How to use a chat-bot to stand out?
Nowadays, customers prefer to have a personalized relationship with the retailer, because their needs are every day more specific. Another aspect is that customers wish to spend less time looking for the right article.
Today’s e-commerce websites need a very good user experience to catch the clients. The user experience will determine how the client feels while navigating on the website and remembers this experience for future purchases. Or it can determine how the client is impacted by our website in comparison to the others and help him remember our brand.Nowadays, customers prefer to have a personalized relationship with the retailer, because their needs are every day more specific. Another aspect is that customers wish to spend less time looking for the right article.
Too many choices
This leads to some problems. The retailers need to have multiple articles to be able to fit the needs of their customers. But they don’t scroll all over the website or application to find what they need. To simplify this, we can create categories where each article can fit in. But when you have a lot of articles, this means also sub-categories, which means different layers where you can lose the client, as in a labyrinth.
In one of our cases, we designed a website that would allow a customer to get a quote for a vehicle repairs. This meant that the client had to know the details of the vehicle problem if he wanted a precise quote. However, not everyone knows how each single spare part of the vehicle works, which makes it difficult to target a big audience.
One choice to rule them all
Instead of leaving the customer the freedom to navigate over all the products, we chose to have a single entry point: a chat-bot. What’s a chat-bot? It’s a simulated conversation with a chatting interface. Simulating a conversation will allow the customer to be more comfortable as if it were facing a real person.
Let’s now see the different parts of the chat-bot.
The chat-bot will be the single entry point. It’s the way to start a conversation with the application. It’s the way the application will try to know what the client wants. The chat-bot was presented from the homepage. Leading the client into a set of questions.
The chat-bot asks multiple questions (the less the better) to know what the client really wants. But the chat-bot should also prepare the answers. It should display a list of possible answers. Don’t allow the user to insert any text he’d like, this will complicate the search system. Letting a free input text allows the user to insert whatever he wants. It could be an expected answer, but it could also be a long text describing what it wants. Or the client could even misunderstand the question and answer with another context.
The questions should focus on the need or the problem, not on the articles themselves. If you let the client tell you the article he wants, you may not know what he really wants. Looking for his needs we may offer him a product that better fits his demand.
The system behind the chat-bot was a decision tree, where each question/answer will lead to another subset of questions/answers until the final product is found. Choose wisely the questions and answers, to avoid having too much and bore the client.
Sometimes, some articles are very common to sell (some spare parts, or some common reparations), but the chat-bot asks the user to follow the complete workflow of questions/answers until the final quote. To avoid the complete workflow, we allowed the system to be able to start at any point. This means that it should accept some kind of parameters to indicate the state of the decision tree where it should place the client. This kind of behaviour was very interesting with the landing pages which had a promotion of some article.
The landing pages were used for temporary promotions. Once we had done all the work to attract the customer into the landing page, we didn’t want to lose it by doing all the workflow of the chat-bot. So we let some parameter guide the client from the landing to the chat-bot at some advanced position in the decision tree, or with an already prepared quote.
Finally, after some questions/answers to the customer, the system found the article he was searching for. Now it’s time to present it, with the summary of the questions and his answers. We showed the customer that the article matched the answers he made. If it was not the case, the client could go to a previous step and change his answer. But maybe a change in his responses will request him to perform the workflow again, as new questions have to be made. But the objective at the end of the chat-bot was to find the ideal article for the client, so we had to be sure to ask the correct questions.
Nevertheless, we can’t always have all the answers. Maybe the answers proposed to the customer didn’t match his needs. This is normal, as we can’t handle all the possible answers given by a person. For those cases, we had an emergency exit. That means having an answer as “Others” which redirects to a workflow where the client can contact some real person to guide him.
The usage of this new section must be monitored, to study if we can create another category that can match an incoming demand.
The main purpose of having a chat-bot is to guide the clients into our articles. Because we knew that vehicle reparation is an area that most people fear and are not very comfortable with. This way, we avoid losing them with a complex search system and interact with them as if it were with real people. Because a client lost halfway is a lower conversion rate, it’s short-term lost sale and probably a long-term lost client.