The Outlook
Perspective: Jonathan Smalley

Digital Product and Event Tracking

So, what is “digital product?” I think that definition has grown and evolved over time. Traditionally, it's thought of as what's your direct-to-consumer site, your digital storefront if you will. I think over time, we're seeing that evolve a little bit where it can be a mobile application or an in-app experience. Depending on the platform where that product lives, you'll have varying degrees of control over that digital product and experience.

That's where I think based on the level of control you have, on your digital, direct-to-consumer site, you have ultimate control and can really customize anything depending on your level of technical ability, all the way down into being able to check out through something like Instagram, where you have significantly less control; that leads you to think about how to do semantic event tracking, which is the best way of understanding how users are behaving.

Digital Product Behavior

The best place to start is on digital products that you own. A DTC e-Commerce site can be just as applicable on a mobile application, on a mobile site, and on any digital product that you completely own the experience of. With that in mind, the reason why these things are so important is because that's where you interact with your customer. In the same way that you want to have highly trained associates in a physical retail store giving the best possible experience, so that you know, dopamine is releasing in people's brains, you wouldn't have the same types of high-quality, branded experiences through your digital product in order to transact people, in order to create brand affinity, and all those different sorts of things.

Zooming in more on event tracking, it's the best way of understanding how consumers are behaving on that site and how you're interacting with your customers. All of that ultimately drives down towards the point of transaction. The point of anything, if you're a retail company, is to transact someone. Traditionally when people think about digital products, that gold standard metric is generally conversion rate. What percentage of traffic coming to our product is ultimately purchasing and transacting?


As e-commerce has evolved, we have to think about those things, plus conversions, a bit differently, particularly for high product value items. We work with a number of brands that are selling high hundreds to thousands of dollars, and conversion rate can become quite a lagging indicator as people might have a longer consideration cycle. (traditionally, when you look at the first wave of digitally native brands like Warby Parker and Glossier, we think about conversion being a good, leading indicator for the quality of traffic that we're sending as we're bringing new and different types of products online).

I think conversion rate is a great place to start– it needs to be thought about in quite a different manner depending on the type of your business. That's where event tracking can become both important and complex to think about. Eventually, it can be instrumented in a number of different ways. The most ubiquitous method is through a Google Analytics pixel that you would set up on your site, but there's also other platforms to do that. Segment is probably the best in terms of deep instrumentation and being able to track every customer movement. Regardless of what platform you use for semantic event tracking, the key touchpoints that ultimately lead to transaction starts with two bookends, the first being what is the total unique traffic, the unique users that are coming to your site from different platforms. Over time, as businesses become more sophisticated, you can get more in-depth and more complex and think about how individually you define something, like unique users, for your business.

For most of our unique users,  Google gives a really great way of looking at overarching traffic. Something like Segment might define a unique user as a user whose session time is greater than an actor, or if they've viewed a certain number of page views, and a user as just any unique individual who comes to your site within a time defined duration. Google does a pretty nice job of doing that right out of the box for you.

That's where everything starts: how many people are coming to our site, and then from there, everything cascades to that other side of the bookends, which is conversions. I think with those two bookends in mind, then the question becomes what other actions are leading indicators for customers transacting?

Object Action Framework and the Funnel

Now for some context: what are we selling, and what's the value? If we're selling something pretty complex, like furniture, for instance, we're going to look for different behaviors than someone who's just selling clothing.

Similarly, if we're selling a subscription, we will want to have that instrumented as well. Can we see the number of users who took an action of like switching from a one-off purchase to a subscription purchase, but then that ended up doing a one-off purchase? What percent of people end up selecting a subscription purchase?

Taking a step back, it's good to explain semantic event tracking. Basically, it’s using some type of tags, whether that's Google tag manager or Segment, in order to track the actions that a customer takes on your site. Segment has done a really good job of creating great formats around what they call item action, which is basically product add to cart. It's whatever that item is, plus the action that's taken, like product detail page (PDP), view page, or viewed.

The object action framework is a really nice way of thinking about semantic events. To understand the behavior of how someone who views blog posts, then views product detail pages then maybe adds to cart; a good framework to think about is object action framework. 

Thinking about event tracking, ultimately it comes down to understanding how customers move through the funnel of transacting.

For me, I always start at a high level of what's our total unique traffic. The next step is what percentage/absolute value/how many people viewed a product detail page. While there are other actions you could instrument against to show intent towards a product; an awful lot of sites have a quick view, or you can purchase or add to cart directly from a quick view; still the average customer is going to view a product detail page before they transact, especially a first time customer. 

Generally, as we think about moving down a funnel, I think about unique traffic then into what percent of those PDP pages. The next part of the funnel that I generally look at is the PDP view to add to cart rate. So, a couple of things there to think through; what's the size of the catalog that we have? A business that has one to maybe five products should be looking for a much higher PDP to add to cart rate. Then, take J. Crew that has probably tens of thousands of products that you could view at any time, I think that also then goes into understanding what is the average when you look at carts, what's the average number of products added to cart? I think for custom companies with quite a broad product category, this can be really helpful to understand the intent of customers when they use a cart: are they looking to add a number of products to the cart and use it almost like a wishlist? I think with apparel, cosmetics, or even home goods, if someone is looking to build out a room, are they only adding to cart when they're going to transact?

