What are the elements to check in terms of tracking to limit statistical differences?
What are the counting differences between tracking technologies?
In digital marketing, an advertiser has three datasources:
- traffic vendors.
- measurement solutions.
- own back office.
It is therefore observed that the two main types of discrepancy are, on the one hand, those between what traffic sellers will charge and the measurement solutions, and, on the other hand, between the measurement solutions and the back offices.
Hence four concrete cases to exemplify the problems of the advertiser:
- train desk reporting has far more impressions than the measurement solution.
- the retargeting solution requires a lot more clicks than the allocation solution.
- the affiliate platform counts more transactions than the analytics solution.
- not all e-commerce back office sales are necessarily in the allocation solution.
What explains these discrepancies?
We can identify exogenous and endogenous sources as being responsible for such discrepancies. Among the first, there are browsers with different functionalities. From the endogenous point of view, differences are often due to the configuration of the analytics solution (deduplication, tagging on the advertiser’s site, cookie life, cross-device or counting method).
Impact of cookies on tracking
In 99% of cases, the cookie is still the basis of what is done in tracking (except for mobile applications). Cookies all have the same nature and become “first party” or “third party” only in a certain context. If the cookie is linked to the domain name visited by the user, it is located in the first part; otherwise, it is found in a third party.
Impact of ITP on tracking
In the case of the Safari browser, ITP 1.0 was released in 2017 to limit the duration of use of third-party cookies to 24 hours. A year later, on ITP 2.0, Safari will no longer grant any time for third-party cookies, which will have a huge impact on user tracking and advertising.
Then the browser will attack the first cookies, which, in turn, come from a domain other than the one visited by the user, limiting its use to 24 hours. Recently, ITP 2.3 included the limitation of storage space to 7 days.
Apple announced the same thing, unlike Chrome and Firefox. The latter blocks only the cookies of the advertising industry (disconnect list) and finger printing (recovery of information on a computer to try to make it a single key).
Impact of adblocks
Adblocks are an extension of the browser. Some are business oriented; others are free but block everything. All work through lists (like Easylist Privacy) that aggregate a set of third-party domains. We also find adblocks on mobile, native, and integrated with their own lists or even without lists. Today, these adblocks are very efficient.
How can induced problems be solved?
In order to counter them and streamline tracking, it is necessary to systematically identify where errors come from, which requires time and a little curiosity. This makes the entire marketing data chain more reliable.
To understand the discrepancies, the method is quite simple. The first step is to analyse the biases of each solution, then make a point-by-point comparison of each bias. Some combinations of bias on both sides lead to very large differences. In other cases, the two solutions will cohabit, and thus minimize the differences.
There are essentially three specific points of vigilance to monitor:
- Semantic Deviation Logic - there is a difference between regulated clicks and inbound clicks that are counted at two distinct times, which is important because of the latency of the site. The visit definition also plays, depending on whether a fixed duration or a reset duration during the session is assigned by the solution. Similarly, the definition of a print differs between the IAB (a second on 50% of the surface) and the solutions.
- tag management and CMP: the tag manager, installed on a site, conditions the call of other sites, which simplifies the integration of new tracking tags, and the CMP manages the user’s consent. In case these two technologies are blocked, the tracking is necessarily distorted.
- cross-device distortions: it may be observed that two conversion paths were made by the same user via two different browsers. These two paths must be reconciled otherwise there will be a poor distribution of sales according to solutions and devices, hence attribution discrepancies.
Here are some practical cases to consider.
Number of inconsistent visits
The difference between the number of visits counted by a traffic provider and an analytics solution must be evaluated according to the possible blocking of one or the other by an adblock or Firefox, then by checking the tagging. If the solution is not blocked and the traffic provider is whitelisted, there is no tracking gap of more than 5%.
Different CA between solutions and back office
We have to ask ourselves the same questions as before, but by adding the verification of the implementation of the solution on all the e-commerce sales niches and its possible integration in a blocked TMS.
Differences between affiliation platform and allocation solution
In addition to the usual questions, it is important to know whether all attribution models work properly. The distortions in tracking can then come from the same causes as before, but also from too specific attribution models.
Cross-device attribution (Facebook)
The gap between analytics and Facebook is increasingly being discussed. In addition to the rest, we must consider the fact that Facebook has a very powerful cross-device. Facebook also recovers a lot of sales initiated on mobile and completed on desktop.
Emmanuel Brunet concludes: “There are three priorities for advertisers: understanding the differences between solutions, understanding the specifics of each, and being able to follow technological advances.”
Noella BOULLAY, Directrice Déléguée - CPA
Emmanuel Brunet - CEO - Eulerian Technologies
Côme Filippi, Acquisition & Webanalytics Manager - MisterFly