GA 2.1: Mistakes in Your Google Analytics Setup That Can Trash Your Data
- How Mistakes in Google Analytics Setups Can Ruin Your Data
- Accounts, Properties, and Views in Google Analytics
- Importance of Cross-domain Tracking
- Making Sure You Set Goals For Your GA Reporting Views
- Verify Your Filters in Test View First
- Exclude Staging Environments From Filters
- Exclude Query Parameters from Filters
- Why You Need Event Tracking
- Building Remarketing Lists from Segments
- Usefulness of Custom Dimensions & Metrics
- Tackling Spam & Undesirable Traffic in GA
Aiden: Welcome back and happy new year. Jill and I are back now for series two, with SEMrush of course, around all things analytics once again.
We're going to be talking about how Google Analytics can turn into a little bit of what we would call a dumpster fire if you don't set it up properly. We're talking about the human mistakes that make this particular tool, from a technological perspective, not work all that well.
A massive hello to all of you. Now, Jill, you've seen, over the hundreds of audits that have kind of crossed our various different desks, mistakes ad infinitum at Google Analytics. I think we agreed that there was one account that we've seen that was nearly correct but wasn't quite. As for the rest, not so much. Can you tell us a bit about that?
Jill: Yeah. With Google Analytics, because it's a computer program, you can't always assume that the people that created the account for you have done it correctly, or they understood your business well enough to track the things that matter.
When we've done audits, when we've trained people in Google Analytics, the same mistakes happen, and that's the reason for this particular webinar.
Aiden: I'm going to ask the lovely Jill to kick off her extremely interesting talk about some of those mistakes and indeed what you can do about them. Jill, quick, take it away.
Jill: Thank you very much. Okay, let's do the screen share.
How Mistakes in Google Analytics Setups Can Ruin Your Data
Right. Mistakes in your GA that can trash your data.
Now, why should you care about your GA setup? I'm going to tell you a bit of a story. Some of you may know this story already. The reason that we called our company The Coloring-In Department, is because people had used that almost as a slur, as a mean thing to say to me when I was in the office.
When the budget was getting cut, or things were being moved, my department, my budget was the first one to go, and I'd have people say to me, "Well, we know marketing does stuff, but we don't always know what the impact is."
That's when Google Analytics can really help you, because you're reporting to people, and you don't want them shouting at you saying, "I gave you this money and what did I get from it?" You want to be able to turn around and give them a number and show them some data.
Going back to my favorite quote from our data scientist, "Without data, you are just another person with an opinion." You want to make sure that you can validate what you're doing. On that thought, you also need to make sure that the data that you are providing is correct, because otherwise you will lose your credibility and people will shout at you, and nobody wants that.
Crap in, crap out. One of the things with any program, including Google Analytics, is that it's as good as the settings that you have. If you have a bad configuration, if you haven't turned a certain toggle on, then the data that's inside those reports, those reporting views that you're using to give information to your boss, your board, your client, whatever. You've got to make sure that the data that you're presenting is actually valid data.
This messy door, this crap that can be in the back of your admin settings, I'm talking about this, the account settings. You arrive here by looking at the little left-hand cog, which will show your analytics account setup.
Accounts, Properties, and Views in Google Analytics
One of the mistakes that we're going to talk about, is actually getting your house in order to make sure that you have the right account, properties, and views. We did this in the first webinar in series one, where we talked about analytics being a little bit like a house. You have your account level, and I want you to think of this as the roof of your website, your little house that you have. You'll get given a UA number, so hypothetically UA1234.
You then have a property, and I want you to think of these properties as a floor, an actual living floor that your visitors are going to come in and sit down on your sofa and do stuff, like buy stuff from you, send a lead form, whatever. You've got different reporting windows, which are your view settings. Then to make those windows work you use filters, which is like adding drapes and blinds.
If you have a basic website, so you've got like our website, thecoloringindepartment.com. If you just have a single website, you have essentially a bungalow. You'd have a roof setting, UA12345. You have a property, so it'll be UA12345-1. You should have four windows, four reporting views.
