The Day After Labour Day, How Did Your eCommerce Promotions Perform?
Some get excited for the holiday shopping, trying to find the best deals on the products they want. For others, specifically those of us in Digital Marketing, it is a little more nerve-wracking of a time. Did our promotions work? Did we reach our target audience? Was our messaging effective? Did we hit sales numbers?
So how do I look at performance? And why do I dislike Google Analytics method of displaying revenue in reports? Let me show you.
It is rare for a customer to be introduced to your brand and purchase something on their very first visit to your site. Customers will find the product, do some research, look around, and then come back to purchase. In the case of a Labor Day sale they might get an email, view the promotion, then later that day purchase.
So how does Google record this revenue? By last non-direct click attribution. Let us assume that a customer visits your site four times before purchasing via a paid ad, social, email, then a direct visit (they go directly to your site via your URL or a bookmark) for a purchase. Google Analytics gives 100% of credit to the last non-direct click throughout its reports. Here is a graphic to better undertsand this
In this case, 100% of the credit for the sale goes to Session 3, email. This doesn't give any credit to the other three sessions in the conversion path and isn't giving you a great view of how your marketing is actually working.
Luckily, Google Analytics does have attribution modeling where you can change the model itself. There are a couple standard models or you can even create your own. You can choose from time decay, equivalent (flat) attribution, position based, and others.
My favorite? Position based.
Now, this looks better! We are accounting for all four sessions in a logical fashion. This model gives 40% of credit to the first session, 40% to the last session, and divides the remainder between the middle sessions. You can adjust the distribution but in general, you want to give most of your credit to the paths that found your customer and also pushed them to purchase. This model does a great job of showing that.
You can also take it one step further. If you have been properly using UTM parameters you can further segment out your model by campaign name, source, medium, term, or content. This enables you to answer questions like which Facebook audience drove the most sales, what email was most effective, or what version of your banner ad performed the best.
Attribution modeling and segmentation moves us away from (mostly) meaningless metrics like clicks and towards ones that actually matter, sales & conversions.
Questions? Thoughts? Leave a comment.