De-Duped Success Metrics
When working with SiteCatalyst clients, I often see them ask questions related to how often a particular Success Event takes place at least once during a visit. Examples of this might include the following questions:
- In what percent of visits do visitors add an item to the shopping cart?
- How often to visitors who add items to the cart reach checkout?
- What percent of visits do visitors conduct an onsite search?
At first glance, these seem like easy questions to answer, but I see clients making mistakes with these questions. For example, let’s say that you want to answer the first question above and see the percent of all visits that add items to the shopping cart. Most clients would approach this question by creating a calculated metric that divides Cart Additions (scAdd) by Visits. While this seems logical, it will not give you the correct answer, since visitors can add multiple items to the shopping cart within the visit. If Visitor X adds three items to the cart in the visit, the formula in our calculated metric would be:
The issue is that since most people look at this metric for all visits, the individual Cart Addition numbers are obfuscated and you are often seeing an inflated percentage for Cart Add/Visit %. In fact, the same issue applies to all of the questions listed above. If you are looking to compare Cart Additions to Checkouts, multiple Cart Additions or Checkouts taking place in a visit could inflate your ratio.
So how would you resolve this issue? There are several ways in SiteCatalyst to accurately report on the preceding questions so I will share the various methods at your disposal.
Using De-Duped Success Metrics
The easiest way to resolve the preceding dilemma is to set an additional “de-duped” version of metrics that you want to see in calculated metrics like the ones above. Personally, I wish Adobe provided an easy way in SiteCatalyst to see a de-duped version of every Success Event, but that is not currently available. Therefore, you will have to create a second Success Event for those metrics that you want to use in these types of Calculated Metrics. Keep in mind that you are limited to around one hundred Success Events so you won’t want to do this for all of your Success Events, so use your best judgment.
In this case, let’s assume that you are interested in seeing an accurate percent of visits in which a Cart Addition took place. To do this, every time you set the normal Cart Addition Success Event (scAdd), you should set a second, custom Success Event and call it something like “Cart Adds (De-Duped).” For this second Success Event, you will want to apply Success Event serialization to prevent the event from being counted more than once in a visit. I would recommend using “Once per Visit” serialization since it requires less tagging and can be enabled by ClientCare. By setting this new Success Event, you will have a count of how often visitors add items to the cart, but it will only be counted once, regardless of how many times the visitor adds items to the shopping cart within the visit. When this is complete, you can create a calculated metric that divides this “Cart Adds (De-Duped)” metric by the Visits metric to see an accurate ratio for visits in which at least one Cart Addition took place:
To see the impact of this, let’s imagine that you had five website visitors that performed the following actions:
In this scenario, if we used a calculated metric that used Cart Additions and Visits, our ratio would be 120% for these five visitors. Obviously, this isn’t representative of what really happened. However, if we use our new “Cart Additions De-Duped” Success Event, we will only count one Cart Addition per Visit and see the following data:
Doing this provides a more accurate representation that 60% of visits contained at least one Cart Addition. And since you now have a Calculated Metric that is trustworthy, you can see the answer to this question trended over time using a report like the one shown here:
This Calculated Metric can be easily added to a SiteCatalyst Dashboard and can be used like any other Calculated Metric.
Note: Some companies implement the Carts (scOpen) Success Event at the first shopping cart addition and de-dup it using “Once per Visit” serialization. This is a similar approach, so if you are doing this, you can use the Carts Success Event divided by Visits to see the same cart rate.
As you can see, the addition of one more Success Event allows us to greatly improve our reporting for cases in which you want to see if something happened at least once in a website visit. If you look at the other questions posed above, you will see that the same concept can be applied. For example, if you want to see an accurate ratio of times that visitors do at least one Cart Addition and one Checkout, you might create a “De-Duplicated” version of Cart Additions and Checkouts and use those versions in your Calculated Metric.
Keep in mind that these new “De-Duplicated” metrics will not be accurate when used in Conversion Variable reports (i.e. Products Report, eVars, etc…) since they will only be counted the first time the Success Event takes place. This means that if a visitor adds three products to the shopping cart, only the first product will be associated with a value in the conversion variable (i.e. Product XYZ). These new “De-duplicated” metrics should only be used in global website calculated metrics and the normal metrics (i.e. Cart Additions) should be used in detailed Conversion Variable reports.
If you are averse to using more Success Events to answer the questions above, it is possible to answer them using Segmentation. I think this method is more cumbersome, but will describe how to do it for educational purposes.
To use Segmentation to answer any “how often did X happen in a visit” question, you will have to create a Visit-based segment that isolates visits in which the Success Event in question took place. Using the preceding example, if you wanted to see how often visits contained a Cart Addition, you would create a Visit segment and add the Cart Addition Success Event to the segment as shown here:
Once you have this segment, you can open the Visits report and see how many Visits took place in the desired timeframe. For this example, let’s use the data for the five visitors we described above. In this case, three of the five visits would qualify to be included in the segment so our Visits report would show a total of three. Now you can write that number down and then remove the segment (go back to “All Visits”) and look at the same Visits report for the entire population. In this case, you would see a total of five visits, so you can divide the three Cart Addition visits by the total visits to get the same 60% we saw above. If this is something you will be doing on a recurring basis, you can automate this process using Adobe ReportBuilder. To do this, you would create two different data blocks in Excel – one for all Visits and one for Visits with the above segment applied. Then you can create a formula that divides the totals of these two data blocks and trend it over time using a custom graph.
As I mentioned previously, I think this approach is more time consuming, but it does save Success Events if that is a concern.
Page Name Approach
In theory, there is another way to answer these types of questions, though I don’t recommend it. This approach involves using Page Names. To do this, you can use Adobe ReportBuilder to isolate the specific page on which a Success Event takes place (i.e. Cart Addition page) and look at that pages’ Visit count and divide it by total Visits. However, since page names can be unreliable and it still requires work in Adobe ReportBuilder, I don’t recommend this approach.
If you ever have questions in which people ask you how often something took place at least once in a website visit, I hope that you will think about these concepts and make sure that you are accurately answering them for your organization. While some of these concepts are a bit complex, they can save you the embarrassment of reporting inflated conversion metrics to your organization.