Digital Analytics Scholarship @ CXL Institute — 12th Week Review

In the 12th and final week, I went through 2 separate courses: Google Analytics 4 and Attribution. It wasn’t a difficult week, because I had managed to leave the easier stuff for the last part of my time. I thought I’d be exhausted by then and I think I was right!

Let’s see what I’ve been through:

-1- Google Analytics 4 by Charles Farina

• One of the biggest changes in Google Analytics 4 was its method of tracking. Instead of all those hits that were sent to GA in universal analytics, now we can rely on events. Because it is the method that is used by GA4. Even pageviews are sent as an event.
• In GA4, you can measure scrolls, downloads, mouse clicks, and a lot more by default. If you are a fan of tag managers, you have to know that we used GTM to measure stuff like these. But now they are being measured by default in GA4.
• The thing about GA4 is that it doesn’t work in retrospect and previous data can’t be seen in it. You can’t have the data you had in your UA implementation in this new GA4 container. The method of measurement has been completely revised in previous data and isn’t compatible with it.
• Charles Farina says that this new version of GA is much better at managing funnels. I don’t know what he really means, because the only thing that I think wasn’t available in the previous version was custom funnels. And this very thing was available for 360 (paid) users. Anyway, this is what Charles says and I’m eager to understand what he means.
• Another thing about GA4 is that it offers a lot of features that were only available to paid users to free users. So you can leverage some of the 360 features for free. For example, you can have access to BigQuery integration for free.
• In the past when GA4 wasn’t branded and known by this name, they called it App + Web analytics. Meaning that in GA4, Google Analytics has been merged with Google Analytics for Firebase. Now you can measure your iOS and Android applications as well as websites in GA.
• A big question that was a matter to me a lot of others was this: should we ditch the currently working universal analytics and migrate to GA4? The answer is a sharp NO because Google is still experimenting with the new installation and it hasn’t still fully landed. Lots of features are missing in GA4 and you need to have it alongside UA.
• GA4 comes with lots of changes and improvements in the user interface which might take some time for you and me to get used to them. But it is the new normal (inspired by CoronaVirus Age) and we need to accept it from a point in time. Consider that someday, eventually UA will be deprecated and google won’t support it anymore.
• This was all a preview of GA4. Let’s go and see what has changed and what needs to be learned.
• In the next lesson, Charles took our hands and showed us how we should install GA4 on our website. He made a simple website and installed Tag Manager on it in order to install GA4 using a GTM tag.
• To be honest, installing GA4 is much harder than the previous universal analytics version. Here you should first build a data stream and only then you are allowed to choose how the data can be collected.
• GA4 has its own dedicated tag in google tag manager and it is only used for configurations. Meaning that using the configuration tag, you can only set up a pageview tag and nothing else. Every other thing you want to send to GA needs another tag which is in charge of events. GA4 events I mean of course.
• It was fairly easy to set up GA4 and I want to know what else has changed.
• Google Analytics has changed a lot in GA4. one of the biggest changes in my opinion is the removal of views. Universal analytics had accounts, properties, and views. But in the new GA4, views are completely removed. Now that we don’t have any views, all of those filters, goals, eCommerce settings, channel grouping, or content grouping are gone or relocated.
• Charles says that data streams are kind of like views in universal analytics. But I am thinking about what about those tests, raw data, and master views that we used to build in universal analytics?
• One of the things about GA4 is that it can detect and filter internal traffic out of the box, meaning you can exclude your teams’ traffic so that it doesn’t affect your measurements. But it acts in a weird way and you should look for an awkward place. You need to open your data stream (whatever it is) and only then you can filter internal traffic.
• Filtering for this moment only can be applied to internal traffic and if you want to see whether it is including or excluding the traffic, you need to go to another menu item in the settings. It looks and is unintuitive, but this is what google has served us and we have to accept it.
• A big thing about filters is that they don’t support regex rules. Can you even believe that? Regex rules are at the core of technical marketing and they don’t exist when you want to create filters. This really sucks.
• Another thing that we need to know is that the referral exclusion list doesn’t exist in the newest version of google analytics (at least for the moment). So what should we do for setting up cross-domain tracking? We have to go to our data stream, select tagging setting and then, go for the “configure your domain” option. There we can select what domains we want to be considered for cross-domain tracking and done!
• One thing that is actually cool regarding filters in GA4 is now they have stated. Once you build a filter, you can specify whether it is active, testing or deactive. So now you can test filters in a controlled manner (trying to replicate test views? maybe)
• One of the changes that are obvious in the new UI is that Google has tried to mimic GTM’s interface in the new GA4. they look and feel almost the same, especially in the admin settings.
• Oh, and your data only stay in the google analytics platform for 2 months, unless you go to settings and change it to 14 months. So why is this happening? Why do we have only 2 and 14 months and nothing else? Who knows?
• But I have to say that I am really excited about the direct and free integration between GA4 and BigQuery. BigQuery is a great data warehousing platform that lets us mess with data at large scales and query them. Of course, using BigQuery is a paid service, but having basic access to it from within google analytics is something that seems to be valuable.
• I have one criticism for the CXL team: the audio quality is terrible in this course and the sound gets cut in lots of situations. I don’t know if it is only for me or others who are having this issue too. I accept that this course has been prepared rather quickly and this GA4 is still a trending topic among marketers. But this level of quality isn’t expected from a great team like CXL.
• We said that events in universal analytics and GA4 works quite differently. The difference lies in the process and methods that have been used in them. For example, if you remember how event tracking works in universal analytics, you might know that we have event category, event action, and event label. In GA4, there are no such things. Instead of all of them, we have parameters.
• Parameters let us define what we want the thing to be called, not necessarily category, action, label, and value. You simply name the event to something that you want and then define these parameters one by one. It is a little bit annoying, but it is what it is.
• You have to know that these parameters won’t be displayed in google analytics the way they used to be in universal analytics. By default, GA just shows the name of the event that has been fired. If you want to see these event parameters, you need to define them using custom dimensions. I know it feels a little bit awkward and bizarre, but you have to get used to them.
• Do you remember that we had goals in universal analytics? Now instead of goals, GA4 has something that is called conversions. Conversions now can be turned on and off and only those that are on will be counted as a limiter. You have a limit of 30 open slots in conversions (instead of 20 in UA).
• Conversions can be made out of events, meaning you can set the events to be counted as conversions. It might be useful in a lot of cases. Another cool thing about GA4 events is that they can be made using existing events, right inside google analytics. You can just set a subset of an event to be counted as a conversion and use it for a specific purpose. You can also modify events inside google analytics. To be honest, I don’t like all of this cool stuff and the previous version seems much convenient in my eyes.
• Another feature in GA4 that I forgot to talk about is its debug view. If you use google tag manager to configure GA4 or have installed Google Analytics Debugger on your Chrome Browser, you can use this new addition. Whenever you enter preview mode in GTM, GA4 automatically detects this and shows you the things that are happening in it.
• Do you remember the time that you wanted to create custom dimensions in universal analytics and you set the scope (user/session/hit/product)? As we discussed earlier, hits are completely removed from GA4 and everything is based on events. But what about those user scoped dimensions? GA4 has something named User Properties that lets you create user-level dimensions.
• If you want to take data out of google analytics, you can count on BigQuery linking. It has been built right into GA4 and you can directly connect GA to BigQuery. Of course, it is a premium service, but it is a great thing to have it in here.
• Charles Farina also talked about Data Deletion that lets you delete a part of data. So if part of your data has violated data governance rules, you can delete them and everything can be ok.
• Charles started talking about standard reports and how the UI has changed in the new GA4 that you might not find things that you already know in it. It is somehow odd to see things having this level of change that even professional users can’t find the reports they want.
• Another thing about reports is that your data won’t be sampled in the new GA4. even if you are in the free version and have applied segments or secondary dimensions, nothing will be sampled at all. This is very pleasing to hear.
• Another thing that has changed is the way GA4 treats data. Previously in universal analytics, we had acquisition, behavior, and conversions. Now we have a new set of reports: acquisition, engagement, monetization, and retention. Do you remember that in behavior reports, you couldn’t understand how much any page has impacted your desired goals? Now it is possible to see them in the engagement report.
• In universal analytics, we have access to advanced search and we can apply any kind of queries, including regex queries. But in GA4, none of those things are possible. We can only query based on the “contains” option.
• GA4 has a new setting that lets you change the attribution model on the fly. Now you are not bound to “last non-direct” attribution. You can change it to other options as well
• One of the new features in GA4 is the Analysis module. It allows us to do the things that we were never able to do before and this was only available to paid users. It allows us to do meaningful funnel analysis, pathing reports, and all sorts of cool stuff.
• I remember times at which I needed to create funnels not only based on destination (page URL) but based on events as well. It was called Custom Funnels in google analytics and only 360 members had access to it. Now we can create a new funnel and in it, we can define the steps based on dimensions, metrics, and events that we have created. This is a whole new level of freedom.
• Once you are in the funnel report in the analysis section, you can create segments based on the users who have gone through your funnel steps. You can use these user segments (audiences) in your next reports.
• You can make your funnels open or closed. An open funnel lets your users enter them not necessarily from the first step, but maybe from the second or third step. We don’t have this option in closed funnels.
• Another new addition to funnel reports is that funnel visualization can be done in a new form which is trended funnel. A trended funnel shows us a bar chart that depicts changes over time. This option wasn’t available before, even for 360 members.
• One of the greatest improvements in google analytics 4 is that you can analyze the path to conversion very easily. In the Analysis section, you can choose path analysis as the technique, and then, you have two things to specify: the starting point and ending point. If you specify the ending point, you can do a backward analysis and understand what led to a specific conversion. It is something doable in GA3, but rather difficult.
• Another technique that is available in the analysis section is segment overlap. It creates a simple Venn diagram and you can understand where your segments and dimensions have overlap. Nice addition, huh?
• Analysis has another technique that I believe is important because it gives us the features we don’t have in GA4. This technique is called exploration and it creates a simple table that we can fill with the dimensions and metrics we prefer. Here we are allowed to perform all sorts of filters, visualizations, regex, and other stuff.
• One of the things that separate GA4 from universal analytics is the way it targets audiences. We have two types of audiences in GA4: one of them is temporary and is good for ad-hoc analysis. The second type is built in the audiences section and collects data from that point forward.
• The difference between these two types is that one is temporary but retroactive. The other one is permanent and completely ignores preview data. You can turn a temporary audience into a permanent audience.
• Audiences are directly reported to google ads, so you don’t need to send them over manually. This is a new change that has come with the release of google analytics 4.
• There was something called an audience trigger which I didn’t understand at all. I don’t know what it is good for, but it was right there in the audience builder section.
• That is it. I had google analytics 4 with Charles Farina. How was it? The content was rather good. But do I need to say once more that the sound quality was terrible?
• I don’t feel it is time to migrate to GA4 because it is under heavy development. Everything might change so rapidly that renders our knowledge obsolete within days or weeks. I personally tend to stick with the UA and postpone the migration to a point that using UA isn’t an option anymore.

13- Attribution by Russel McAthy

• This is one of my favorites, especially because the tutor is a talented young guy who says he won’t merely explain different attribution models. We are about to understand what attribution is and what we can do after understanding it.
• So what is marketing attribution? It is basically a context in which marketing data will be meaningful. We have to know which one of our channels added the most value to the customer journey and brought the most paying customers.
• But Russell says we want to talk about consumers, not just paying customers. He says we want to study our consumers and find out how they behave throughout the entire journey.
• Marketing attribution is something that can be related to both B2B and B2C markets. But since most of our marketing activities are centered around the b2c market and decision-makers in the b2b sectors are generally more than one, we are focusing more on the b2c segment.
• We have a couple of players in the world of attribution software. but as you might know, Google and its Analytics is the major player in this world.
• This marketing attribution thing will help the head of digital marketing and/or CMOs to determine their return on investments (ROI). So understanding it will help them make better decisions for their companies.
• In the next lesson, Russel talked about the fact the attribution is not only a thing for online marketing. Offline marketing should have a role in the attribution because it too is part of our marketing efforts.
