All sectors and sizes
of companies are potentially affected by the unreliability of their data. Small
and medium-sized organisations generally do not have the time or resources to
assess quality and/or resolve reliability issues.
Large companies tend
to generate a lot of campaigns and new web pages. They cannot keep up with the
pace of publication while respecting data quality standards.
The poor quality of
analytical data can take different forms, with sometimes serious consequences
for your business:
- loss of income
- reduction in the ROI of marketing actions
- loss of quality in decision-making
- contamination of other data projects (CRM, Data Lake, CDP, etc.)
- decreased internal trust and credibility
The intrinsic risk in web analytics
Without preventive action, the quality of the data is altered by nature. Sources of error are diverse and inherent to some web technologies: unmeasured data, robot traffic, browser inaccuracies, traffic jams, etc. We have identified, quantified and represented here the main risk factors that threaten the quality of your analytical data.
The
critical phase of collection
The data collection phase is critical
because it is permanent. Each optimisation, functionality, new campaign or new piece
of content poses a risk to the quality of data collection. An effective
collection strategy brings together all the company’s decision-making players
and adapts to each development on an ongoing basis. This reflection on data
collection must be taken into consideration when defining a data governance policy.
The more you update and enrich your mobile sites and
applications, the more likely you are to inadvertently alter your analytics
tags. This seems to be an elementary mistake, but it is very common to find
missing, defective or duplicated tags, especially on large sites with a lot of
content. While these tagging problems, sometimes minimal, can be difficult to
detect, they have a significant impact on performance. It is of vital
importance to be vigilant over the integrity of tags! Checking the source code
of all pages is therefore essential, but who has the time to do this tedious
manual task? Crawling tools allow you to
automatically browse the site, all pages and sections combined, to check the
presence of digital analytics tags. Others allow you to check your tags live once they are
implemented on a site. They can also issue a report indicating the problems to
be solved.
Robot traffic
According to some estimates, robots (or
“bots”) account for more than half of web traffic. To know exactly
the actual volume of flows, it is essential to have the means to identify and
exclude the part generated by the robots that visit your sites. However, some “bad
bots” can be very difficult to detect; hence the importance of working
with a digital analytics provider that has the
experience and means necessary to recognise and eliminate this traffic. The
ability to discard flows caused by robots has direct consequences on data
quality.
In addition to the qualitative aspect, the manual
sorting of this polluted traffic is enormous, if not impossible for the person
analysing the data. As a first step, your Web Analytics provider should be able
to identify these robots using the official exclusion list published and
updated regularly by the IAB. It must then offer you the possibility to regenerate
your data, over the desired period, excluding this robot traffic.
Source allocation biases
Some events, such as Facebook’s overestimation of
video viewing time and the temporary suspension of two Google indicators by the
Media Rating Council for “non-compliance” with measurement
guidelines, have given companies reasons to question the accuracy and validity
of the data they receive.
At a time when transparency seems to be lacking, one
can begin to question the accuracy (and impartiality) of indicator calculation
in these restricted access systems. Here is an example with a very simple
question: can we really rely on the figures provided by an analytical tool for
the source of a “search engine”, when it is at the same time
generating the revenues of this tool? One of the last and most striking
examples is the bias in the measurement of source attribution by Google’s
analysis tool. In other words, the conversion is automatically assigned to the
Google source (engine or sponsored link) if the visitor has clicked, even once,
on a Google link in the last 6 months. The measurement tool thus completely
ignores the sources of direct traffic (link in favourites, automatic entry in
an engine for example) to assign the conversion to itself. In other words, if
the source is not determined, Google takes it over. The result: conversions
that add up and numbers inflated in the counters of advertising channels like
Google Ads. Nearly 20% of conversions are overestimated
due to misallocation of sources.
Fortunately, it is possible and simple to act to
reduce risks with the appropriate tools and procedures. The most difficult
thing is to be aware of the potential sources of errors.
AT Internet offers a wide range of tools for quality control of analytical data. Fewer errors are therefore likely to alter your
data and influence your decisions.
If you are keen to find out more about data quality,
download our latest guide:
