Roll-Up Property Considerations for Google Analytics 360/
November 21, 2017
For Google Analytics 360 users who are considered taking advantage of the special Roll-Up Property feature, extra time should be spent planning what data to combine and how that will impact decision-making and reporting. As a reminder, Roll-Up Properties are special properties that Google will grant GA360 users by request only, so you cannot create them on your own.
Even though there is a Property ID that is formatted like all the other property IDs that you are familiar with, this Property ID is not meant to go anywhere on your site or in a tag manager since it behaves differently. It can be thought of as a master folder that will sync and copy data from your chosen properties automatically (only data moving forward from the roll-up property creation date). It collects copies of the hits from multiple locations but you can still configure familiar property and view settings.
Our previous post on rolling up your data introduced the feature as well as an alternative parallel tracking solution that will work for non-GA360 customers. Since then, more Google Analytics changes, additions and updates have been introduced which affect decisions and strategy for these roll-up properties. With these new opportunities and additional 360 capabilities in mind, it’s time to think beyond basic implementation. This post will be focused solely on the roll-up reporting feature for Google Analytics 360, but several of the broad ideas and concepts here can translate over to other aggregate tracking solutions that you may be using.
To add a regular property to a Roll-Up property, it must be within the same account. This was a disappointment when Roll-Ups were first released but with the newer ability to move properties, there is so much more opportunity to use this feature.
Thinking about why the data should be aggregated and rolled into one property is the first step. Not every organization will need this feature and using it for the sake of novelty will lead to wasted effort and time. I feel like most data questions are frustratingly answered with the same question: “Should we do this/Are we successful/etc.?” followed by the response of, “What’s the context?” … but that’s exactly what I would say in this case. Let’s look at a few examples.
Multiple Revenue Models
Many companies don’t fall neatly into the high-level model description of Business-to-Business (B2B) or Business-to-Consumer (B2C) types. A portion of your websites may be dedicated to one focus while another portion is dedicated to the other; they are completely different audiences in most situations. The marketing strategy will be very different as well as the product type or level, so should they really be analyzed together?
The higher-value purchases and leads of the B2B sites may skew the data when analyzing marketing efforts as well as metrics related to revenue or conversions. Also, there likely would be little user crossover between the two models’ sites. My recommendation would be to roll-up all B2B sites into one roll-up property and roll-up all B2C sites into another.
One potential argument against this is the ability to see if the sites’ content matches the audience by looking at a portion of users that land on one site and convert on another. However, an alternate and simpler solution would be to add an event for links to the other site and analyze segmenting in each property’s view. To take it a step further, you could also link Search Console data to the individual properties to see if users’ searches are matching the appropriate product – this would partially account for those who land on one site and go to the other through future organic searches. If only a small amount of users fall into this category, why waste the additional hits on your GA360 billing account?
To get a broad picture of how both types of sites are performing at a quick glance, reporting with Data Studio or a similar solution would give you overview data in one place.
For mobile websites that are either simply a mobile version of the desktop site or an “m.” subdomain, our recommendation is to use the same property ID as the regular website and always take advantage of segments and mobile-only views. Unless there’s a rare case of the mobile subdomain serving different content and a different user experience, there’s no reason to separate this into a different property, which means you would not need to use the Roll-Up property feature.
Previously, integrating AMP tracking with the Google Analytics Client ID to retain sessions and users across AMP and non-AMP pages was not a default behavior and involved significant effort to code and configure (luckily we have uber-experts Simo Ahava and Dan Wilkerson!). For those who did not use custom code to retain the Client ID, a recommendation was to keep AMP in a separate property so that session and user metrics weren’t inflated in the main property. This relates to the Roll-Up feature if you are in a situation related to the next point where a decision has to be made.
With the new and useful native Google Client ID feature, the integration is much simpler and AMP hits should be sent to the regular property and exist alongside regular website hits. It isn’t a default setting but adding it is as simple as a line of code and a field addition for Google Analytics. For those that have been using AMP for a while and have been sending hits to a separate property, using the Roll-Up feature may be preferable – this would preserve legacy AMP data while taking advantage of the Client ID feature since Roll-Up properties can be configured to unify sessions.
Mobile Applications and Firebase
Mobile applications are downloaded from the Apple Store and Google Play and are not served from a browser – these are tracked differently even if they appear the same and deliver the same content. These should always have their own property and the different versions (iOS/Android) should be separated as well. With Firebase, mobile application data may even be in two places at once – tracking to Google Analytics for Firebase as well as a more familiar mobile application property.
Roll-Up considerations are especially important here. First, Firebase cannot be added to a Roll-Up property. It makes sense because it is a completely different reporting platform. However, the next step is to decide if your mobile application properties should be added into a Roll-Up with your website data. If the content is similar or the same, it would absolutely be worth it to use a Roll-Up and is especially relevant for publishers.
Since Roll-Up properties have custom dimension mapping and calculated metrics, a content ID or something similar can be passed on both platforms to have the ability to analyze overall data. This has the added benefit of cross-device tracking since the User ID feature is available for both web and application tracking.