• But how can we measure attribution and ROI when we don’t have any data in the offline world? Russel says we can use short URLs and coupon codes to track offline activities.
• TV stations and satellite channels also give you some data in the new modern age. Besides, most of the media channels have turned into on-demand services that are fully digital and data can be captured from them.
• When you are working in a marketing team, all of the members should believe in and follow a unique attribution model.
• You might ask do I need attribution? If you multiple touchpoints with your customers the answer is yes. If your sales cycle is not a short one, the answer is yes again. This is something that will give you all the insights you need to know your consumer better and address the pain points (sort of)
• One of the greatest problems that marketing teams and organizations face regarding attribution is knowledge. Having access to the technology and all the data that it generates isn’t enough. You need to have people literate enough to utilize the tools the way they are designed to work.
• Attribution models and all the things that we have heard in the past. I think you know what they are and what is the difference between a last-click vs. linear. Each one of those has its own set of pros and cons and you can decide on what to choose.
• I would personally rather the position-based model or as Russel mentioned it, the bathtub model. It gives the most value to the first and last touches. The one that initiated the process and the one that closed the sale.
• But we have another model that isn’t predefined and uses machine learning to assign dynamic values to specific touches. It is easy to say that using a machine learning model is the best option, but it is the most expensive one and takes a lot of time and effort, as well as a budget to develop a custom model.
• Russel said something that really touched me and I found it relatable. He said anytime we as humans try to assign values to things, we are using our intuition and our intuitions are highly flawed. We don’t know a lot of things and this causes a lot of repercussions in the future.
• What kind of data do we need for attribution? It is better to ask how much data we need to do such a thing and the answer is a lot. Besides analytics data, we need to have all those events, cookies, javascript, locations, language, and a lot more data.
• When it comes to data gathering and data measurements, we have to consider that we have macro and micro conversions. Although macro conversions are much more important, it is micro-conversions that eventually lead to macro ones. Measuring these conversions are crucial for you to have a successful attribution implementation.
• And don’t forget that you don’t only need data from your own website visitors. You also need some data regarding the events that happen outside of your own territory to have a big picture.
• You need to know what is going on in search engines, what happens when a user sees a display ad, and what type of social ads he/she clicks on. Some of these things can be measured using APIs and some other needs thoughtful inspection and creativity for answers.
• Russel talked about the role of CRO in the attribution, but I didn’t understand what he was trying to say. I accept the fact that if a single landing page converts better, the whole process will work better. But can we assign any value to CRO activities? Is it rational?
• And in the following lesson, he talked about PPC and its relationship with attribution. When we run PPC campaigns, we target different users with different keywords. For example, we bring strangers to our website using generic keywords. But for those who know exactly what they want, we use more specific keywords.
• Each one of these keywords plays a role attribution-wise and should be looked at with a different perspective.
• I feel all of the lessons have a high level of theory attached to them. I am expecting practical and real-world examples and so far I haven’t got any.
• I think these lessons that are centered on tactics are very boring. For example, I just watched a new lesson on SEO and its connection to marketing attribution. We know what SEO is and how invaluable it is in the marketing funnel. But to be honest with you, I didn’t understand what I am supposed to understand from this lesson.
• It might seem a little bit odd at first, but it is the truth. Of course, we optimize specific landing pages for specific keywords, some of which are broader intent-wise and some others are more specific. We drive traffic using them to our website, but should we assign any value to them. Of course, we should, but how should this value be distributed?
• There are a lot of gray areas in these lessons and I think and feel they aren’t complete, even in theory.
• And after that, we get to the display section. Display advertising is part of the attribution, and it is tough to measure. Display advertising is measured by impressions and clicks. But the point is that some of the impressions won’t immediately turn into conversions, or at least won’t convert immediately.
• In Display Advertising, we need to calculate cost per impression and cost per click. But it is very difficult to assign a monetary value to each impression and attribute results to them. As I take it, we need to think about it for our own and find a way that does the job for us.
• Affiliates are a great marketing channel, but measuring them and their attribution to outcomes is quite a headache. We might have a lot of situations in which we have done lots of activities and then the affiliate comes in and steals everything. I am talking about an attribution model based on the last touch.
