An event where you do not know in advance who will attend or what the talks will be; it is all created on the spot with paper cards attached to a time schedule. These are not long talks: each lasts about 25 minutes and can be of a different level of difficulty. This year MeasureCamp was also held in the Czech Republic, in Prague, at the ČSOB headquarters on Radlická. The adventure itself began with the attempt to get tickets, which are offered for free because the event is supported by sponsors. On a certain day and time registration opens on the internet, and the fastest manage to get a ticket. We succeeded and made it among this year’s 215 conference participants.
The language of the presentations was Czech as well as English.
Rich refreshments were prepared for us, both for breakfast and for lunch. We also had several interesting discussions, one of them a reflection on how GA4 data can be turned quite easily into a CDP (customer data platform).
We also had the chance to give a talk ourselves. Tom presented to the audience the topic of Anonymous Measurement, which was a great success. He explained the issue of consent, how we collect data, where we send it, how we report it, and showed a specific report. The classic questions we hear quite often came up too, such as what to make of undecided rate = 100% – (consent rate + decline rate), that is, skipping the banner, blocking it with plugins, or the bounce rate we know from the previous Analytics.
MeasureCamp definitely did not disappoint us and once again led us to many interesting ideas.
Michal Pich
Steen Rasmussen – The Decision Economy
Steen reflects in his talk on the value and function of data. He urges us to collect only data that we will actually use, and at the same time shows that the absolute majority of analysts deal only with descriptive analytics, that is, with what actually happened. During the talk he urges us to see that the value of data lies rather in the other parts: the diagnostic part – why what happened happened; the predictive part – what will happen in the future; the prescriptive part – how we can achieve what we want to happen. At the same time he shows what value data has over time.
An ideal talk for the beginning of a conference; Steen is a great speaker and at the same time a veteran of analytics.
The Second Talk
The second talk dealt, rather than with presenting data, with general presentation skills. A relaxing part, most suitable for people who present directly to a live audience. There were also a few tips for video calls.
Frederick Werner – Incrementality of Analytics
Are we actually driving value for the business? How do we know? Let’s discuss!
Frederick, a big advocate of Adobe Analytics , deals with several ideas in his talk. The first is that a lot of “expensive” tools can end up cheaper if we use their full potential. For example Adobe Analytics, a much more expensive tool than GA4, can save us a lot of time on reporting, because GA4 reporting is currently very sparse.
In the next part of the talk he gets to the incrementality of analysts itself, that is, how to think in general about calculating the contribution of analysts. The most concrete idea he tried to pass on was to calculate the percentage impact on the people to whom the analyst supplies work and to attribute this total impact to the analyst. An idea that is certainly interesting and much more accurate than pretending that no work would get done without the analyst; unfortunately, I cannot really imagine this percentage expression in practice. In a sufficiently mature company this approach could be ideal.
Vojtech Kurka – Cross-domain tracking you have never seen before
Vojta gave a technical talk on how they try to work with cross-domain tracking at Meiro. Technically a very interesting talk. At the end a small discussion broke out about the ethics of this approach.
Viet Anh Chu aka Chuan – Dataform & GA4 ecommerce data
Chuan presented a new package that they are developing at Optimics for Dataform. A very interesting idea, especially because Dataform is integrated directly in Google Cloud. I am very curious how many people will move from DBT to Dataform and how much these tools will influence each other in the future.
Mája Remešová
David Vallejo – Analytics Firewall
His talk was devoted above all to how to protect data against attacks. It covered recommendations for securing data, how and at which point of its journey to analyse and sort the data, and how to encode sensitive data or not publish it before it reaches a server that an ordinary user has no access to.
He mentioned individual tools that can be used for filtering data on the client side, on the server side and in the ETL layer. David basically gave us interesting metrics for sorting data at the input. He also recommended masking the ID, hashing it and hiding it in a phrase that is recalculated on the server before being sent to Google Analytics. It was interesting to follow the discussion that arose from this. Several people in the audience raised the problem that hashing data works but at the same time breaks sending data to other systems such as Google Ads. So it is usable, but it depends on whether the data is sent purely to Analytics or to several systems, or whether a way to handle this situation appears in the future.
Fred Pike – Tips for GTM debugging
Fred is the author of the newest courses on GA4 in its current form on the CXL platform, from which we very often draw new ideas and techniques and where we regularly educate ourselves. In his workshop he showed us what we can use when setting up analytics in GTM so that it is easier to debug and easier to identify possible errors, especially when you have a complicated container (for example, several companies worked on it) and send to several systems at once.
His recommendation was to send initial data in the Google Tag (formerly the configuration tag). So for more complex implementations where there are several authors in one GTM and lots of tags, it is useful for debugging to use a shared event parameter (formerly the event parameter in the configuration tag) named gtm_info, which carries the built-in variable for the container ID and its version ID. In the event tag itself he recommended using an event parameter tag_name, into which we put the name of that tag. When an error then shows up in the reports, we can easily identify in a table from which version of which container the error came, or which specific tag caused it. Unfortunately there is not yet a way to get the name of the event tag we have chosen into a variable, so for the tag_name value there is nothing for it but the well-known Ctrl+C, Ctrl+V. Debugging should then be easier even for a person who did not do the implementation themselves.