A good metric to look for there is basket size: pre-transaction basket size, post transactions, or units per transaction. Then, as we marched down that funnel, other helpful pieces to look at are what's add to cart to checkout-started. So that will show us our average intent of a customer when they add to cart, are they looking to transact or just build this wishlist? The last piece of that is checkout initiated to actual transaction completed. I think if we have that users to PDP view PDP to add to cart rate, add to cart to transaction started, or checkout started, and then checkout started to actually convert or transact, that funnel regardless of your business gives some really nice baselines to understand how A) the intent of our average customer, and       B) the efficiency of our site. If we're seeing a lot of customers start checkouts but not complete them, there's almost certainly something up with our checkout experience and we should spend time optimizing there.

Whereas say we're a company that only has five products and very few people are adding anything to our cart. That means, A), we're marketing to the wrong people, or  B) our PDP is bad and we needed to fix it. Or, we're just selling a bad product that no one wants to buy.

By breaking it down into those different pieces, we can pretty quickly have an idea of the performance of our digital products and the intent of our customers. 

The Modern Levers of Digital Product

In regards to the new eCom climate, in general, you have much higher intent users now than maybe you did a year ago. Maybe a year ago, someone's going to come to your site, they're going to see something they like, and then either if you have a physical retail store or if, there's someone who sells your product through wholesale, or if you have a competitor in the area, you don't have a physical retail store that, you know, in the past, maybe a customer is going to look at that, they're going to do some research online, but then they want to go and have some type of physical interaction with the brand or the product.  Now, people are less interested in going to stores or a store might just be closed here might not be options around them, so the overall intent has arisen.

I think when you look at things like year-over-year conversion rate, everyone should probably expect to see higher PDP to add to cart rates, higher total conversion rate then they would have seen last year. So, that's certainly in some ways a benefit.

That's something that as consumer brands you'll have to be mindful of: if and when things get back to normal, I think there's a lot of behaviors that are being learned and changed and probably kept for the long term. Maybe before, someone wanted to buy shoes or furniture or something like that in person, because they wanted to have that physical experience. I think we should all expect to see that a lot of those behaviors stick around even whenever this is over, whenever that might be, especially since it's pretty clear at this point that it's not going to be any time in the very near future. 

Then, I think some of the other factors that I think I would bring into consideration when looking at those metrics based on what events have been instrumented– I think first, what is the industry? Going along with that, what's the price point inside of the industry? (Of course, if you're a more luxury product, you should expect to see a lower conversion rate than someone else in this space). Also, what's your average product value? The higher the product value is, the less likely you are to have a number of units per transaction– the least likely you are to have a higher conversion rate. What's the size and the breadth of your product catalog, and are there any natural cross-sells? For instance, if you're selling an air purifier and you have a filter that goes along with it and you start seeing an increase in units per transaction, that might not be because people are like loving your air purifier or wanting to buy more of them, it might be because marketing is doing a discount against buying an extra filter with a purifier or something of that nature.

So I think it's always important to understand the specific nuances of the business. Keep those in mind as you're instrumenting against those things. Our feeling here is always the crawl walk, run, today, almost everything we talked about is like the crawl version of digital product and semantic events. I think until you have some of those foundational things done, it's awfully hard to think of leveraging deeper event tracking until you understand where there other opportunities to pull different levers.

The Future of Retail 

I think it ultimately comes back to this fundamental shift away from at least traditional physical retail, particularly right now: can we talk to the right person with the right product at the right time inside of the right context? That is the future of all retail. 

In this case, we're mostly talking about digital branded commerce. The number of locations where that exists I think is only going to increase. A really simple example of this is if you have the right event tracking done and you've set good baselines and you have a grasp on how the nuances of your business might impact those things, then that means our team can worry less than maybe about add creative if we see that people are dropping off on the PDP data cart page, and if we've instrumented a little bit more deeply, we might see that any customer who actually scrolls down the PDP page and views something like highlights of core values. Allbirds, for instance, and you look at their PDP pages, we know their team pretty well, they have these highlights of values that [the shoe is] durable and supportive. They have these core features of like, renewable materials, that's pre-expanded, which I think is another interesting thing to look at. 

A good event tracking plan can tell you that customers on one of your product details page scrolled down to see that everything is renewable and machine washable. Maybe people who viewed that event were 10% more likely to convert. So, let's bring those higher up on the page. Let's bring that above the fold so that a higher percentage of customers use it.

Those are the types of learnings you can only make if you have a well-instrumented PDP page, and it keeps you from saying, oh, this PDP page isn't converting because we've laid out the product wrong, or we need to do an overhaul of that page. Maybe it's as simple as we just need to reorient the content on the page, or that's a really reasonable test to run. Those types of insights, we can only get to through well-instrumented event tracking.