User Tracking Methods
Understanding users’ behavior right from the time he engages with an ad content is crucial from an advertiser’s standpoint. Advertisers are keen on capturing maximum data points they can from an ad click (even ad view) till conversion. Choosing the right tracking method is important to get a strong hold of the information that one needs. With growing business and challenges in the Ad-Tech industry, various methods for tracking the users are evolving. Let’s see them in detail and understand the pros and cons of each.
Client to server
It is a cookie based tracking method. ‘Client’ here is the user’s browser. When a user clicks on an ad, a cookie is dropped on his or her browser. Tracking pixel(s) can be placed on the final conversion page or on multiple pages of the site to which user is redirected after clicking on an ad. These pixels gets fired and conversion data along with cookie information is sent to the server (which had initially dropped the cookie).This not only helps in unique identification of the user, as each user is assigned a unique cookie, but also helps in collecting other valuable information about the conversion.
Along with cookie, which a server can read from the users’ browser the server also gets information about user agent and IP address.
It is easier to do tracking and attribution using this method as information is readily available and heat mapping of users’ browser can be easily done. It requires fewer technical resources to implement this method of tracking, as only tracking pixel needs to be placed at the desired page. All the information is stored on the client side, so it also saves on server costs.
If the client has disabled cookies this method fails, as there is nothing you can store on the client browser. Tracking is lost once the user clears the cookies on his browser. In short, your previously identified user becomes a new user once again. It is also difficult to track clients having Ad blockers as they interrupt the tracking path. Sometime because of some unknown issues, this tracking method doesn’t work at all. Modern browsers like UC Web, Opera etc. face more challenges in terms of tracking and attribution when this tracking method is used.
Server to Server
Server to server is also known as server-based or cookie-less tracking. It works by placing all the user and tracking related information on advertisers’ server (2) whenever the user clicks an ad or views an ad impression. The information is stored on server (2) along with some unique identifier for every click or impression. These identifiers are generally click id, session id etc. The unique identifier assigned is passed back to the click/impression generating server (1) when the conversion is notified via a ‘Postback URL’. This not only identifies the converting user, but also helps in collecting other valuable information related to conversion.
Does not require cookie to be dropped, so no issues related to modern browsers or cookie being cleared by the user. Data of each user is present on a server which helps is understanding user behavior progressively. It’s more safe and trusted way to track.
Requires more technical resources and expertise to implement the system in place. The client, being that it’s the browser has easy access to user specific attributes, such as cookies, IP address, user agent and referrer. While on the other hand, collecting this information using s2s requires effort on the client side.
In s2s, especially for mobile app, user tracking can be done in following ways
Device Identifiers based tracking
Device identifiers are unique identifiers assigned to each device by their OS .For iOS devices, it is termed as IDFA (Identifier for Advertisers),for Android devices, it is GAID(Google Advertising Identifier) and for windows it’s windows AID .This is mainly used for re-marketing, re-targeting and behavior-based advertising. These identifiers are accessible only through a mobile app and not m-website. A user can be clicking on ads through multiple apps installed on their devices, but the device identifiers (IDFA or GAID) remains same. This helps in understanding the likes and behavior of each user uniquely. User profiles can be created using the unique device identifiers and the data generated through engagement with Ads or an advertisers’ app. In the long run, a lot of learning is available for targeting the users as per the needs of advertisers.
Device identifiers are global and is not specific to a server as in case of cookies. Cookie dropped by a certain domain can’t be accessed by others. Very few users reset their identifiers so tracking based on these identifiers is more accurate.
You do not have access to this identifier through mobile browser. Users have the freedom to reset IDFA or GAID of their device which means, if a user changes his identifier periodically then the system will lose his past behavioral data and treat him as a new user.
Whenever a browser makes a request to the web server, lot of information is passed to the server which can be used to identify the device uniquely. This information is generally IP address, browser details, screen size, screen resolution, device OS, font style, installed plugins and many other ’n’ things (If you want to have a look of what all is being passed through your browser on making a web request you can visit noc.to) .Since these ’n’ things together make a unique combination for each device, the advertisers assign a unique identifier called as ‘fingerprint’ to each combination on their servers. Each of these fingerprints is unique to each device and thus tracks each user uniquely.
Fingerprint match (as we do in real life for crime detection) helps in identifying fraudsters, especially on e-commerce sites. No privacy concern (till date) for the user as the information which is used for fingerprinting is not treated as personal. Cross-device tracking is possible using this method with 0.6–0.8 probabilistic score.
A single device may have multiple web clients installed, or even multiple virtual operating systems. As each distinct client and OS has distinct internal parameters, one may change the device fingerprint by simply running a different browser on the same machine. Since this method relies on the availability of client-side scripting language to gather information about various parameters, someone using privacy software cannot be identified using fingerprinting. Not highly accurate because of probabilistic model used.
So next time you click an ad you know what all is going backstage!