The Keys To Web Analytics Data Quality – Get Your Consent Management Right

Poor data quality is a silent killer for many websites. Misleading data analytics undermines every decision-making. One of the first steps often is to improve consent management.

Consent management is not only about privacy and customer respect. It’s also about securing high-quality and reliable data. Missing this, you won’t hit good strategic decisions. It’s like a path full of hindrances and garbage that makes you stumble. Your optimisations are prone to failure. As a proper setup can be hard, let’s try to clean up this path – and give a focused look to the crucial aspects.

This post is part of my blogpost series “How to get Web Analytics right“. In the series I talk about the 3 most fatal flaws in Web Analytics setups that I observe most often:

• Flaw No. 1: Missing Web Analytics Strategy, Goals & KPI.
• Flaw No. 2: Bad data quality. Consists of 2 parts. Part 1: Consent Management (this article). Part 2: Track the complete Customer Journey (not published yet).
• Flaw No. 3: How to finally use your data (not published yet).

Today, let’s go and fix your Consent Management! I’ve packed lot’s of my collected knowhow into this article.

2. The problem of low consent rates

You have solved your Consent Management data quality? Great!

Now, here comes the second crucial challenge:

Many organisations generate too few data. Having enough data is crucial for each organisation and project in the digital space. Each Web Analytics strategy can thrive or fail exactly by this. For many organisations only few users are willing to give consent. Because people want their privacy to be respected and protected. The more technical your business is, the more careful your visitors will be. In my case, the visitors of my website – folks like YOU 🙂 – are typically VERY privacy aware. 

But you indeed can get both: Good quality and correct consented data. AND enough data for reliable decisions. And then also use it. I can help, to finally achieve all this. This is the goal of this blog posts series.

But first, now let’s get even deeper into the problem:

2.1 low accuracy leads to Wrong conclusions  

If you don’t have enough data or minor data quality, you’re more likely to draw incorrect conclusions. Data reliability is key!

The main question is whether the data you have – hopefully a not too small piece of data – is representative of all your users/clients. If all types of people i.e. customers are not equally represented in your, this can lead to biases and manipulate accuracy.

Thus, since only a portion of the total population is analysed, there is always some degree of uncertainty regarding the accuracy of the results. Sampling errors can cause estimated values to shift from the actual values of the overall population. If the data you have accessible is not randomly selected or certain groups are underrepresented, the results may be misleading. 

Working with i.e. only 30 or 40% of your data (as this can often be the consent rate of a consent management tool) will result in a smaller sample size than ideal. Smaller samples tend to have higher variance, meaning that the estimate outcome is less precise.

To reduce these problems, it’s important to either make sure your data is arranged very carefully with a focus on correct statistical methods. As only few do this, especially with “normal Google Analytics data”, my advise instead is quite simple: Go and get more data. This way it gets more representative and more reliable!

2.2 How to find out, if you don’t have enough data

There are two important borders to make out, if you have enough data, or not:

1. You should have a big enough share of data.
2. You should have a big enough absolute amount of data.

First, let’s talk about the share of data.

Getting insight into at least 50% of your share of data is critical. Why? When you’re below that line, the majority of your website visitors remains invisible to you. As a consequence of this, false conclusions become more likely.
For example my website, has a consent rate typically somewhere between 10-20%. And I’m not the only one with that low consent rates. During the last years I’ve seen MANY clients and organizations who get only consent from below 50% of their users. 
Sure, you can also be below that line. But then you should start investigating your data more thoroughly and make sure you avoid wrong conclusions.

The second important border is the absolute amount of data.

You could have 100% consent of all your visitors. And you could have no losses because of browsers or plugins blocking tracking technology. And still, you are prone to wrong decisions. Because of too few data.
To find out your organizations individual border of your needed amount of data, you can calculate statistical significance. It’s important to reach it. 

Statistical Significance. how to calculate it.

To put it plain: If you don’t really use data, then you don’t have any problem when there’s not enough data. But in case you want to do & improve something, then the question towards “enough data” emerges.

Start by understanding which questions you want to get answered. Which goals and topics do you want to improve? Maybe, you want to test a hypothesis about your website visitors, or a segment of them (i.e. users with mobile devices). For example, you want to optimise a calculator on your website. Today, too few visitors finish the calculator and see its result. Let’s say, you would like to reach double amount of finishers. And let’s assume you have a very good idea, what needs to be done in order to reach that goal.

In the next step, now you can use an online calculator to find out, how likely it is that your observed effect is really true. What are your ingredients? 1. The amount of your calculators total visitors. 2. The amount of your calculators visitors that reached finish. You need both values before and after havin applied the optimisation, i.e. for the time span of one month each. Type these numbers into the calculator.