Google recommends that you have a raw window, so if I have a look through this window I see absolutely everything. You have a test view, this is where you play around and check if something's going to work or not. You check a filter. You check a goal. You check your settings and make sure that you don't flatline your data. Then when you're happy with it, you then create your reporting view, which is the one that you go in and use to validate your decisions and your strategies.
Now anything that we do to our settings if I start messing around with my roof if I whack a massive hole in it, it's going to impact the floors beneath it. If I do something to a window, by adding a filter, it only changes the data to that particular window. You've got to make sure you go into all of your settings and audit them, and not assume that if you do something at one end of the house that something's going to magically happen at the other.
Importance of Cross-domain Tracking
Now if you have a multiple brand scenario, so you have different websites that sit independently from each other. You have a website, maybe you have a separate business, maybe you have an app, but they all sit separate to each other, you've essentially got a flat or a multi-story building. They all sit under the same roof, so the company name I'm going to make up a company named JillLikesGin.com.
I've got several floors for different things that I'm selling because I've decided I don't want to do gin anymore. I want to sell vodka, or I want to sell ice cream or whatever. If they sit separately to each other, they're separate floors with their own little reporting views, and I can see what is going on for those particular properties.
Now the mistake that I see more commonly these days is something where you have multiple domains for a website, but they've built as if they were a flat. I see this a lot in eCommerce sites at the moment. Their domains will change, so you might have website.com where you're sending all of your beautiful visitors to go and have a look at your products and services.
They decide they want to buy it, so they click on a button and they go to shop.website.com. The website still looks the same. The design still looks the same, but the URL has changed. Then when I've decided that I'm going to buy from the shop and I'm going to create an account, another domain pops up called account.coloringin, or whatever it is that you've got.
Now, the mistake that can really ruin your data is if you set up a multi-domain website where the ecosystem is actually within the same business, and you've set them up as separate floors, you are unable then to track the customer journey. Because when I log into my analytics, like a floor, an actual physical floor, I'm on floor four of our building, and Aiden's on floor seven. I can't see what's going on on the floors above me. I can only see what's going on on this particular floor that I'm on at the moment.
If you have a goal for somebody creating an account, or somebody buying something, if I'm in the reporting view for coloringindepartment.com, but the goal happened on a different floor and you haven't tied that all together with something called cross-domain tracking, then you are not able to build your goals. You're not able to see the journey. You can't see what marketing brought people into that particular sale, so everybody's having a bit of a hard time.
The first thing that you need to think about when it comes to your analytics setup... double check that your roof and your floor and your windows are correct. Double check that if you do have an ecosystem where you've got several domains but they all really need to be packed into one, have you done cross-domain tracking so that you're essentially treating all of these different domains as one.
Making Sure You Set Goals For Your GA Reporting Views
The next mistake that we see a lot, which seems really obvious but they're interestingly absent in so, so many audits that I've done, is not having any goals on any of your reporting views. Sometimes when I've done an audit where there are several windows, there are goals on one, and not on the other. I've had so many people come back to me going, "Oh, I thought if I created it in this view they would be created magically in the other." I'm like, no, it's a computer program. You're going to have to go in and build that goal another time. If you've got 17 views, then you build that goal 17 times.
For any business, you want to be tracking a significant profitable customer interaction. Now, that can be I signed up to your newsletter, I scrolled down to the bottom of the page, I submitted a contact form, I bought the damn product.
Whatever it is that you're doing, you have 20 goals available to you per reporting view. 20. What we normally see are either one or two big goals, so these are your macro goals. Think of these as, if these don't happen, then I'm out of a business. Then your micro goals are the smaller meaningful interactions as people edge ever so close to doing what you want them to do.
Imagine having 20 goals, and being able to understand the role of a particular marketing channel. You might see that Facebook is really good at getting people to watch a video, but they're not going to convert. Like that was the early awareness stage of the journey, but email was really good at getting them to start a free trial.
By doing this you're able to give the managed expectations of your marketing channels and how they're actually performing and the role that they have. Not every single channel in every single campaign is going to be the person that scores the goal, but they have a role. If you have 20 goals, your micro and macro conversions, it gets a lot easier to actually present that information to people.
