Advanced Analysis in Google Analytics 360: Part IV – Finding Insights with Advanced Analysis
Now that we know how Advanced Analysis works (as discussed in article parts one through three), how can it be used to answer business questions?
Let’s say that on checking our weekly dashboard in Analytics or Data Studio, we noticed that our conversion rate was down as compared to the previous month:
We want to investigate this and figure out the reasons behind the decline. In the Advanced Analysis Hub (homepage) I click on the Blank template, and this takes me to the Analysis Editor.
I select the date ranges for the analysis and enable the toggle Compare to which has the Previous period value.
The Users metric is the only one applied to the report. I decide to test my hypothesis that a specific traffic channel impacted the overall conversion rate. I add Medium to the Dimensions selector in the Variables panel and then drag it to Rows in the Tab Settings panel. Medium shows up in the table. Similarly, I’ll add Sessions, Transactions, and Ecommerce Conversion Rate into the analysis canvas.
Now we can see how mediums contributed to the CVR decrease. Since Ecommerce Conversion Rate is a calculated metric, we want to pay attention to Sessions and Transactions in case a particular channel demonstrated a lift in low-quality sessions or, alternatively, dropped in conversions. I change the Cell Type in the Values selector to Heat Map for a better representation of the areas we need to focus on:
Based on initial findings from the data table, we can isolate some traffic segments and figure out what factor impacted performance. For example, I right-click on the organic row and select Include Only Selection. The filter including only organic traffic is applied to the table.
Furthermore, we want to know if this only happened to traffic from a particular search engine and apply the Source dimension to the view. Having isolated the source we’re interested in, we can easily drag other dimensions directly into the data table, such as a landing page, to determine if a particular high-converting site page received fewer visits in the reporting period or if it just started converting worse. In the latter case, we’ll also have to check whether the volume of search decreased or if there were changes in site ranking. For example, we may see a fewer number of users visiting a landing page with promo codes and, therefore, a lower conversion rate.
On the other hand, other landing pages may become discoverable in Google search results and are getting more sessions, though still require optimization from the performance standpoint. Adding the Bounce Rate metric will help us see whether users find these landing pages engaging or not. Thus, the data exploration capability may reveal some useful insights for our SEO and UX teams.
We also noticed that another channel driving the WOW decrease was Direct. Since this is traffic Google doesn’t have much information about, we have limited capabilities for digging and the Exploration feature will help us discover some immediate insights, if any.
Let’s take a detailed look at the Direct traffic. Clicking on the right-click menu in the corresponding row will show “(direct) / (none)” as a traffic filter. Then change, or add, the Country dimension to see the locations where the surge of traffic arrived from. The United States leads in traffic and is mostly responsible for the increase. Then, upon seeing that the US was driving additional numbers of low-converting sessions, we may want to slice the data by additionally applying the Region dimension. There could be an increasing interest to our site from a particular state where we may not provide delivery or where our stores are located, thus telling us these users are likely to purchase from a store. If the distribution of the lift in users seems equal across the US, we’ll continue exploring the data by adding other dimensions such as: Device Type, Browser, Network, Service Provider, Operating System, and Hostname.
It may appear that the increase in users were registered on iOS. If so, we’ll want to check if we recently launched new campaigns targeting iOS web users, and whether they have been tagged correctly and/or why the referrer information is missing. Other scenarios may include test or development traffic streaming into the production property or a segment of users from outdated browser versions which can be classified as bot traffic.
Then, we might create a segment from the (direct) / (none) row filtered by country by selecting this option in the context menu. The segment will automatically include all users from the US with the Direct source. The next step would be creating an analysis of the Segment Overlap type in a new tab.
Now we’d like to know if the spike in direct visitors was impacted by other channels. We have to create and export segments to the analysis canvas (e.g. Display, CM / DV360 and Paid Search, Google Ads) as in the above example. This will give us an understanding of whether some of those users arriving directly have ever been touched by Display or Paid Search.
Changing the date range to compare to the previous period, gives a full picture of how the overlap between channels has changed WOW:
Hovering over the intersection of the two segments (Direct from USA and Display) will show the week-over-week change. The two percent change doesn’t seem to explain the increase in the above case, so we can conclude that Display didn’t drive the surge. In any case, we can create a segment off this overlap and explore it additionally by checking what advertisers, campaigns or sites are driving direct visits.
Finally, we may want to identify the conversion funnel steps where users tend to drop-off as compared to the previous week, resulting in a conversion rate decline. Create a new Funnel tab and drag Organic Traffic to the Segments comparison control. Now we are ready to create a funnel by clicking the pencil icon next to Steps. If your website Analytics tracking includes an accurate Enhanced Ecommerce implementation, you can take advantage of it by specifying the funnel steps as Shopping Stages – from PRODUCT_VIEW, ADD_TO_CART, CHECKOUT_1 etc., down to TRANSACTION. The All Sessions step needs to be created independently.
Looking at the graph and data table allows us to reveal actionable insights such as the biggest drop of users occurs after the Product Detail step. With this knowledge you may consider optimizing this step to provide a better user experience or to retarget this audience with a promotion or deal for the product that was viewed. Building an audience to be used in both scenarios is available directly in the context menu.
You can also check how the Completion Rate has changed WOW, to get a better understanding of why this traffic’s conversion rate has decreased. In the above example, we notice that the biggest decline in completion rate occurred in the Cart step, meaning that users were less likely to proceed to checkout. You may re-market to these users with a promo persuading them to complete a purchase or explore the Cart step users in depth to determine if a higher completion rate in the previous week was related to visiting a particular page – which is doable with the Segment Overlap functionality.
Troubleshooting and identifying the how and why behind data fluctuations become easier with Advanced Analysis. The three techniques work together to address an analytics request or solve a business problem – from digging into data with Exploration to visualizing user journeys with Funnel. Not only does Advanced Analysis extend the existing analytical capabilities but also provides an intuitive and flexible way to interact with the data. Having this powerful tool, marketers and analysts are getting closer to resolving one of the main Analytics’ challenges – building out a complete and centralized view of the customer journey.