Mapping the Consumer Genome: Ep 3
Featuring a conversation with Simon Carpio of CarParts.com on:
- How eCommerce is evolving
- What type of data is important to collect
Conversation with Simon Carpio, Vice President of Product at CarParts.com
00:06 | Debjani:
As we get started, I would love for you to take a few minutes to introduce yourself, your experience and your work at US Auto, and we’ll go from there.
00:16 | Simon:
Sounds good. My name is Simon Carpio. I’m the Vice President of Product at U.S. Autoparts. My background has really been building a better shopping experience on the website. Our flagship website is carparts.com, and really a lot of our efforts have gone into there. More recently, though, I’ve moved into not just the shopping experience, but that post-purchase experience as well: looking at things like shipping, communication of WISMO, power contact center handles, different customer complaints, and then really folding that back into an end-to-end experience for the customer.
How is eCommerce evolving?
00:56 | Debjani:
So, Simon, in your capacity in the U.S. Autoparts, you obviously have a very executive-level bird’s-eye view on where eCommerce is heading. What are the things that are front and center for you? If you could share a little bit about how you think about the evolution of commerce, and what is in store in the next few years.
01:22 | Simon:
Yeah, sure. When I think about the shopping experience online, I think we’re moving more and more towards relevance. I think the ultimate goal is to be that exact need for the customer, at that point in their shopping journey. The ultimate goal for me, and I think the best way to look at it—in terms of forward-looking—is really creating that personalized experience for the customer, and only injecting things that the customer needs at a very specific point in time.
Right now, what we’re seeing a lot of is just a bunch of shopping widgets up. They [customers] will find what’s helpful, they will sort it all out, but that can come at the cost of attention, focus, and ultimately, conversion rates. Not showing the right thing to the customer at the right time is really going to be detrimental moving forward. It’s those companies that can get it right most of the time, and then eventually all the time, that’ll really start to see them adding value for the customer.
What type of data is important to collect?
02:29 | Debjani:
There are a lot of conversations happening in regards to how to get that data together, and what are the right approaches to it. Is it housed in one place? Is it a CDP? Is it housed where it is and gain access to it? There is a data management layer. How do you think about it, and how have you been thinking about it?
02:52 | Simon:
That’s interesting. In our industry, we have basically two cohorts, two big customer segments that we like to call: DIY (the Do It Your Self-ers), so people who actually get to work on vehicles themselves, and then you have the DIFM market (Do It For Me), they might have a mechanic that they know. Maybe a friend, a brother, a sister, someone else. So, those two segments exist, and if you asked me a couple of years ago, how do you create the shopping experience for the customer—what different customer journeys and different features on the site show up? I would have told you to segment by a new user, a returning user, and a repeat purchaser. So, very historical-looking.
But what we’ve found is that…there’s no mutually exclusive line between a DIY’er and a DIFM’er that can be found when you’re looking at cohorts inside customer segments, historically. You really need to listen into the in-moment signals, those real-time signals that are happening. And I think, a lot of people will see this is that, when you’re moving from moment to moment within the session, it may be the case that a customer is searching specifically for a part number and only really looking at, at least in our environment, an interchange part number. They’re looking for maybe an OE number, a parts link number. But then they can move into another category where they pogo stick back and forth between the listing page and the product detail page. And so, here you have one customer; what segment do they ultimately end up in? It’s that moment to moment listening where you can actually start to surface the right features and the right experience.
I think one of the main hurdles—still—is being able to architect the site in a way that can respond to real-time signals. When you think about certain widgets; in our industry, it could be things like how-to guides, installation guides, tools is another big one. When and how do you service that? So your algorithms now have a bunch of data that can reflect real-time moments, but how do you listen into that? How do you make sure that that’s capturing, and that’s changing literally the user interface of your website? That’s sort of the things that I think about longer-term to create those very, very hyper-personalized experiences.
05:37 | Debjani:
You know, it’s interesting. What you are referring to is something that we call short-term user behavior. Essentially, the premise is that each person coming on the site—even though you might know them historically—in that moment in time, in that context, is showing a set of behavior that is specific to that moment. And therefore needs to be listened to, and probably has higher bearing in the ability to add value to that moment, than does the historical.
06:14 | Simon:
Yeah, definitely. Weighted for recency is sort of how I see it.
How is AI and ML being applied to short-term user behavior?