Verify Your Filters in Test View First
Now onto one of my favorite topics, filters. When we are talking about our reporting views, as I mentioned, we have our little windows where I can look in and see what all of my lovely users are doing and if they're engaging with my content, if they are buying stuff. Filters have probably been one of the main culprits of the data just being god awful because they are quite powerful. If you haven't done it in a test view, and you haven't checked out that they were working, it is very easy to flatline your data.
I'm going to give you some examples of where this has happened. This is an example of a reporting view that was created. Now, this view was meant to only show people from the United States of America. Now, when I went into the reporting view, and I went into audience, I think we can all see quite quickly here, because of the gorgeous heat map, it's not just showing people from the United States of America.
This company was saying, “Right, we had 52,000 users in America this month.” That's wrong because actually, only 48% of this data was from America, the rest came from different countries. When they said to me, “Jill, we have a problem with our data. There's a data discrepancy. We don't understand what's gone wrong.” I was like, let's have a look at those filters. Let's see what's happened.
Essentially they had the right idea but the wrong filter. Now if they did this in the test view, they would have gone, "Hang on a second there Bob. You can see that the traffic is clearly not just showing from a particular country. Let's go back and see what we've done and make some changes." They didn't do that. They didn't do any of the test views, which would have fixed this particular obvious problem.
Another tip here, you have to write the country exactly as it shows in the Google Admin settings. If they put in France with a capital F, then you have to put the filter pattern that exactly matches what is in the Google Analytics settings. Because this is a computer program, there is no empathy here. They won't look at this and go, "Oh you put the USA there. Did you mean the United States?" No. It's going to just flatline your data and say, "Computer says no." You've got to have a look at your filters and double check that it's giving you the information that you want.
That was a nice easy one to get us started. The next set of filter mistakes will show you how you can have real errors in your revenue or your page views, or just generally the data's not making any sense.
Exclude Staging Environments From Filters
The first mistake that I see with filters is to not exclude your staging environments. This happens quite a lot actually. There was one audit I did for a global company last year, where they had 20 different staging domains, and they didn't put a filter to say, "Exclude staging.website.com, dev.website.com."
You've got to remember when it comes to conversion rates that Google is calculating those conversion rates based on your sessions. If some of the sessions are your dev teams, or your agencies working on stuff, you're diluting your conversion rate. You might be getting kicked for a conversion rate that looks really bad, and in truth, it's actually not bad.
You can still, for the reverse, have a view that says, “Only include traffic from staging.website.com.” You can see what's going on, on that staging. That's fine, but you do not want to include this in your reports, because it's just going to give you those data discrepancies.
If you have got any staging environments, you want to go into your view settings for each view that you want to do this in. Do it in your test view first, check that it works. You want to go in and say, "Hey Google, please exclude the data from dev.whatever the website name is, or staging.websitename." Do it in your test view, check that it works, and then roll it over into the reporting views. Then the data is going to be a little bit cleaner, and life will be a little bit better as a result.
Exclude Query Parameters from Filters
Another problem with filters can be down to your content report. When we are going into our behavior reports and we want to understand what people are doing, we can go into our all pages report. This was from a client that we did some work for, and they said, "Our content is amazing. We have 11,416 page views. Go, team." We went, "That's fabulous. That's wonderful." They didn't set up their site search correctly, in terms of adding a filter to exclude the query path.
Now you may or may not know this, but Google will record the URL at the time of a user's session. For this particular website, they had a lot of people using their site search, and I mean a lot. Every time somebody went onto the little search bar and said, "I am looking for insert keyword." Whatever the keyword is for the content that they were looking, that URL, where it pulls in the query parameters included in that particular search string, gets recorded as a page in your analytics.
When we added a filter to exclude query parameters, those 11,000 and odd pages actually got stripped down to about 3000 pages. That was an uncomfortable conversation to have with them, to say, "Actually people aren't reading your content. They're not looking at your pages. You have a usability problem. People can't find the information, and they are resulting in your search features to try and help them find that information."
Why You Need Event Tracking
One of the other mistakes that we've seen in accounts is not using event tracking. Now event tracking, for me, once you've got your house set up and you're tracking things correctly in your acquisition reports, is to actually understand what people are doing on your website.
