Skip to main content
Table of contents

Criteo OneTag technical notes

The Criteo OneTag includes a set of safety and quality mechanisms to ensure that there will not be a negative impact on your visitors’ browsing behavior and that the visitor’s privacy is protected.


User experience and performance

The Criteo OneTag doesn't alter your site visually and is completely invisible to your customers. What's more, the Criteo OneTag is asynchronous, so it won't affect your site loading time.

Device / bandwidth detection

The Criteo OneTag considers the current device of the user and the latency and bandwidth of the user's connection. Depending on that, the thresholds and the potential number of requests will be more strict.

Loading time thresholds

Time limitations are put in place to make sure the entire loading time does not exceed a certain threshold. All requests triggered by Criteo will immediately stop once the threshold has been reached.

Cached resources

The Criteo loader is a JavaScript library that will be cached by the user's browsers. This happens automatically as soon as the user interacts with the Criteo retargeting services on any website globally.
 

Compressed calls via HTTP-GZIP

All calls made to Criteo are HTTP-GZIP compressed to ensure a quick interaction between the user and Criteo’s data centers.
 

OnLoad-Event

The initial tracking request will be made as soon as the Criteo request has been fulfilled. All possible further requests triggered by the OneTag will only be made after the browser has successfully downloaded all primary resources including images, script files, CSS files, etc. from your website.

Example Criteo Calls: Inventory selection and Publisher calls are made after LOAD (red line)


Encapsulated calls

All potential publisher calls are triggered as an image request within an iFrame. This ensures that 3rd parties cannot place or read data on your website.

 

Smart inventory selection

We identify the most relevant publisher networks for the current users and also keep records of which publishers have already been informed of the current user. This way we make sure there are no redundant requests.

A user who interacted with Criteo in the past will see much fewer publisher tagging calls.

New users will see the maximum number of calls. Most monitoring and performance systems fall into the ‘new user’ category and because of that don’t reflect a realistic user browsing experience.

 

Publisher service monitoring

All publisher services are tested periodically and monitored 24/7 by Criteo around the globe. Loading times, response types, certificates, etc. are being monitored and checked constantly. Publishers are automatically removed from the inventory selection if tests are not passed completely.

 

Loading and uptime of data centers


Loader

This request (down)loads a file hosted in our data centers, there is no processing time between the request and the response. We have five data centers worldwide with exceptional availability (uptime).

 

Event

This request will be processed by our servers before they respond but the request is asynchronous. This means that the content of your website will continue to load as the browser waits for the response. Regardless, our average response time is only 25 milliseconds and our infrastructure team monitors latency 24/7 so that incidents are resolved within minutes.

 

Inventory selection

Much like the Event, this request is asynchronous, and our average response time is only 50 milliseconds. Our servers’ latency is also monitored 24/7. Most importantly, we only begin to tag users for publishers once the onLoad event has been reached and the page is fully usable for your visitors. This means that our tags will not affect the perceived load time of pages on your site.

 

Can certain publisher calls be removed?


It is not possible to remove specific publisher calls, except if there are technical issues. This case will result in a global deactivation as explained above. In case of issues related to that please get in touch with your Criteo contact. 

Criteo OneTag HTTP/HTTPS calls explained 

The Criteo OneTag allows us to know who to retarget and what to recommend to them across our network of publishers. A smart algorithm evaluates where the user can be approached in the most efficient way. As a result, 3rd-party HTTP calls may be made. No personal or advertiser-related information is shared with other vendors.

 

What calls are triggered by the OneTag?

When our tag fires on one of your pages, it results in a call to our loader and three main HTTP requests: a request from the user's browser to our servers and HTTP requests to our publisher partner’s ad servers. All requests are asynchronous.
 

  1. The initial call downloads our main JavaScript library that functions as a framework and prepares all of the following requests

  2. The second call is the actual tracking event. This event sends the user-browsing information to our servers such as which products are currently being looked at

  3. The third call request is for inventory selection. This call builds a list of all relevant publishers from our server

  4. Depending on the result of the inventory selection call, publisher http-calls might be made. This is when publishers are informed that we would like to show ads to the current user whenever they browse the publisher’s network. These requests are made as image requests and are therefore also called 3rd-party pixels

 

FLoC IDsProposed by Google as an audience targeting scenario for the Chrome web browser, FLoC, or “Federated Learning of Cohorts” creates lookalike audiences based on machine learning. Google began testing FLoC as a replacement for third-party cookies in March 2021. 

Criteo will be collecting FLoC IDs on a portion of the traffic in eligible countries (Australia, Brazil, Canada, India, Indonesia, Japan, Mexico, New Zealand, Philippines, and the U.S.) during the trial period set by Google.

For more information on FLoC, visit https://web.dev/floc/.