With foot traffic reduced and global travel restricted, in-store shopping has taken a hit. In the U.S., department stores like Neiman Marcus and JC Penney have gone bankrupt. Since the pandemic began, there has been a 33% surge in online spending due to worries about health amongst consumers. However, this story is not only about our current pandemic as the trends towards eCommerce have been clear for years even as the pandemic accelerated these trends. For example, Green Street Advisors in the U.S. say the disease outbreak contracted a five-to-ten-year trend of department stores exiting as mall anchors to the point where a little more than half of U.S. mall-based department store could close for good by the end of 2020. Yet, retail sales continue to grow.
In February 2020, the National Federation in the U.S. had predicted 2020 retail sales would grow between 3.5% and 4.1%. That was before COVID-19 pandemic caused states to issue shelter-in-place orders. Just one month later in March 2020, the NRF predicted that the small retailers industry would lose $430 billion in revenues in the third quarter of 2020. Despite these dire predictions, retail sales are bigger than ever according to the U.S. Federal Reserve reaching a record of US$553 billion last month. In Europe, retail sales have grown for the past six months and Japan saw YoY growth last month.
So, what happened? eCommerce happened. The growth in eCommerce has more than made up for the loss in foot traffic. 47% of consumers say they are making more eCommerce purchases than they did previously. We can simply look at the growth in internet traffic to see the growth in the infrastructure behind this change. The millennial buying power is increasing YoY and make up a larger segment, especially in the luxury sector. Gucci reported that 50% of its sales came from millennials, who purchase many products online – thus, brands have to face an audience that have higher demands for convenience, speed and also a personalized experience.
So which categories are growing? Any global internet discussion cannot be made without commenting on Amazon and eBay, the global giants. eBay reported over 25% YoY growth in Q2 of 2020 and Amazon’s revenue is greater than US$6.5 billion.
However, this masks some of the real winners in the move online. Instacart has seen 150% increase in business at the start of the pandemic and in general foods, home fitness gear and home furnishing have been doing exceptionally well. Luggage and cameras have not been doing well at all with declines of over 75% in some cases.
Where do consumers go? They often turn to the big marketplaces but as we have noted in previous posts, using Philip Kotler’s perceived value, the brand accrues almost nothing and price becomes the only differentiator. This is good but only temporarily for consumers until quality lapses. European luxury brands rushing into China found that the route through Alibaba was fast, but the plethora of fakes and cheaper imitations cut into their perceived value. So, how do you retain that brand identity? How do you foster loyalty when price is your only perceived value?
If we look at the trends that have helped merchants build brands in eCommerce, it is not in the U.S. or China, but rather a company in Canada. Many merchants have sold products on Amazon or eBay or Alibaba and the reason for this are two-fold.
First, the barrier to entry is low. You can have a virtual store up and running in no time on these marketplaces. Second, the brand value of these marketplaces ensures that there is foot traffic. In the same way, a physical mall can guarantee a minimum number of visitors, large platforms ensure visits to their site due to the name recognition but these visits come at a high price. The NY Times estimates that “if a business lists an article of clothing on Amazon for $50, Amazon gets $8.50 in commission; if the seller opts to advertise on the site, Amazon likely gets at least another $6.50. And if Fulfillment by Amazon is used, Amazon’s total cut gets closer to 40%”.
It is no wonder that merchants whose margins are less than 40% or who want to build their own brand may want to build their own site away from the marketplaces. Here is where Shopify, WooCommerce, Magento and other eCommerce platforms come to the rescue. The platforms allow merchants to construct a website, install a payment gateway, find a logistics company, and customer support from one central place. More and more find they are being served with good marketing and SEO tools allowing brand identity to be built and this has been a success. Shopify merchants registered an astounding US$5.1 billion in sales this recent Black Friday and Cyber Monday.
Although we may want our brands to emulate the successes like Warby Parker, Allbirds and Shinola the amount of work to design a store can be daunting. Even with Shopify, how do you stand out from many other shops? This is where Visual AI comes into play. If you are a grocery store, you may have between 15,000 – 60,000 SKUs. If you consumers want a replacement for their favorite brand, you can offer them the opportunity to take a picture and automatically match with a similar item. Voila! Your customer is now shopping from a range of similar items as they would in-store but the magic does not stop there. As the consumer moves online, the shelf space they see is much smaller than in a physical store but crucially, it can be dynamic. With Visual AI, you can begin to create a better and more engaging online experience for all of your customers by harnessing this dynamism.
With Delvify’s SMART SKU, you can create your own personalized banner that recreates the product shelf any way you see fit. Within that banner, you can create a recommendation system to help your customers. 35% of Amazon purchases and 75% of Netflix viewing is driven by recommendations according to McKinsey. So we know they work!
You may have dozens of reviewers that allow for collaborative filtering, but your eCommerce site will have enough information that can be used to create an enjoyable and unique visual customer experience. By utilizing the browsing behavior of your customers, you can super charge your recommendations and create a unique shelf presentation to each customer. Customers who come to your site will want to know what is popular and what is useful.
How does Delvify make this work? It all starts with the eye. Your customers will be extremely efficient at seeing and interacting with the products they like. So give them this visual stimulus through visually similar objects but also work with them to provide interesting and unique variations on what they see. Due to this unique virtual shelf space, your brand experience will be better than a chance encounter on Amazon. Delvify helps you stand out from the crowd by incorporating the customer behavior into your virtual shelf space, Delvify’s recommendation tool makes the search results into something special – a unique customer journey for each customer.
A customer will be looking for items, but also want confirmation that the item is a popular one. With a careful machine learning approach, we can understand the behavior of thousands of users and incorporate this learning to boost the more popular items or the more likely to be purchased items to the top. The AI and machine learning can be combined to help make your super hot items be front and center.
For some companies who are proactive in pricing and inventory management may choose a different route. By highlighting less seen or overlooked items, you can increase the customer discovery part of their journey. Each visual element is analyzed to create the perfect feed. With affordable AI, the Delvify SMART AI tool goes beyond the simple match to create a unique journey – every step of the way.
Generally, when we think primarily of recommendation systems, we have collaborative, content, or some combination of these two systems. Content generally refers to a simpler approach where previous items are assumed to be indicative of future interest. This is usually determined through the calculation of the Euclidian distance or Cosine Similarity, the idea being that you can identify different features of an object, color or description and the vectors created for each item are embedded in a matrix which can be used to find similar products. Collaborative filtering is a bit more complicated and uses user interactions (purchases, clicks, views) and/or ratings to create a matrix of related items.
The baseline approach to collaborative filtering is matrix factorization. The goal is to complete the unknowns in the matrix of user-items interactions (let’s called it RR). Suppose we have two matrices UU and II, such that U\times IU×I is equal to RR in the known entries. Using the U\times IU×I product we will also have values for the unknown entries of RR, which can then be used to generate the recommendations. Using neural networks to classify (clustering for example) or for regression analysis is one way to think of these systems. Best of all, you can avoid the cold start problem. We have a paper about the technical side of recommendation systems used by YouTube, which you may read for some interesting ideas.