Do we ever wonder how clicks are recorded? Well, today I would like to talk about the standard click measurement methodology as proposed by Interactive Advertising Bureau & Media Rating Council. Here, I would consider click as one of the media buying currencies and focus this study as when the click happens, an actual revenue transaction takes place. Different click counting organizations such as search engine providers, publishers, networks, SSPs, & DSPs employ various enhancement technologies to existing click methodology, (to gain a competitive edge) for competitive differentiation. These amendments are valid as long as they are meeting the minimum acceptable conditions.
Fundamentally, there are four stages of click lifecycle after an ad impression happens. Note that, Fraudulent clicks may occur from non-human agents, automated plugin, robots & spiders etc. Advertisement Organizations must ensure that fraudulent clicks should be excluded from valid click stream so that click count contain only actual human initiated clicks. So, when user-initiated click event takes place in the advertisement, it causes an HTTP 302 redirect on the server side, which re-directs the user from publisher website to advertiser website or landing page. There is another type of click measurement method that does not redirect users to advertiser landing page or website, but that does cause an HTTP302 server redirect and thereby log clicks. These click on the ads are interactions with various HTML5 elements on the ads that advertiser wants to measure. Let's talk back about four stages of click referral cycle. These are Initiated Click, Measured Click, Received Click & Resolved Click.
Initiated Click- When an ad impression happens and a real user interacts with that ad, an initiated click event takes place.
Measured Click-While an initiated click happens, an HTTP GET request goes to the publisher server or third-party ad server from the user’s browser and server response it back responds with successful HTTP 302 redirect. At this moment, Publisher server or third party ad server may count click event for reporting purposes, provided the conditions to validate click against fraudulent clicks are met.
Received Click-After Measured click is logged by Advertiser web server, the user is now redirected to new click destination and a received click is registered.
Resolved Click-After received Click, the user will be redirected to the advertiser’s landing page, where advertiser server measure resolved clicks with HTTP 200 status of final destination.
Below diagram show the process step by step,
Ad Served by Publisher
Ad served by Third Party Ad Server
There can be a slight variance between a Measured click(Origin server) and Received Click(destination site), because of various reasons such as latency, user aborts, 404 errors etc. So, that means click measurements are not always equal to a website redirect measurements made on the landing page. Rule of the thumb is Prefiltered Measured Clicks should always be equal to/or higher than received clicks.
A click may be considered valid only when either of the two counting methods is used and two important conditions are also met:
One-Click-Per-Impression Method: This method represents measuring Click on Unique ad impression.
Multiple-Clicks-Per-Impression Method: In this method, a rule of repeat-click-refractory-period* is applied to prevent inappropriate click counting due to multiple clicks by users unintentionally due to certain reasons such as UI/navigational mistakes etc.
*Repeat-click-refractory-period is the time between the click and the previous click on the unique ad impressions. This time is defined by the advertiser or publisher server.
The two must conditions are impression-staleness-window & invalid click filtration as explained below,
a) Impression-staleness-window: This is the amount of time predefined by publisher web server or ad server. This rule is explained as the time between the ad impression and the click, which should be less than an impression staleness window.
b) After Initiated Click, when the measured click is happening and before the user is redirected to the landing page, a click must be subjected to Invalid click filtration steps as we will discuss in detail below.
Some other volumetric metric/conditions can also be established such as Maximum level of clicks per users on a creative within a fixed time frame or some click scoring methods like the estimated likelihood of clicks. This will help us to explain click stream anomaly on the basis of website and user behavior.
Click counting Organizations are also encouraged to maintain click transaction records in consistent manners and unique click identifiers (unique numeric or alphanumeric string associated with the click) should also be used. These things will help in auditing and investigation for invalid click activity later.
Invalid clicks can happen due to multiple reasons such as from a user agent bots, spiders, auto click functions, internal publisher clicks, internal users, proxy servers etc. They get access to URL and launch it in invisible miniature popups, new tabs or new windows without user permission. These invalid click should be removed from click count stream.
Organizations should also maintain a record table of suspected robots and compare the incoming clickstream data user agents/IP address against it to filter down invalid clicks.
Note that: There aren't any verification partners that can report on click fraud due to a couple of reasons,
-technical limitations &
-the MRC will not allow 3rd party verification vendors to audit clicks because they are NOT ad servers.