Article-Multichannel

The first article of this two-part series highlighted two extremely important points:

  • the issue is more complex than one might think
  • we must beware of reductive and categorical arguments, of which analysts are inherently wary

This naturally leads us to multi-channel tracking, where we’ll see that different companies’ offerings hide some significant constraints.

Managing and analysing “multi-channel” performances: they don’t tell you everything

Building on the assumption that the multi-device experience (desktop, tablet, smartphone, smart TV, etc.) is undeniable, thanks to the multi-channel brand interactions it offers consumers, let’s now take a look at the most frequent tracking and recognition use cases from two complementary angles: relevance and confidentiality.

To correctly monitor user experience (and therefore optimise it), we need to be able to associate a single consumer with a series of items, some of which are exclusive to the consumer in question, some of which are shared. Without this key, all results would be unreliable, as would be any analyses stemming from them. And there are different ways of obtaining this key, some of which raise problems with regards to consumers’ informed consent.

Let’s study a few examples:

When using everyday consumer services online, most of us have been asked to share a phone number, in addition to providing our email address, with the rationale of “protecting” us, the consumers. But behind this display of goodwill, might this actually be a simple but masked way of recognising mobile visitors and linking our web and mobile activity?

“Fingerprinting” is already used in the United States for the purposes of targeted advertising.

In the United States, a class action suit against Google – who allegedly scans Gmail messages for “spam blocking and viruses” – is another more underhanded example. While it’s unlikely that Google will expose the minute details your private life, is it unreasonable to think that this data collection might be used to qualify databases, and be linked with the notorious AdID (unique advertising ID)?

As companies happily fill their databases without consumers’ prior informed consent, let’s take a look at the most common recognition techniques.

What are the different recognition techniques used?

#1 The simplest of them all – the identified visitor: Each person uses the same log-in details, no matter what channel he or she is on. This is a web analyst’s paradise, as the consumer has voluntarily identified him or herself and therefore accepts to be recognised. The consumer will then be recognised across all channels used, and the analyst is able to track and study the user’s experience with the brand.

  • Relevance: Accuracy and a comprehensive view guaranteed
  • Confidentiality: No problem here, as long as the consumer has been clearly informed of data usage.

#2 Recognition by cookie (text or flash)

cookie-glass

This small file stores non-confidential data on the device that has been used.

The cookie will be updated during each visit. If cookies are refused by default (browser configuration), then there will be no recognition. If cookies are deleted on a regular basis, the visitor is treated as a “new visitor” for each visit after the cookies have been deleted. If the cookie has never been deleted, there is still the problem of its lifespan.

  • Relevance: While the device is recognised, the user, on the other hand, is not (except in the case of fingerprinting). Let’s take the example of a family computer where 3 visits from 3 different people have been recorded, including both parents and their child. The 3 people have all been in contact with the brand but with slightly different user experiences. For example, if we consider a music site: one of the parents is a reggae fan and searches for old recordings of Steel Pulse, the other parent loves Stromae, whereas the child swears only by Daft Punk. We can reasonably question the relevance and accuracy of the user profile that will be attributed in this case…
  • Confidentiality: Provided that there is no hidden cross-referencing of databases, confidentiality will be respected and the data will remain anonymous.

#3 Recognition by cross-referencing of databases

This system is widely used by ad exchanges and Data Management Platforms (DMPs) to manage ads. Another recent example is the large TV media groups who have partnered with social networks. Assuming that these media groups will only track comments on these networks, you might wonder what the real point of the partnership is, since a simple API could be used to achieve the same results. But if we assume that the partnership includes the exchange of private data, the true significance of the partnership becomes clear: Private data coming from social networks can be cross-compared with other databases containing a common identifying element (name, telephone number or email address, etc.).

  • Relevance: In this case, relevance is inversely proportional to confidentiality
  • Confidentiality: Confidentiality may be stretched here. Private data may be used, as it is provided voluntarily by the user (but for another use). Once recognition is complete (association of channels), the data can then be made anonymous.

Effective multi-channel analysis relies mainly on visitor recognition through several different channels and interfaces. Such recognition is very difficult to carry out in a comprehensive manner, except through processes which more or less contradict the principles of confidentiality and respecting private life. We can safely conclude that this analysis is extremely accurate for sites which require private access, whereas in other cases the analysis only applies to smaller quantities of data, except for where private confidential data is used without the visitor’s knowledge.

The themes of confidentiality and use of private data need to be clarified: Can data provided for reason X be used for reason Y (without informed consent) as soon as it has been made anonymous again after a correlation has been established? Since the final data is anonymised, there is no violation of confidentiality, and it is along these lines that the majority of companies will play.

And that leaves room for a very wide debate…

Comments are closed.