By Santiago Gallino
Charles W. Evans Distinguished Faculty Scholar, Associate Professor of Operations, Information and Decisions
Few things are as frustrating for online shoppers as a lousy website. They want to add to cart, but the page won’t load or there’s too much third-party content slowing it down. Santiago Gallino, a Wharton professor of operations, information and decisions, is the co-author of new research that should serve as a warning to online retailers: That clunky website is costing some serious cash.
Their study found that even a small delay in website speed significantly increases the chances that a customer won’t complete a sale or even come back to the site to try again later. The paper is titled “Need for Speed: The Impact of In-process Delays on Customer Behavior in Online Retail,” and the co-authors are Nil Karacaoglu, operations and business analytics professor at Ohio State University’s Fisher College of Business, and Antonio Moreno, business administration professor at Harvard Business School.
Gallino joined Knowledge at Wharton to talk about the study. Following is an edited transcript of the conversation.
Knowledge at Wharton: It’s intuitive that if you make something too hard for customers, they’re more likely to walk away. So, it wasn’t that surprising to read in your paper that a slow website can cause purchase abandonment. But your study quantifies how much this is costing retailers. Can you take us into those numbers?
Santiago Gallino: I agree with you. When we started working on this project, our main interest was not in finding whether a slow website would have a negative impact on customers, but how big of a negative impact. It’s reasonable that if you frustrate your customers, they are going to buy less, convert less. But nowadays most websites are pretty fast, so our question was: Can small fractions of a second impact the conversion and the sales? It turns out that the answer is yes. We are able to estimate that a 10% decrease in the website’s speed can reduce sales by 4.2% and conversion rates by 2%. If you think about how much effort online retailers put into attracting customers, if you are frustrating them down the road, it’s almost like you’re throwing away money. We thought — and our collaborators on the project also thought — that this was a meaningful impact.
Knowledge at Wharton: Where did the dataset come from?
Gallino: We collaborated with two main sets of partners. One group was the retailers that were willing to share with us the data on transactions, conversions, and visits to the different websites within their main site — like if someone was at the main page, a product page, or a checkout page. Also, we partnered with a company called Catchpoint that specializes in measuring the performance of these sites. How fast is the page loading? How fast is the page reacting to what the customers are trying to do? They are very, very precise with those measures. The tracking is important because we are measuring reductions in speed that are relatively small. The average page load is around 3 seconds, and a 10% decrease is a fraction of a second.
Knowledge at Wharton: What are the implications of this research? What should brands be thinking about when they’re designing their websites?
Gallino: There is a natural reaction around the idea that there are things you cannot control. “I cannot control the overall traffic on the internet. I have no control over what’s happening with a server downtown, so why should I be focusing on these things?” That’s partially true. Depending on the scale of your company, you can do some things to address those big issues. But the more you look at what you can control, you start to discover that there are a lot of things you can do. For example, you can decide on the third-party content that your page is loading in the background when your customer is trying to access the main page or the product page. It’s very common that we say, “When you’re online, you can track everything about a customer.” That’s true, but that also will slow down the performance of your site. That’s a decision you might want to consider in terms of how much third-party content you want to add.
Another thing you can do is think of the design of your online store for mobile and desktop. While it is true that you don’t control the overall performance of the wireless networks, we all know that LAN connections when you are working on a desktop tend to be more reliable, higher speed. You’re at home, so the interface is a little bit more sophisticated. And the customer has the patience to wait because of the context, compared to what I do when I’m commuting and I’m on my cellphone. The performance is different, but my predisposition when I interact with the site is also different. These are really two things that retailers can think about. What is the level of third-party content that I want to include, and are my customers going to be landing on a mobile versus desktop application?
Knowledge at Wharton: You also mentioned that companies should think about making the checkout process easier, correct?
Gallino: Yes. One additional insight from our study is that the impact of slowdowns through the purchase process is not homogeneous. People can get more impatient, depending on where in these purchase funnels they find themselves. For example, we found evidence that the checkout process is a place where you really need to be mindful of the speed, compared to the landing page. There are some behavioral theories that support this finding in the sense that when a customer is more in an exploration phase, trying to understand what the product is like and trying to learn and ready to engage with a company, they have one level of patience that may be different from when they get to the checkout stage. Retailers have studied this for many, many years in the physical world, and it’s interesting to see that has a parallel in the online world.