Do you know what people are actually doing on your website? Who's scrolling down the page? Who's clicking on images? Who's printing a page, or downloading a PDF, or playing a video, or clicking on an email address?
This does not come pre-baked into Google Analytics. The reason why it doesn't come pre-baked in is because you've got like a million, million, million websites using analytics. Google doesn't know what you want to track. They don't know what you want to call it, so you have to do this. You have to set up events.
Now the mistake that I'm going to focus on here is just getting the brief wrong from the offset. You need to think about the things that you want to track, and you have to talk to the computer program the way that it wants to be talked to. They want to have category, actions, and labels. Think of your categories as big, broad buckets for you to organize the things that people could be doing on your website. The action describes the doing, what is it they actually did. The label is going to further describe that action.
You've got to go into your event tracking, which will be found in your behavior reports, and I want you to audit what you currently have. This is how not to do it. I did an audit for an eCommerce company, and they said, "We've got event tracking, Jill." I was like, "Fabulous. Let's have a look at what you've got."
We had one bucket, one bucket, for eCommerce. I'm like, all right, let's have a look what's in here. Two actions associated with a category, non-interaction, and interaction. I was like, okay, I'm going to open door number two, to see what the interactions were labeled, and they were all labeled to not set. That doesn't help anybody. You've basically just wasted a load of opportunity here.
When you go in and audit your analytics, which is the fix, you have to go through your pages, your home page, your money pages. Your money pages being the pages that you want people to actually do something. Fill out a form, buy the shoes, buy the bottle of gin, book the holiday, whatever.
You want to go through it and make sure that you have independent categories, and that the actions associated with those categories don't overlap. This is one of the issues that I've seen with event tracking. I might have somebody have an action, so this is the doing, and they've called it click. That's associated with one or more categories. When you look into the report, you've overinflated a potential action because you've named three different things the same thing, if that makes sense. It's just really heartbreaking when that happens.
You want to make sure that you don't duplicate any of the naming conventions for your categories, actions, and labels. Because when it comes to building segments, or it comes to digging into this data if I've got three different separate categories. Let's say it was playing a video, or downloading a PDF, or clicking on an email address, and all the actions for those categories were labeled as a click. When I go into my analytics and say, "Hey, show me all the people that watched the video." It's going to count all the other things as well, and that's going to give you a false number, which you don't want to have.
Building Remarketing Lists from Segments
Now a tip here, because I'm aware that we're speaking to people that are going to be using a tool like SEMrush for their ICO and their PPC. If you have event tracking on your Google Analytics account, which I strongly suggest you do, because you're going to need them to build goals, you need them to build segments. I want you to start building remarketing lists based off of what people nearly did.
Imagine having events firing correctly where somebody goes on a webpage, scrolls down the page, downloads a PDF, plays a video, half filled out the form and added to the basket but didn't buy. Build a segment, and if you've linked up your AdWords to your Google Analytics, and again we go through this in the first webinar that we did in the series, then you can click on that little button called actions and build an audience.
That's going to build a remarketing list that will sit in your AdWords, which means you can remarket to the people that nearly bought. They would be my first go to people that I would do any remarketing on. Then you can drill down further as you so wish. There's a nice little bonus point for you there.
As you can see with a lot of these mistakes in analytics, it's just down to not fully understanding the implications of the features that you have. Namely, a bad account set up in the first instance. You don't have the right reporting views. You have a flaw, or several flaws, for different sub-domains when you should have done the cross-domain tracking. You don't test anything, so you don't see any filters that are making mistakes. If you are doing filters, you want to make sure that you're at the very least removing your staff. You want to remove any staging environments. If you're going to be looking at particular markets, then you want to make sure that you have the correct filter to isolate and only see those particular users.
Usefulness of Custom Dimensions & Metrics
I now want to touch on a mistake which is just a mistake by not using them. Custom dimensions, and metrics, and data import. My main point here is that your business is very unique. The insights that you want are also very unique.
Now, creating a custom dimension in your property settings will take you minutes. It will take dev a lot longer to actually get this working for you. The thought process, of thinking about what else can help my data analysis?
Let's go through these one by one. In Google Analytics there is no dimension for a refund. If you are selling stuff, you can go, right, this month I sold 20 pairs of shoes and I made 1000 pounds. If somebody then decides that they don't want to keep the shoes anymore and they send them back to your company, Google Analytics doesn't know that that happened.
