The essentials
- Four methods: a CMS extension, a partner integration, a Tag Manager server container, direct development; from the simplest to the most flexible.
- Deduplication: each event carries a unique identifier sent identically by the pixel and by the server, without which conversions are counted twice.
- Quality score: Meta rates event matching from 0 to 10; aim for 6 as a minimum, 8 for an e-commerce site.
- Expected gain: 10 to 30% more attributed conversions, and a cost per result measured closer to reality.
Meta's conversions API sends events from the site's server rather than from the visitor's browser, which recovers part of the conversions lost to blockers, browser protections and cookie refusals. The general principle and the expected gains are described in CAPI: definition ; this article covers the practical implementation on an SME site, from choosing the integration method to the final check. The pixel is still needed: the two work together, as explained in the Meta pixel.
Choosing your integration method
| Mode | Difficulty | Cost | Who for |
|---|---|---|---|
| CMS extension (Shopify, WooCommerce, PrestaShop) | Low, a few clicks | Free to €30 per month | Most e-commerce sites |
| Partner integration (Zapier, Segment, email platforms) | Poor | €20 to €100 per month | Sites with no official extension |
| Google Tag Manager server container | Medium, technical configuration | €10 to €120 per month of hosting | Sites that want to centralise all tags server-side |
| Direct development via the API | High | 2 to 5 days of development | Bespoke sites, specific needs |
For an SME on a common CMS, the official extension is enough in the vast majority of cases and takes an hour to set up. The server container becomes worthwhile when several advertising platforms have to be fed in the same way.
The events and parameters to send
- The same events as the pixel: purchase, add to basket, checkout initiated, form, sign-up. An event present on one side and missing on the other creates a permanent gap.
- Value and currency: essential for optimising on ROAS; without them, the algorithm optimises for volume.
- The matching data: email and phone hashed in SHA-256, first name, surname, city, postcode, external identifier. The more there are, the better the match.
- The advertising identifiers: the parameter
fbcfrom the click and thefbpfrom the browser, to be passed to the server to link the event to the campaign. - The event identifier: the same value on the pixel side and on the server side, the key to deduplication.
- Consent: server-side sending is subject to it just like the pixel; it must be conditioned on the signals from your consent platform.
Deduplication, step by step
This is the point where most installations fail. The principle: generate a unique identifier per event, for example the order number, and send it in the event_id field from both the pixel and the server. Meta then recognises one single event. Three checks: the events manager shows a deduplication rate, which should approach 100%; the number of purchases reported should match the number of orders in the back office; and a sudden jump after activation signals a duplicate. Without deduplication, the ROAS shown doubles, bidding runs away and budget decisions become wrong.
The event match quality score
| Score | Interpretation | Action |
|---|---|---|
| 0 to 4 | Low matching, few events linked to an account | Add hashed email and phone, check the format |
| 5 to 6 | Acceptable for a services site | Add the click identifier and the postcode |
| 7 to 8 | Good level, the target for an e-commerce site | Maintain and monitor |
| 9 to 10 | Excellent, rare outside accounts with authentication | Nothing to do |
The score is read in the events manager in Meta Business. It measures Meta's ability to link an event to a person, which directly determines attribution and optimisation.
Checking after activation
- The live test: the test events tool shows the pixel's and the server's sends in real time, with their identifier.
- Volume: a rise of 10 to 30% in events received is expected; a rise of 80% signals a duplicate.
- The comparison with the back office: over two weeks, the number of purchases reported should come close to the number of actual orders.
- Cost per result: it falls mechanically since more conversions are attributed to the same spend; the benchmarks are in how much Meta Ads costs ; see the SMA KPIs.
- Conversion tracking: overall consistency is checked using the method described in conversion tracking in SMA.
The frequent mistakes
- Forgetting the event identifier: systematic duplicates, ROAS multiplied by two.
- Sending unhashed data: refused by Meta and non-compliant.
- Ignoring consent: server-side sending is still subject to the GDPR.
- Not sending the value: no way to optimise on ROAS.
- Install and forget: a site update frequently breaks the sending; the quality score should be monitored every month.
How GreenRed helps
Rather than juggling several tools, GreenRed's SMO and SMA social media tracking brings these metrics together in a single dashboard, compares them over time and tells you which actions come first. You can try it free, with no card, from the Pricing.
Frequently asked questions
How do you install the Meta conversions API?
Through an official CMS extension for most e-commerce sites, through a partner integration, through a Google Tag Manager server container to centralise tags, or through direct development via the API. For an SME on a common CMS, the extension is enough and takes an hour to set up.
Should you keep the Meta pixel?
Yes. The pixel captures the browser signals, the server fills in what is missing. The condition is deduplication: a unique identifier per event, sent identically by both channels, failing which each conversion is counted twice.
What is the event match quality score?
A score from 0 to 10 given by Meta that measures its ability to link an event to a person, based on the data sent (hashed email and phone, click identifier, postcode). Aim for 6 as a minimum, 8 for an e-commerce site.
How do you check that the installation is correct?
Through the test events tool, which shows sends in real time, through the deduplication rate, which should approach 100%, and above all through the comparison between the number of purchases reported and the number of actual orders in the back office over two weeks.