me
Is iOS 14 still affecting my Meta ads in 2026?
Yes, and it has been for 5 years. About 3 in 4 iOS users still opt out of app tracking when prompted. Meta's view of iOS Safari traffic dropped 30 to 50 percent in 2021 and never fully recovered. CAPI partially compensates by sending signal from your server, which the iOS opt-out does not block. Without CAPI in 2026, you are running on 50 to 70 percent of the conversion signal you had in 2020. With broken CAPI, you are at 60 to 70 percent. With CAPI deduplicating cleanly at EMQ above 7.0, you can recover most of the gap. iOS 14 did not go away. The accounts that adapted just stopped feeling it.
Most CAPI installs at $20K+ in monthly Meta spend pass 2 or 3 of the 5 verification steps. The other 2 or 3 quietly cost 15 to 25 percent of reported conversions. Run the 5 checks on your account this week, or have a Free Quick Scan do it for you in 48 hours.
If you don't have four to six hours, or you want a second pair of eyes that's managed $150M+ across Meta and Google, the Free Quick Scan is what I built for that. I'll record a private 5 to 7 minute Loom walking through the leaks I find on your account using public data only. You'll have it in 48 hours.
Get Your Free Quick Scan →Keep reading on Meta signal quality and the audit method
Aditya Chaturvedi is the founder of BTB Audits. He has managed $150M+ in ad spend across Meta and Google for DTC, SaaS, and lead-gen brands. The CAPI deduplication patterns in this post come from auditing accounts at the $20K and above monthly Meta spend tier, where broken signal quietly costs 15 to 25 percent of reported conversions. Read more on the BTB Audits blog.