Robert Petkovič
Robert’s talk appealed to me a lot. It focused on how to present from a psychological and analytical point of view. He summed up what is key for a talk to capture attention and for people to take something away from it, so that someone can still say the next day what they remember from it.
We got many recommendations. He advised us to read a book in the language in which the speaker will be presenting (this makes it easier to adapt to the particular language, as the speed of delivery and the style of presenting get aligned). It is also key to realise who the listeners are, whether they are experts or beginners, and to adapt the presentation to them, and many more tips…
Krista Seiden
Krista is an American specialist who works closely with Google and directly with the development of new components and reporting. She also presented a course that she worked with Google on creating. It is an academy about GA4 on Skillshop (Google Analytics Academy), available online to anyone interested.
Tim Wilson
Tim’s talk was a very pleasant experience. In its content he dealt with how to work with data. According to him, the key is to align 3 things:
- Performance measurements = looking at where we are today compared with what we expected a few years ago;
- Hypothesis validation = what is our idea that, if we carry it out, will have a positive impact on our future business;
- Operational enablement = how we can process data for practical use within a process;
At the same time he dealt with a similar idea that has been bothering me for some time, namely that a terrible amount of data is measured. A large part of this data remains completely unused, users do not look at it, and it is measured so that it is available in case someone happens to need it. The whole thing has an even more serious consequence: the owner often gets lost in the flood of data, so does not use it for growing their own business and starts to perceive analytics as unhelpful.
He therefore presented an alternative to the concept of data collection, where the key parameters of the business are first discussed with stakeholders, owners and marketing experts, and on that basis data collections and clear reports that users can find their way around are created. They would rather not measure a billion data points from the start just so that everything is there immediately, but spend time analysing what the client wants and what is relevant for them.
Vlad Sidion – Measurement protocol
Part of this talk was to sum up the basic prerequisites that must be observed when using the Measurement Protocol, its use mentioned for measuring from devices in the offline world and in shops, and above all what to watch out for and which part of the data we lose by sending data to GTM or GA4 this way.
Tomáš Ondříšek
David Somar – Using AI (ADA) in Data Analysis
David dealt with the ChatGPT tool and how to use it for data analysis. On the paid version of ChatGPT 4 he showed us a clear example of analytics. He uploaded sample data as a CSV into the tool and, using a complex command, obtained a clear and complete analysis.
In the second prompt the intended output was a business strategy. So he again told ChatGPT to create, on the basis of the data it had available, a meaningful strategy aimed at maximising ROI. The tool generated a roughly five-step strategy with the given procedure, focus, tools used and budget.
Jan Javurek – Struggles of GA4 Cost import
In the talk we dealt with a tool or plugin that they created and that is able to import into GA4 a report showing how much our advertising costs us. Honza showed a practical example of how to make a report via URL addresses, whereupon a new chart is created in Google Analytics 4. This chart visualises the complete campaign costs.
Because we do not use this way, it was interesting to see another method that reaches the same goal. We build these processes in Data Studio, where we insert two different tables and then blend them.
Krista Seiden – GA4 Issues & tips
Krista focused on classic beginner problems in GA4, such as measurement errors without custom dimensions, switching off enhanced measurement in data streams, and the like.
The rest of the talk was devoted to how to make custom reports in GA4 that you can share with a specific team of your client. This function is suitable when we have, for example, 5 different reports named “Marketing team”, into which only those reports with the given filters that this team specifically needs are placed.
Again it is interesting to see a different way from the one we use, as we once more rely on Looker Studio.
Parameters in Looker Studio, or: Can You Make 1 Report for 1000 People?
The speaker presented a specific brief from a client, the start-up pickey. In the administration, where they have the basic metrics, they created a report that will show the number of subscribers, how much money they bring in, the number of views of posts and similar data. He showed us ways to create such reports.
There were several ways before he arrived at what really works. It can be done with parameters in Looker Studio, into which the author’s name can be put. Parameters have the advantage that they can be connected to BigQuery and work like SQL queries. That is, once the data gets into BQ, it can be queried through the parameters. In the early phase of the platform it was done by putting into the administration a link through which a URL would be launched and would immediately pull out the name of the author who clicked on it. The platform is still developing, but in the future it is being considered that Looker Studio would be embedded directly in the website by programmers.
Alternatively the Streamlit tool (an app for building data apps and reports) could be used, which would solve all the problems including integration into the administration.
Meiro.io – Hit Log Log (HLL)
A very advanced talk from Meiro.io focused on the HLL method, which they explained on the basis of binary code. To explain the method they also used a comparison with bitcoin mining, which helped to make the whole issue clear.
The Hit Log Log (HLL) method is a native function of PostgreSQL and works within the BigQuery SQL database. More or less it is one of the commands in an SQL database where it is possible to query data. The difference between classic querying and HLL is that HLL is several times faster in processing. What a classic SQL query takes 4 seconds to do, HLL processes in 0.5 seconds. That may not look like a big saving, but in the end queries are quite large and the longer it takes, the more data it costs. At the same time, when data is pulled from BigQuery into Looker Studio and joined with other data on top of that, it makes a big difference in the end result whether processing one table takes 5 seconds or 0.5 seconds.