• We have a lot of affiliates. Some of them are based on discounts and vouchers, but others are educational and focused on brand awareness. Put into considerations which ones are targeting the top of the funnel and which ones are focused on converting to customers. These types are fundamentally different and should be approached differently.
• Emails are great resources for winning back the customers who have already purchased something in the past. Some of the emails are only for confirmation and not designed to be clicked at, like purchase invoices. But others need to be studied in depth.
• For example, we need to know the open rate or click-through rate of our email campaigns. Then we need to do some calculations and find out the monetary values assigned to each action.
• A customer might receive a lot of emails during the journey. Some of them might be more successful than others and attribution is crucial in these situations. We have to know which types of email yield the most beneficiary results and which ones suffer severely.
• If you run tv commercials, it is a best practice to limit its air time and then see the change in your SEO and SEM performance. Does it change your homepage visits or does it hopefully increase your sales? If you do this, you can find out if your activities were fruitful and made an impact.
• Sometimes sending customized direct mail stuff to your existing customers might help your brand and build repeat customers. You can place smart URLs in your print media or a QR code that can be tracked easily.
• You can also assign campaign terms to the URLs you place in your direct mail stuff and follow the performance accurately. Simply send out the stuff to your customers’ address and see if they return in the upcoming days, then you can attribute that traffic to the direct mail channel.
• These printed media that are sent directly to the customers’ addresses can have an impact on your brand perception. For example, your brand search might get a boost on search engines or customers might search “a product category + your brand”. This is the power of the offline world and can be easily related to the online measurable world.
• Russel says that you have to study a lot of customer journeys in order to understand what patterns emerge. Some of the journeys have commonalities that might be precious to you and you can use them to repeat a previous success.
• You can also go a step back to see what was the action before a real conversion and what led to that specific action. In other words, you need to reverse engineer the process and see what can be turned into a repeated success.
• Time is another factor. How much time is passed on average between each step and how can we reduce that time? Is there any way that we can optimize this process from a time perspective?
• Optimizing the number of touchpoints in a journey is another thing that needs to be studied. A 5 touchpoints journey is usually better than a journey having 6 touchpoints.
• Now Russel has decided to talk about strategic terms and concepts because he feels the one who needs to know about attribution needs to know about branding as well. The argument is a valid one and can be taken into consideration, but measuring it is rather hard.
• If you want to be known, some specific metrics should be tracked. But when your main goal is to increase your product propensity, a whole set of different metrics should be looked at. It is a little bit theoretical and I don’t like to open it here. By the way, I don’t think I am pro in describing it either.
• I don’t know why Russel is talking about lifetime value and how it is supposed to be related to marketing attribution (of course it is related and I’m unaware). Anyways, he talked about CAC (customer acquisition cost) and the fact that it should be less than the amount you are going to sell to that customer over the period he is interested in you and your business.
• You might be willing to spend a lot of money to acquire a customer for the first time, but you have to calculate the probability for that customer to rebuy something from you in the future. And of course, this probability should be specified and the period of time in which the customer is a potential customer should also be clear. You should have a number that explicitly says this customer or this group of customers are supposed to be worth this much over the next x months.
• When you are doing attribution projects, you might be tempted to collect personally identifiable information (PII), but it isn’t accepted, legally and morally. Besides that, we need to look at segments of people, not individuals. Studying segments gives us much better insights than individuals because we can find out what type of customers we have.
• But to study individuals isn’t that bad if you want to identify outliers and look at their purchase behavior. They might not be that much, but it will satisfy your hunger for data!
• In attribution, you need to know how much every user is worth to your business, or at least you have to know which one of them is more ready than others to convert. By having access to this kind of data, you can save on budget and time and thus push the qualified one further down the sales funnel.
• Lead scoring systems give you the number of times a user has visited particular pages and then you can send them customized messages to attract them toward making a purchase. You can also adjust your PPC bids for these users and target them with different kinds of messages.
• This was the whole attribution course or what were my takeaways from it.
• This last review became a little bit longer and it was inevitable because it was my last week in the program.
• Thank you for being with me.