For me, I want to understand, how much did I sell? Then what got refunded? Then hopefully tie this together so I can understand the marketing channels. The way that we're going to stitch this together is with your eCommerce data and your product data. Provided you've set that up correctly, you can stitch that together by creating a custom dimension and a metric where you can punch in that refund data. You can either load this up manually through your property settings, or you can just pump it in through your API. When you've got this, it's so, so cool.
When we are looking at things like content, so if you are writing a lot of content, then it would be useful to know if I've got a lot of people on staff writing, who gets the sale? Looking at things like the author, because that doesn't exist as a dimension in analytics. I can tie that to the page URL, so say, "Hey Google, this is the URL, and this is the person that wrote it, and this is the category."
For anything, if you can build segments, you want to build a segment. You might say, "Show me everybody where the monthly recurring revenue is like 200 pounds." If your average order value is like 50, who are your best users? Show me all of my users that are the B2B, or the B2C if you're a two-sided marketplace. You're able to split out that data. You're able to split out those different users.
Again, as I mentioned, if you can build a segment, you can build a remarketing list. You can say, "Hey, show me all the users that are the buyers." They haven't bought anything for a while, so let's send them a message around the web, and follow them, and give them a little message to say, "Hey, remember us? Come back."
Tackling Spam & Undesirable Traffic in GA
Aiden: Jill, there seems to be a lot of chatter around, well actually how do we exclude bots? Is this an issue? What about spam? What if we're getting traffic from sources that perhaps are undesirable? Things that we might have disallowed, say for instance in search console, are still coming into our Google Analytics reports. How might we address these two relatively synergistic issues, in your opinion?
Jill: Yeah. That's a very good question. Spam, we'll tackle that one first. Fake websites sending you traffic, but they're not actually real visitors. Those visitors are showing up for sessions, and with them showing up for sessions they're screwing with your data. If you go into your acquisition reports, and you drill into the referrals, so websites sending you traffic. Then I add a secondary dimension to say source, so where does the link live? You can start to find particular users and visitors where the bounce rate is 100% and time on site is set to zero. You'll see things like freesocialsharebuttons.com. You start to pick out common culprits that are basically spamming your site.
If I find a couple of different domains that are rubbish, I will go through the process again, where I'll go, "Right, let's have a look at all of these different links." I add a secondary dimension to show the hostname.
If you find that you've got several domains that are all spam, but there's a mother ship where the host is faketraffic.abc.com, I will build a filter to say, "Exclude all traffic from the hostname, the mothership."
Also, in your view settings, there is a little tick box that doesn't get ticked by default, which is exclude all known bots and spiders. You want to make sure in your view settings that you tick that little bad boy to make sure that it excludes Google's kind of nice list.
Aiden: A couple of other things came up. Cross-domain tracking is a bit of a repetitive one. That one came in on Twitter actually.
Cross-domain tracking, thinking about connecting up our various different mother ships, so to speak. Are there issues with that that we need to be aware of? Is the UA code involved in some way? What would you say in a nutshell in making sure cross-domain tracking is set up properly?
Jill: You need to understand which domains do you want to track as a whole ecosystem. We'll take our example. We've got Coloring In Department, and then when you do our online courses it'll be like teach.coloringin. We want to stitch those two websites together essentially, those two different domains. I want to see the full ecosystem.
Instead of having several floors, where I've got different properties with different property numbers, I'm going to say, “Hey Google, we're going to treat each of these different floors as if it's one really big floor.” I'm going to use the UA1234-1 on Coloring In Department and teach.coloringindepartment.
If you know that you've got an ecosystem that needs to be tracked, then cross-domain tracking is what you need to have set up.
Aiden: Thank you, Jill. Thank you, everybody, for watching this wonderful webinar on all things GA and mistakes there. Hopefully, you are less terrified, or you feel a little bit more able to troubleshoot maybe some of your own account issues or challenges that you might have. We are 100% out of time now. However, we will be back on February the 28th, as I'm sure many of you are already aware, at 5:00 PM GMT, so that's London time.