This will be the year that will not be forgotten. For the first time in contemporary history, a worldwide pandemic has effectively stopped the world. The images and feelings of vacuity and discomfort we all experienced, will never fade.
We've seen Manhattan's vitality vanish into empty streets and the desolation of Shanghai's subway. The once crowded corridors of the Louvre now feel like a lonely collection of paintings; and the major airports in the world are reduced to nothing but huge giants full of empty lanes. All this cannot be forgotten overnight
But what to take from this? Kahlil Gibran, libanese poet, philosopher and painter has stated that:
"Nothing will prevent the sun from rising again, even the darkest night. Because beyond the black curtain of the night there is a dawn that awaits us."
Keeping this in mind, we at Toblee have developed a solution to help the small and medium enterprises (SME) to digitise their day-to-day activities. All of a sudden Covid-19 has forced SMEs with almost fully traditional operations, to find alternative ways to do business. The business environment quickly became an hostile and increasingly challenging environment, with the main objective that shifted to survive in the hope of a better future. We suddenly realized the importance of an online presence and of effective communication channels
What it does
Toblee has embraced this challenge, carefully listening and sharing ideas with both customers and store owners, learning about each business’ specificity and need. We designed a chat-based solution leveraging Facebook Messenger capabilities, scalable to the vast majority of traditional brick and mortar stores, simple to use for the end customers, easy to interact with for the store owner. We leveraged the latest technologies among which Natural Language Processing to create an AI-based chatbot that allows customers to interact seamlessly with the shop via general questions, booking and ordering intents, as if a shopping assistant was always there for you. The agent deals with all customers' requests, taking care of the questions with tailor made answers.
The chatbot is also provided with an e-commerce solution and an in-chat booking system, with all the requests which are routed to a single external chat kept between Toblee and the shop owner.
Once the request is passed to the store owner, he can either accept, decline or propose a new time to the customer. This task can be accomplished with a single tap on the message he receives. Furthermore, the chat on the shop owner side can be shared among multiple smartphones, allowing for both employees and owners to have full visibility on the incoming requests.
The customer will then be notified about the status of the order directly into his chat. To provide a better UX, we decided to rely on an external landing page for the ordering process, while providing a fully in-chat experience for the booking system, leveraging the same backend infrastructure for the accept/decline process.
The full flow is shown in the image below:
This way, the shop owner avoids dealing with hundreds of calls and wasting time manually recording the orders, which easily interrupts the core day-to-day activities. Moreover, this solution is very scalable and flexible for multiple purposes, such as remote ordering, delivery and at the table orders. It can be used for remote ordering, delivery, as well as for at the table orders. it perfectly fits different situations: from takeaway at pizza shops to booking a table in a restaurant, from ordering an ice-cream while avoiding the queue to collecting grocery from a local farmer. This also removes problems related to the queue, that have always been a barrier in the customer purchase checkout process and nowadays represents a true obstacle due to social distancing and in-store quotas of people. The solution is completely commission-free on the order or booking, removing the burden of crazy-high commission charged by delivery apps, which discourages several sellers to do business online as their margins are completely wiped out.
Needless to say, it leverages a 1.3bn monthly active users customer base. It can be integrated directly into the Facebook Page of the shop and into the Instagram bio. The chatbot natively supports English and the local language of the business. It is the perfect solution to reach a different segment of your customer base, from a more young picture-driven type of users who navigate Instagram, to a more adult segment of customers which use Facebook extensively. Furthermore, this is a quick and much cheaper solution for shop owners compared with the setup of a website and of an ecommerce. In fact, many small shop owners in the European food segment don’t even have a website for their business and rather rely on a Facebook Page through which they post daily content. Toblee has developed this solution with a clear goal in mind: providing an effective and easy-to-use framework to cover the current and future needs of both customers and sellers in the retail space.
Despite being still fighting this pandemic, we at Toblee believe that dawn is just around the corner. We want to be the preferred solution for customers and sellers today and tomorrow, creating a new concept of interaction and enhancing the digital presence to appreciate even more the physical one, when we will go back to our normal lives.
