The rising trajectory and stellar performance of e-commerce on a global scale has been phenomenal, making it one of the biggest internet success stories of all time. Having evolved into an efficient and formulaic framework over the years, it now bears little resemblance to its offline peers in the physical world. While online shopping scores highly in most ways over shopping in traditional brick and mortar stores due to its functionality, it leaves one box unchecked and shoppers nostalgic over a crucial aspect in the purchase cycle – the human interaction. This calls for a middle ground that marries the successful features of e-commerce with the human elements of retail stores, transitioning the shopping event from ‘consumption’ to ‘experience’.

What are some of the human elements of shopping?

Shoppers in a physical store seek assistance and expect the attendant to be knowledgeable, to communicate well, and quickly gauge what they want, while bearing in mind that they will likely be willing to compromise on some features and prioritise others. The assistant engages in asking, understanding, recommending, and selling.

Additionally, consumers are influenced by the people whom they trust, such as friends and family – as well as celebrities or Instagram influencers. They seek their advice and recommendations before making decisions.

So, how well are these human elements being reflected in e-commerce?

Human advice and recommendations: A study reveals that 97% of consumers read product reviews before making purchases. E-commerce sites have boldly embraced the user-generated review and rating system, which provides users with candid and credible opinions about products. Sites like Lush lay huge emphasis on testimonials, reviews and ratings, which, through their web layouts, take precedence over other product information – it’s an approach that their A/B testing reveals to have succeeded. Taking this a step further, Guuru enables e-commerce sites to allow their shoppers to connect directly with other consumers for their first-hand feedback.

Human influence: Brands are becoming increasingly conscious of the desire for human interaction, and are using social media in a fun and human way. An example is that of Facebook Shops, which allows anyone from a sole trader to a large brand to sell products, end-to-end, via Facebook, Instagram, and soon Messenger and WhatsApp. In its endeavour of combining e-commerce with more human interaction, it allows brands to tag their products to images, videos and even live broadcasts, encouraging them to connect with their followers using their personality, knowledge and insights to boost sales.

Shopping assistants: The trickiest human element to emulate in e-commerce is perhaps the knowledgeable, helpful shopping assistant. Human assistants can naturally communicate and understand both simple and complex language. They can also pick up subtle cues and nuanced phrases that help them understand consumers better. Many of them have a good understanding of the products they are selling, and can use their broader knowledge to guide consumers to make informed choices. However, the investment of deploying human assistants for one-on-one interaction on an e-commerce website is prohibitively costly.

This is where an AI-based shopping assistant can come in to fill these roles. However, as discussed in an earlier article, AI personal assistants currently seen in the market are far from human in their approach when it comes to helping consumers shop.  This, however, need not stay that way – but how? The AI shopping assistant must be re-imagined from the perspective of the human touch.

How can an AI shopping assistant provide the human touch?

Advancement in Natural Language Understanding (NLU) is making it possible for AI assistants to correctly interpret more nuanced and subtle phrases. Through complex understanding of the language, an AI assistant can ask the right questions and understand consumer search in a more human way, for example, knowing the difference between saying “Bluetooth would be nice” and “I need Bluetooth”. It can also use the product data contained within the site to offer consumers insights into products, such as price differences and product comparison based on several features, as well as making intelligent recommendations and suggesting alternatives.

The agent can work in a human-like fashion, assimilating information from internal and external sources. For example, if someone were to look for a tent for a trip to Scotland in August, the agent could find the local weather and suggest the right kind of tent for the shopper. Furthermore, it can also use assumed logic to ask the right question or make suggestions. So if a potential house buyer has looked at a 4-bedroom house and is asking about nearby schools, the agent can assume the user owns a car, and based on this, it can ask the buyer about their preferences regarding parking.

Summing up

Adding human elements into e-commerce is resulting in greater marketplace success, and this will continue to be a driving force in the years ahead. While we become increasingly detached from physical stores, the need for specialised knowledge, helpful information, interesting insights, and real-world experience is still something that shoppers crave. For a meaningful interaction, shopping assistants must ask questions, as opposed to just finding answers. Only those e-commerce sites that embrace the human touch will get the best of both – the online and offline worlds. According to customer service expert and keynote speaker, Shep Hyken, “The greatest technology in the world hasn’t replaced the ultimate relationship-building tool between a customer and a business – the human touch”. While it may seem unrealistic for technology to replace the human touch altogether, more can be done to bring human elements into the online shopping experience.

At Traverz, we have developed an AI shopping assistant that takes a Conversational Recommendation approach. The smart assistant asks relevant questions, remembers user preference, and offers contextual advice and suggestions — almost akin to a human shopping assistant. Click here to learn more about how we’re humanizing the product search experience.

To find out how you can build an e-commerce platform that delivers better insights and greater brand loyalty from customers, get in touch with us.

Here at the Traverz office, we live and breathe product search technology. So we like to stay on top of what other companies are doing to update the online shopping experience.

Recently we’ve taken a look at how two of the biggest players in online shopping, Amazon and Google, are using AI to innovate product search. Over the past two decades these two companies have dominated online shopping and product search and are constantly innovating to try and stay ahead of the competition.  But just how well are their latest product search developments delivering for their customers?

Hey Alexa, let’s talk about AI assistants

Alexa has become something of a cultural phenomenon.  The ubiquitous voice activated smart assistant can be used for many things, from playing music, to checking your schedule and, not surprisingly as it’s owned by Amazon, also buying products online. So how does searching for a product with Alexa work? It’s actually incredibly simple – the user asks about a product and Alexa will offer a single recommendation that the user can either buy or ignore. That’s it! Such simplicity can be a beautiful thing when you know exactly what you want, you can go from “Hey Alexa, I want to buy an iPhone XS” to it arriving at your door that same day.

