Dialogue-based systems – whether text or voice-based – have become commonplace in our lives. One billion devices sold and delivered with Google Assistant, 1.5 billion active WhatsApp users, 100 million Alexa smart speakers and over 1 billion users on Facebook Messenger demonstrate the indelible impact of these systems on our society.
In this guest article, our Shopware Solution Partner Pixup Media explains what potential the increasing number of digital assistants offers retailers.
Most recently, voice-based systems – and specifically Amazon Alexa and Google Assistant – have taken the market by storm. The products of both companies feature an open platform architecture and the devices are marketed with increasing success by themselves or via third-party providers.
The continuously increasing number of digital assistants and the previously established base for messenger services underlines an extraordinary potential for brands and retailers. Voice and text-based chatbots form the basis for continued contact and interaction with customers outside the company’s own shop. Specifically, the direct interaction options (e.g. automated product advice) for customers who are currently not on the shop page don't just create plenty of new marketing and sales potential – they also deliver a completely new kind of shop frontend for potential customers.
How do chatbots work: the technology behind KI chatbots
Text-based chatbots are not really a new idea. An unstoppable wave of hundreds of thousands of chatbots hit the market once Facebook opened up its Messenger app for chatbots in 2016. Most of these bots were based on a fixed rules tree. The bot became useless as soon as the user deviated from what it expected. That is why new solutions were needed, and it became clear that self-learning systems achieved significantly better results.
This professional chatbot system is based on a Natural Language Processing engine, aka an NLP. This engine ensures that the bot will understand various phrases with the same content, synonyms and even dialects correctly and will deliver the expected result. That requires a training phase for the NLP engine. Let's have a look at a simple example:
We sell T-shirts, jumpers and shirts in a wide variety of colours in our fashion shop. We train our bot to understand the phrase "I’m looking for red T-shirts" and tell it that "T-shirts" is a category name and "red" is a filter, which means that the user will be redirected to a filtered listing page. We do the same for another colour and a jumper. The bot now knows what the NLP engine should do if a user looks for a white shirt, without us having to teach it specifically beforehand.
This is a very simplified example, but it does demonstrate the self-learning behaviour that can be achieved via the NLP engine, which is based on machine learning – and therefore artificial intelligence.
New touchpoints and applications thanks to chatbots
In the kitchen while preparing a meal, in the car on the way home or on the commuter train with a smartphone – the number of touchpoints created for brands and retailers is growing exponentially. The three main players, Facebook Messenger, Amazon Alexa and Google Assistant, are ahead of the field because of their open platform architecture and huge proliferation. WhatsApp, the Apple Business Chat and Google RCS are comparatively new to the game or haven't even completed the beta stage yet. And there are a few other channels like Samsung Bixby, which will add another huge potential and an open eco-system in 2020.
Three concrete application areas in which chatbots can be used:
Customer-specific product advice just like in a retail store – and that via chat or voice – is probably the most obvious benefit of a chatbot in e-commerce and puts the viability of the last bastion of offline retail into question. The full browsing and shopping history of a customer will be known and the bot will be able to offer customised product selections based on that information in conjunction with a recommendation engine.
The combination of these two technologies (chatbot and recommendation engine) will give rise to some very exciting solutions that can also be used for highly effective marketing. In our fashion shop from above, we can now bring to life "Laura – Your personal style consultant" as a chatbot. "Laura" will know the sizes, preferences (more of a T-shirt or a dress shirt person?) and price ranges of each customer and can then recommend only those products.
An analysis of customer service requests in most companies will show that far more than half of all queries and requests are similar or repeated. A perfect environment for chatbots who can process questions and concerns about delivery costs, returns and orders around the clock. And should a bot be stumped at some point, then the issue can be forwarded to a human colleague. The latter will be able to unload a significant amount of repetitive work and can focus on much more complex tasks instead.
Marketers must learn to think in a chatbot context. In contrast to email marketing, prospective customers who are targets of a proactive Messenger campaign can ask product-related question directly. It is important here that these questions are answered swiftly and competently as part of the campaign. This will be an amazing opportunity for brands and retailers, because marketing campaigns are no longer simply a one-way street and land directly on the smartphone of the potential customer. Involving a customer in a chat will also be perceived less like marketing, and the opening and interaction rates are by far more promising than in traditional advertising campaigns.
Why chatbots must be recognised as a standalone project
Chatbots are by their nature as distinctive as the product range, the brand and the customers. The technology working in the background is most likely standardised, but the content design and layout will be as individual as a fingerprint. That is exactly why a serious chatbot should not be underestimated in terms of integration. It will basically represent a project in itself, which also needs smart minds and specialists in the background. The integration of a chatbot is never finished. It must be monitored and developed further all the time.
Let's look at some examples in that context:
1. Returns process: This process is basically unique to every individual company. Let's just look at a returns label. When a customer wants to submit a return, then the bot will create the return label for the correct order. In some systems, this can be done directly via Shopware, others use their ERP system or the DHL returns portal. There is never going to be just one solution.
2. Product recommendations: No chatbot will be able to produce meaningful product recommendations without a deep recommendation engine integration. That is simply due to the fact that the bot must "know" the customer and their preferences. Here is an easy example: 1,000 new products were added to the shop yesterday and a customer asks his smart speaker what the latest products in the shop are. The bot will have to know his personal preferences to narrow down the search results to 1 or max. 3 products.
Of course, chatbots won't be taking over the traditional frontend of an online shop any time soon. But new channels like Amazon Alexa or WhatsApp will definitely become an important extension for every online shop. Since digital assistants in particular are added to ever more devices, there will be simpler and less cost-intensive options for brands and retailers to bring the shop to the customer instead of the other way around. Chatbots will also become very valuable in terms of customer loyalty and the general customer journey, as they can provide services a conventional frontend never will.
The early bird catches the worm. When the mobile revolution was proclaimed back in the early 2010s, many brands and retailers could afford to sit back and watch the market development unfold, before a decision was made to have, let's say, a responsive template created over the course of a couple of months. In our age of artificial intelligence, watching and waiting has become counter-productive. Machine learning, i.e. the technology behind the term artificial intelligence, requires a lot of training, data collection, building experience and ongoing further development. Brands and retailers who take the first step today will have a significant competitive advantage over those who only get started in maybe another two years from now.