How we built it
The goal of the system is to deliver a smooth and seamless experience for the user in a number of different devices. We evaluated the possible use of several technologies and eventually decided to use a MERN stack as it is an open source full stack development framework which provides entire front-end to back-end development components. The stack is composed of MongoDB, Express JS, React JS and Node JS and they are used in synergy as depicted below:
The interactive user interface is created using ReactJS, HTML and CSS. The front-end connects to the back-end through API calls (typically POST and GET http requests) that triggers actions on the database using Mongoose, which acts as an interface between the MongoDB database and the application. The whole flow of actions to place an order can be seen in the graph below:
The user begins a conversation with the bot either in Whatsapp, Messenger or Telegram. The message is passed to Dialogflow which applies Natural Language Processing in order to extract the relevant information from the message, and then responds accordingly by providing a link pointing to the online menu where the user can place the order. Once the order is placed, a recap is sent back to the user while a notification is sent to the vendor. He can then accept, decline or change the time of the order, and the user will be notified accordingly.
A similar logic applies to every other type of request such as table booking, or questions such as “are you open tomorrow?” etc.
On the Natural Language Processing training, we took 2.3 thousands training phrases, which were collected via surveys and through a collection of Facebook post comments in reply to Food posts. We decided to use 1) surveys, to understand which intents are the most commonly asked while chatting with a bot, and 2) Facebook’s post comment to reinforce the training of the algorithm. In fact, via Facebook’s posts we were able to collect data points about a more natural and spontaneous way of writing, as well as the include patterns related to typos when posting from mobile. We have built an agent which is trained to answer 34 types of questions, with the full list of intents available in the table below.
Challenges we ran into
First of all, the design of the whole user experience flow was challenging. In the early stage of the process we asked ourselves several questions such as: “what would be the simplest way for the user to buy one or several products from the store?” or “how could we design a service that would effectively benefit both the vendor and the user at the same time?”. Another important design choice was the decision between the adoption of Dialogflow or Wit.ai as NLP service. Eventually we decided to use Dialogflow because: 1) it can be programmed to answer directly to user inputs without using a Webhook, thus resulting in a much faster creation of the agent, 2) it allows for bulk uploads of training phrases, thus allowing to easily increase the number of examples to ultimately improve accuracy.
Furthermore, the connection between different technologies was not as easy as expected because Toblee has to work in synergy with Whatsapp, Facebook Messenger, Telegram and Dialogflow. It took time, several trial and error iterations to be able to deliver the same user experience regardless of the messaging platform used.
Accomplishments that we're proud of
We are extremely proud that we gave birth to our first software product. It has been a remarkable experience, and the journey it’s just at the beginning. What is truly amazing is how steep the learning curve has been for us, and how every day has brought precious insights. From technical topics to pure business or UX discussions, Toblee has been developed in a synergetic environment that has stimulated sharing of knowledge and ideas. We have pushed our limits to new boundaries, we learnt topics that few months ago we would have never expected to become exposed to. That’s the beauty of getting involved to embrace the future. Every day is a new discovery, sometime you win and sometimes you learn, but perhaps the greatest achievement is that since Toblee has been conceived, no day has been like another.
What we learned
The list of all the things that we learnt could fit another entire document, but for the sake of conciseness we can describe with one line: we learnt the beauty and the power of diversity, in all its form. From coding related topics moving to the complete opposite of the spectrum of sales pitches, we got challenged and inspired. We learnt about our clients: their business, their habits, their way of communicating, their needs. We were able to listen, meditate, think and discuss internally about how we could deliver the best solution for them. We did the same with our end users. We learnt how to interact among ourselves too. Business people started to think as developers, and developers as business. We learnt what no course at university has taught us, we read solutions no book has ever provided. We learnt how to move the first steps into entrepreneurship.
What's next for Toblee
We have just moved a few steps down the path to accomplish Toblee’s vision of a more connected retail experience. On the immediate next steps, we plan to enlarge the number of integrations to enhance both the shopping experience (adding payments option) and on the owner experience (providing integrations with order management systems, thus allowing the order to be directly ingested into the shop system once accepted). Our infrastructure has been built with one clear point in mind: be scalable to reach as many customers as possible. We are now ready to move the first steps into this world and to run this beta test with some selected partners.
Feel free to contact us at email@example.com or visit our LinkedIn page.