As a result, you will get the feedback, if the change reached significance or not. Significance is determined by the so-called p-value. Small p-values (typically < 0.05) indicate a strong evidence of an effect, while large p-values suggest only weak evidence. In case of a weak evidence, you should get more sceptic & more data.

You can use this calculator to check it out.

2.3 How to Obtain more data – a respectful approach

There are different possible ways how to obtain more data. First thing is, of course, to reach more people. But today, we are not focusing on this. Instead, we are talking about getting more data from your existing and unchanged audience. There are mainly two approaches for this, how you can obtain more data.

1. One possible solution to this is, to improve the share of people who consent. Consent rate optimization. I am not going into details on this today also.

2. The other approach is to also track – in an anonymised way – people without consent. What many don’t know: This indeed can be legal. If you configure it right. When you remove all personal identifiable data (PII) and user profiling, then you may have no problem when tracking website visitors without consent. The need to get a users consent is all about collecting personal information. When there’s nothing personal in the collected data anymore, then there are no laws against it. 

And even one more tip: The above mentioned low accuracy and high risk for wrong conclusions (i.e. when you have less than 50% consent rate) can be a good reason talking to your data privacy officer. Tell him, that it is really needed to get more data, in order to draw reliable business decisions. This reason can change your situation, and make collecting more data possible. 

Now, let’s have a detailed look into above mentioned second approach “collect anonymised and consentless data”, and how some Web Analytics tools cope with it:

2.4 Web analytics tools – And how they cope with anonymised data

The analytics tools I work most with – Piwik PRO Analytics, Matomo Analytics and also Google Analytics – offer technical solutions and possibilities how collecting anonymised data can be achieved in a legal way.

Piwik PRO Analytics and Matomo Analytics

Both are hosted inside EU. Thus for them it is – according to EU laws – allowed to collect anonymised data without consent.

The tools register if users consent has been given or not. And send – based upon this – more or less detailed data. They send data with personal information like IP addresses and cookies, in case you have users consent. Or, in case you don’t have it, then without such detailed data.

If you start collecting anonymised data with these tools, you need to know about some additional challenges. But principally, it is possible! Maybe I will talk about how to do this some time later. Please let me know in case you’re interested in this.

Google Analytics

Google Analytics is hosted in the US. That is why it is – in EU – more complicated to track data without consent.

However, Google Analytics offers “Consent Mode V2 – Advanced Consent Mode”. Advanced Consent Mode gives also possibility to deliver reduced and anonymised data.

The problem with US-based hosting is, that by “classic” or so-called “client-side” data submission, some users personal data like i.e. IP address cannot be removed. As an online connection between two different servers needs an IP address. Otherwise connection is impossible. For this reason, in case of Google Analytics serverside tag management needs to be added to the setup. This technology then makes removal of IP addresses possible. 

Again, some details need to be taken care of. But principally, when having both, Advanced Consent Mode V2 and serverside tag management, then it’s also possible with Google Analytics to track consentless and anonymised data.

One advantage comes with Google Analytics: If you have achieved such a setup, then Google has a more advanced and AI-driven approach as Matomo and Piwik PRO to analyse consentless data: It kind of renders a more complete picture with such type of data. It models and closes the gap of people without consent. However, this so-called “modelled” data view gets only available, in case you have quite some amount of traffic. Thus, if you have enough traffic, I can recommend this. But in case you have less data available, again, this option will remain unaccessible for you. 

For this reason, I recommend especially for smaller websites, to choose Piwik PRO instead of Google Analytics. With that analytics tool, more data access is simpler possible.

TL;DR

The journey towards transparency and trust requires strength to navigate treacherous terrain of data protection laws and data quality pitfalls. You need to foster trust with your users while maintaining a solid technical tracking basis most efficiently.

The result of this will be a well executed Web Analytics strategy and good data quality, with which you can derive intelligent insights – and improve your experiences as well as your whole business.

By focusing on some very important features and aspects of consent management platforms, on user-friendly interfaces and some core data export features, you can ensure that your consent management not only meets compliance requirements, but also enhances your own data transparency and – most importantly – your overall digital strategy. While no tool is perfect, the recommendations shared above have proven reliability in various scenarios.

If you’ve discovered a solution that fits perfectly or have own experiences towards data quality or consent management to share, I’d love to hear from you in the comments. Together, we can refine our approaches to this ever-important aspect of digital responsibility.

P.S. My next blogpost will be about measuring not only data correctly (what was the topic here), but also measuring the correct data: How to make sure you collect all those relevant touchpoints of your customers journey. See you soon!