Here’s another scenario – let’s say you have one application that serves the same or similar content as your website (the above example) but you also have other applications that serve other content or are for other purposes. This is a case where you can take advantage of the ability to add one property to multiple Roll-Ups*. The content app can be added to a Roll-Up with your content website[s] and the same application property can be added to another Roll-Up that houses all of your mobile application properties. Even though the mobile applications have different purposes, an application Roll-Up is useful for high-level performance metrics such as crashes, exceptions and downloads, for example.
*Speaking of multiple Roll-Up properties, roll-up-ception isn’t possible. You can’t add a Roll-Up property to a Roll-Up property (although I can’t think of a use case for that).
Google Search Console
A well-known limitation of the Search Console linking process is that a Search Console site can only be linked to one Google Analytics property. Also, a property can only be linked to one Search Console site. Unfortunately, this means that the link wouldn’t be useful for Roll-Up properties.
The AdWords link is just as simple for Roll-Up properties as it is for individual properties. This should be one of the first things you do when creating a Roll-Up property to start collecting the click and cost data into your account. Not only that, a great part about having AdWords data in a Roll-Up property is the ability to use broader remarketing audiences. This means that you can designate more conditions for your users as they move across multiple sites. Combined with the new Audiences report in Google Analytics, this is even more powerful.
One of the most exciting benefits of having a roll-up property is the ability to link BigQuery to an aggregated view instead of an individual website view. This saves time creating SQL joins across multiple tables and multiple projects. The linking process is identical to a regular property’s view, so there are no additional or more complex steps to go through in order to get things flowing. A potential decision for linking a roll-up view instead of individual properties’ views is that it may affect the amount of backfill for very large sites. When a BigQuery integration is initially linked, 13 months or 10 billion hits (whichever has fewer hits) of historic data are ported over to your project. If you own sites that have over 10 billion hits across less than 13 months, it may be more beneficial to link each property’s main view separately to get the most amount of historic data.
For publishers, DoubleClick for Publishers (DFP) data is invaluable to have alongside Google Analytics data. A Roll-Up linked to your DFP account or accounts would give important metrics like overall revenue across multiple content sites that the organization owns. However, when rolling up mobile application and website data, it’s worth noting that the DFP integration does not currently support mobile app ads, so that data will not show up or be folded into the combined metrics here.
For DoubleClick Campaign Manager and DoubleClick Bid Manager, a Roll-Up property gives the opportunity to analyze broad advertising efforts such as overall brand awareness across your family of sites. Using segments, demographic data and aggregated engagement metrics from Google Analytics could supplement ad optimization.
To link a Roll-Up property to DCM and DFP, the same process of contacting Google to begin the integration needs to be repeated. The connected properties will not automatically populate the linked data in the Roll-Up property. Luckily, one DoubleClick account can be linked to multiple Google Analytics properties for each of these products.
If you are moving properties to make sure they are in the same account for a Roll-Up, a DFP integration would have to be unlinked before moving the property and then re-linked once it is in the new account with the Roll-Up property. This means that there would be a gap in DFP data in Google Analytics. Publishers should be aware before making decisions around these actions.
For DCM and DBM, the link remains intact and they do not need to be unlinked before the property is moved. However, make sure the initial integration happens before the property is moved since the property ID and account ID must match for the link to be set-up.
Data Import and Cost Data Import
When you link a property to a Roll-Up, Data Import settings will not port over. These tables will have to be added to the Roll-Up property as well. Just mapping the custom dimensions will not populate the field.
Since Roll-Up properties will have much more data than individual properties, custom tables should be considered during the strategy and decision-making process. The sampling levels are the same for these properties and if you are using a reporting solution like Tableau, Domo or a spreadsheet add-on which all use the API, custom tables are essential.
In addition, Roll-Up views still have the cardinality limit of 75,000 rows per table so custom tables would be worth it for page dimensions and metrics.
Some events will be the same in multiple websites like outbound link tracking, scroll tracking, contact events and social shares. Others will be more granular and very specific to the individual website. Do all of these events need to be included into the main reporting view in the Roll-Up? Be smart with filters to have a clean view that’s actionable and makes sense of the aggregated data.
Having each site as a content group is such a simple implementation in Google Tag Manager and on-page code. All you would have to do is set the content group to the website name on the pageview. Why do this? Content Groupings are unique because they can be used in the Behavior Flow report. This will give you a (sampled) visual of users’ navigation between sites.
If you are a GA360 customer, consider the possibility Roll-Up properties. If it does make sense to use them, make sure you get the most out of the feature by thinking about account structure, using property settings, and linking to additional data that you have available to you. The new User Explorer report will especially be interesting to analyze multiple touch points as well as the Multi-Channel Funnel reports for multi-session analyses. Roll-Up data also gets around the limitation of not being about to join multiple views’ data in Data Studio, so aggregate data can be visualized and added easily.
Ultimately, the decisions and strategy around rolling up your data will deal with high-level stakeholders and multiple teams. What questions do you have that currently cannot be answered by having segmented data across multiple properties? What part of the user journey are you missing out on between devices or multiple websites? A Roll-Up property could deliver the answers you’re looking for in your Google Analytics data.