06:20 | Debjani:
Yes. I think that leads me to the question that, over time, this sort of listening to these kinds of signals and being taking action on them, does require significant automation, right? Because—as you can imagine—this is millions and millions of signals in an hour. How do you see the use of AI and ML playing into this? In what form do you see that happening, and just your thoughts on that?
06:54 | Simon:
I really see machine learning scaling very similarly to how it’s probably scaling most people’s marketing campaigns today. We used to think that email blasts and certain marketing campaigns cast a wide net. Now what you want to do is cast a bunch of smaller nets.
When I think of machine learning, you have all of this data in the background, and there’s a structure to it where you can get a start pulling from historical data to make your best first guess. But as that starts to change—and again, that customer segment starts to become fluid within the session—machine learning can then say, “Hey, this isn’t the algorithms you use, these aren’t the features to show. Start putting up these features.”.
What is continuous intelligence?
07:43 | Simon:
I see it as a layer sitting on top of the CDP, and probably a tag on your website that’s making adjustments depending on how your site can react to those adjustments. Some websites can, some websites can’t, so there’s a technological stack investment that needs to happen there as well.
08:04 | Debjani:
Yes, and Gartner has come up with a new terminology for it, which I think makes sense; they call it continuous intelligence. Each individual represents a stream of continuous intelligence, and how the brand reacts to that continuous intelligence, really is to be seen, because it requires you to have a new type of stack in place to be able to react to events as they are happening.
08:32 | Simon:
That’s right. I like that—continuous intelligence, constant adaptation.
What technologies do you plan on implementing in the near future?
08:37 | Debjani:
What does the next two to three years look like in regards to what are the technologies you’ll be thinking of and implementing?
08:46 | Simon:
It will definitely be layers of machine learning that can help with personalization experience. I can’t really go into what our technology stack is today because it is pretty proprietary. But, moving towards that more recent signals, using that in an algorithm, being able to shift elements on the page as necessary, hide, and then show as well. We’re moving more and more towards that.
And maybe at some point, the way I see it is, the customer experience is one experience, so kind of leaning towards what I was saying earlier. The customer sends you something on your Facebook Messenger, your company’s Facebook Messenger, somebody sends you something through live chat. Really creating a feedback loop where the agent knows what’s going on on the site, and the site knows what’s going on at the contact center, and then really creating this one experience.
To make that concrete, a very simple use case here is: a customer calls in for WISMO or sends a live chat message for WISMO. Maybe his homepage all of a sudden has a little widget up there that says, “Hey! Your recent activity: Tracking this order,” until the time that it’s actually delivered because that’s probably top of mind for that customer.
What are the keys to success over the next few years?
10:06 | Debjani:
Excellent. One last question then. So, we have Covid happening all over, we have sort of a different world, but what is the key to success in the next few years?
10:21 | Simon:
I think, for our industry, the key to success is, again, really being able to identify a user as a DIY’er or DIFM’er, and creating either very frictionless experiences where you’re not doing the upsells, you’re not doing the email blasts, and you’re kind of just letting that customer check out as quickly as possible in some categories—and then moving more into a guided experience in other categories. I think that that’s going to be really good. I also think, moving more towards the DIFM side, what other services can you provide? I really want to make sure that certain segments only see things that they want to see, and then they’re able to check out in the way that they want to check out. And that post-purchase experience is always reflected back into the site, and whatever they do on the site is reflected back in that post-purchase experience.
As we move forward, you can imagine some of our categories are very seasonal. Tying in just what’s happening in certain segments, certain geographies of the U.S., and making sure that that customer almost immediately sees what he needs to see.
11:33 | Debjani:
So localization, if you would, and personalization based on location, and all of that.
14:53 | Simon:
And everything, yeah. If a customer has a daily driver, and that daily driver is the active vehicle in that session, we’re probably going to give him maintenance and repair solutions. If the customer has an older vehicle, we’re probably going to give him more reconstruction. And you know, try to build different experiences around, “Hey, you’re working on your old Mustang—here’s a front bumper, here’s a rear bumper. You’re working on your pickup truck—here’s a front bumper, here’s a fender because you probably got into an accident.” So kind of thinking through those sorts of scenarios.
12:16 | Debjani:
Certainly, makes sense. Thank you so much, Simon.
When I think about the shopping experience online, I think we’re moving more and more towards relevance. The ultimate goal is to be that exact need for the customer, at that point in their shopping journey.
– Simon Carpio, Vice President of Product at CarParts.com