Knowledge at Wharton: Let’s go back to that feeling of not being able to control what’s going on with your website. What if the internet connection is just shaky that day? Or the site has a 50%-off sale, and there is so much traffic that I can’t get the item in my cart? Your paper said that doesn’t really matter. Why?
Gallino: Yes, that’s a good question, and I can help clarify how we estimated these things. We looked at big shocks and big disruptions for some websites and some retailers. For example, one retailer decided to add a video to the landing page. Of course, that will slow things down in a major way. But the main part of our study took advantage of all these natural variations that you are describing, things that are not under the retailer’s control. Because we were partnering with Catchpoint, we were able to measure them very precisely. We were able to say that on an average day — not those kinds of holiday season types of disruptions — these fluctuations can have an impact, even if they are a fraction of a second. That is why the design and the level of thought that the retailer puts into what third-party content they are going to use becomes relevant. That’s something that’s going to load every single day, every time someone interacts with your site. The design of a final purchase from the landing page, the product page, to the checkout — is something that is always going to happen.
Of course, you can expect that on those crazy Black Friday or Cyber Monday days, things are going to be slowing down. And maybe that day, you are struggling. But again, consistent with what we found, depending on the mindset that your customer is in, they might be more willing to wait on those days. But I think that the main point of our study is trying to focus on the everyday operation, and not necessarily on those couple of days through the year that things are particularly slow.
Knowledge at Wharton: What about the small guy, the founder, the fashion designer that maybe only sells online because they can’t afford a physical store? A lot of founders have a contracted website service. Is there something that they can do?
Gallino: I think so. In fact, there is one portion in our study where we were able to go deeper into one specific brand. And this brand is very much capturing the description that you placed in terms of an established brand but relatively small. I think that the research applied to them even more because they have a lot of control on the different steps that the customer will make when starting to know about the company because there is more exploration. This is not a super well-known brand, so people will want to understand exactly what is the type of product they’re offering? What is the assortment they carry? They want to learn about the quality of the product, learn about the designer who is behind the product.
All these things, I will argue, are even more relevant for these types of brands. Getting to the steps of the checkout, where people get more impatient, maybe someone suggests to you that you need to capture all the information that you can about the customer. My recommendation based on the study is to say, “Be mindful that those small things that you are gaining when you are capturing new data and new information, and loading additional things for your customers also have a counterpart in slowdowns that can hurt the conversion.” You need to make a balance between the benefit of tracking your customers more closely, the benefit of offering new features and gadgets that will pop up with the additional content, versus the potential slowdown and friction that you are creating.
Knowledge at Wharton: You have been studying omnichannel retail for a long time now. Thirty years ago, we wouldn’t be having this conversation about web performance, which shows how much retail has changed over the years. What are your closing thoughts on the complexity of omnichannel and why it’s such a challenge right now?
Gallino: Yes, I think this is very much why I got interested in this project. It’s this idea that many of the things that retailers have learned and have known for many years in the physical world have a parallel in the online world. I’ve talked with retailers many times about the store layout, the checkout process in the store, the experience of the customer when they go to the physical store. It could be thought that when customers are online, many of these things are not true anymore, that people can handle as many products as they can see on a screen. Assortment issues become less relevant online because you can carry everything you want. The layout of the website can be a little bit cool, a little bit flashy, but we don’t need to think about the flow of how the customer discovers products online. Many of these things have been studied, and it has been shown that this is simply not the case.
I think that our work here helps bring one other perspective. Of course, you care about how much time it takes you when you make a physical trip to a store — from the moment you arrive at the parking lot, where you can park, how far the store is. You care because those are minutes, sometimes hours that you’ve missed. Well, it turns out that if you’re going to invest a few seconds in a website, those seconds can add up, and the customer will end up buying more or less based on that narrow experience. This connection between the physical and the digital world — once again, shows that it is there. And I think it would be a missed opportunity not to think carefully about these performance issues.