The problem arises when you are not quite sure of what you want or don’t say the right thing. One simple change in the above example can highlight this issue. Saying “Hey Alexa, I want to buy an iphone” leads to just one recommendation from Alexa. You are left in the not-so-safe hands of the recommendation algorithm. This is where Amazon may well offer the iphone that will make them the most money rather than the one that best fits your needs.

Ok Google, can you do any better?

Like Alexa, Google’s Assistant can be voice controlled, but also works with shopping apps where you can view and touch the screen. This allows the AI to offer a more complex and detailed product search experience. The Google Assistant can control various “assistant enabled” apps to provide shopping assistance, where you get to specify things like price, size etc. Google will then select 3-4 products, ready to purchase. While it offers you a little more control and choice than Alexa, you are still just forced into their recommendation list via a few basic filters. It feels more like a step back in time than a look to the future.

Worse yet, the process is cumbersome and buggy. You have to guess which of your apps is “Assistant Enabled” and then go through a laborious process that often mishears, misunderstands or even just gives up completely, as can be seen in this painful video below:

What else do they have up their smart sleeves?

What surprises us here at Traverz is that surprisingly little product search innovation seems to be coming through from the tech giants. Both Google and Amazon still remain heavily reliant on the search and filter model that has been the standard for decades – with no attempts to completely re-imagine that search experience. Amazon does seem to be making some gentle steps by building experimental tools, and a couple of these are quite interesting.

Their Style Snap fashion tool uses AI image recognition to analyse an image you upload. It first tries to determine which parts of the image contains clothing products (dress, hat, shoes etc); then once you select which of those products interest you, it will try to find an exact match, along with various items that are similar. It does both of these tasks rather impressively, almost always providing quality matches to the uploaded image. It is however a little limited in its usage. Once you are provided with a list, you can only really set size and price to further influence the search. It is also locked to the Amazon store, so you can only find products that Amazon stocks. For areas like fashion, where searches are often visually driven, it really does feel both futuristic and useful though.

Another interesting tool is their Your Journey preference tool, which appears to be in testing now and is only accessible in the lighting section. Once you are on a product page, you get a large recommendation panel that encourages you to like or dislike a range of additional products based on how they look. After each click, the AI updates those products in line with your desired style. After 3 or 4 clicks, you really do feel the results are in the style you prefer and you can then save that “journey” and narrow it down further with filters. Again, you are trapped in the Amazon store and are really just getting image based recommendations. But the feeling of having some control over more abstract ideas like look and style feels like a big step forward.

Are we seeing the future of shopping from the big guys?

Not quite yet. The AI assistants are very good at simple actions like asking about the weather, but unless you know exactly what you want to buy, the process of product search is currently worse with the Amazon and Google AI assistants than when using their standard website. However we are seeing Amazon innovating with AI in little pockets of their site and what they are achieving does feel both smart and helpful, albeit in a basic, locked system.

Here at Traverz, we are humanising product search by providing the user with a smart, AI driven agent that takes a lot of the work out of looking for products, as well as simple yet powerful tools to control their product lists based on their preferences and tastes. The Traverz shopping experience can work across multiple marketplaces, understanding your style and providing real help to find the best products for you.

To find out how you can build an e-commerce platform that delivers better insights and greater brand loyalty from customers, get in touch with us.

E-commerce merchants know surprisingly little about their customers needs and wants.

This is a bold statement that runs somewhat counter to common wisdom, and it is true that e-commerce generates a lot of data originating from visitor’s search queries, mouse movements, page views, filter selections and purchase histories. But this data provides only an observation of how the consumer is engaging with the available product selection and the existing site structure, including the product selection mechanism (such as filters or search bars).

An example – how does an appliance webshop know that a consumer is looking for a quiet large-load washing machine with a low energy consumption? Or that the customer’s primary criteria is in fact the noise level, and that they are willing to settle for a smaller wash load if it helps bring the noise level down? This is not easy to deduce from the data that is currently collected. The consumer is unlikely to enter this level of detail into a textual search query; purchase histories are not very helpful; and mouse movements and page views are difficult to interpret due to their very low signal to noise ratio.

Hence the surprising truth is that e-retailers often have little insight into their consumers needs and wants compared to brick and mortar shops. Physical shops are able to gather information through personal interaction with shop assistants, creating a much more direct line of communication with users.

Up until now, one of the few methods of feedback that e-commerce sites can use is the product selection mechanism. Unfortunately (and as we’ve discussed in a previous blog post) customers have to a large extent learnt to avoid this feedback mechanism – the filter search system – due to its poor user experience, meaning that sites miss out on capturing potentially useful information. And when they do use it, the data that is gathered is often shallow and focused primarily on elementary characteristics like size, colour and brand while missing out on the long tail of customer interests.

This noisy and shallow data is then used to deduct trends and build a picture of what really interests and moves customers. Unfortunately, it’s rubbish in – rubbish out, no matter how much you spend on machine learning and AI.

So how can retailers capture the full breadth of a customers’ many and varied needs and wants?

To get a more detailed understanding of the customer’s thought process, it’s critical that a consumer can express the full breadth of their requirements as they interact with your site.  Here at Traverz we’ve found that the key is to implement simple and unobtrusive feedback mechanisms, enable the consumer to provide indications of relative importance, and to expand the set of interactive features and options to capture the ‘long tail’ of consumers search requirements (see here and here for how we implement this). Through these changes the consumer not only has a better search experience, but the merchant obtains a far deeper and less confused insight into consumer needs and wants.

The resulting data flow provides detailed insights into customer interests that need no further deduction – the data directly tells the story. This enables merchants to be better prepared to:

To find out how you can build an e-commerce platform that delivers better insights and greater brand loyalty from customers, get in touch with us.