Culstaar is Parisian startup specialized in Natural Language Understanding. Initially focused on Search Engine Optimization, we developed tools to analyze Search data and understand queries in natural language. We have combined our SEO knowledge and the machine learning algorithms developed by our team to offer our clients a new service. For this interview, we’ve talked with Clustaar’s CEO and co-founder, Philippe Duhamel, on how they solve support problems at Clustaar.
How do you get your first customers and how did you start building your product?
We got our first customer through a network campaign on AdWords before we even had a product. They wanted a chatbot and we helped them create one perfect for their needs. We then onboarded clients from another line of our business who we knew wanted chatbots by just letting them know that the service was available. That’s how we got our first five customers.
Once our platform was ready to starting officially selling the service, we launched an AdWords campaign. When a potential client is interested in our services, we call them and will even meet face-to-face. This makes a huge difference because it helps develop a relationship over time. When someone calls, we answer the phone and provide them with a level of care that you do not usually receive.
Talking about chatbots, what exactly are you selling to the customers? What’s the idea behind your tool?
We were completely driven by the technology in the beginning. Our services were introduced in a series of steps, starting with our technology. Then we built our chatbot platform to integrate with the existing technology.
From there, we started meeting with potential customers and focusing on the customer service aspect in order to find a problem we could solve for a small segment of the market.
Then we ended up creating this platform for customer support automation using chatbots that we are now selling. It’s typically used for SaaS or software companies. There are bots inside websites or apps that are trained to answer questions sent by users. Basically, it’s requiring requests. The chatbots are used to address 80% of the questions posed by clients, which is super helpful in easing the burden for the support team.
How does support function look like in your company?
It’s a mix of bots and people. Our chatbot was designed to be a very basic support bot that answers questions and turns customers over the support team when faced with a more complex question. It also takes on weekend duty, so if someone poses a complex question on a day where the support team is not active – it leaves a message and sends a notification so that the team can reach out Monday morning.
The response that someone gets from the bot completely depends on the question asked.
When you hire support employees maybe somebody from a marketing team or from the other teams are working on the support. Usually, it doesn’t tell you whether that someone is going to make the customer happy. So what are these essential customer skills that you are searching for?
Our situation is a bit peculiar, as we are at the beginning of launching a product. So what I ask from the customer service team today is that they are patient, empathetic, and driven to serve the needs of our clients. No matter how large or small the client is, we are dedicated to working with them in a timely matter in order to help make their business run smoother and keep them happy.
So you said you tried to find the perfect persona already in the chat, how do you segment your customers? Do you use third party tools? How do you know right away the prospect is right for you and you can continue talking with them?
It’s really simple. We use a series of tools some tools that help us analyze the new accounts created on our platform. We look at the email address used to create the account to see if it has a corporate affiliation. If it does, we look at the company and see if that fits within our target market. If for example, it’s a tech company – say SaaS online business, then we know that we should reach out to them. That’s the first step.
If it’s not a corporate email, we try to identify the person to see what company they are working for using LinkedIn.
We use two other tools to help us qualify leads based on their email address. One is an intercom mix and the other is called a sales machine. We use them as often as possible in order to identify active users and determine if they are in the space we are focusing on.
The first few steps are manual and then it moves into a more automated workflow using linkspam and sales machine.
How do you measure the ROI on the customer service as well as customer happiness?
We don’t measure ROI right now. Instead, we focus on two main objectives – selling as many enterprise licenses as possible and building efficient bots. We aren’t in the frame of mind to measure ROI yet.
Switching from ROI to self-help support you can reduce a lot of costs. What’s the cost of self-help support or do you use any tools or considering it or you don’t have enough support tickets to use for now.
It depends. When the support team isn’t there, we have the bot direct users to help pages. For example, if a user is trying to reset their password, it’s a simple redirect. If there isn’t a page created that will answer their question, the bot will direct them to a search engine in order to bring back accurate results.
Our support team is busy during the off hours and aren’t available to help users, so the chatbot is an excellent solution. It also helps those users who don’t really want to ask direct questions yet. You have to build this type of product over time as you learn what questions the users are asking and how to best answer them, but everyone seems happy with a self-help option.
Something to keep in mind is that the help pages need to be very efficient with videos or gifs in order to actually help the user find the answer quickly.
How do you empower your employees with support documentation? Do you store them somewhere or are you using Zendesk? How do you build an update on your knowledge base? So probably have statistics or insights when people searching for chatbots.
All of our customer success agents and team members are free to create help pages. If they identify a specific problem or request, we encourage them to address it right away. The help pages are created in English and use photos to help the client better understand what we are suggesting.
We use Intercom to connect with our users and see what they are doing on the platform. This is a very important key to the chatbot universe, as we have the potential for better understanding user questions.
The first technology is to understand markets through queries. Each time we showed our clients all of the different questions that people were asking in Google about their markets, they were very surprised. They had no idea that people were looking for these things.
When people have a problem, they typically go to Google which provides us a lot of insight for specific markets. We do the same thing with chatbots. The users go to a chatbot and ask an array of questions. We record all of the interactions and analyze the data to make sure that we’re addressing each of the needs posed by the users. New problems or questions are not already linked, so we use these new queries to enrich our current knowledge base with the new questions that people are asking. We also are proactive in identifying new questions that allow us to create better content.
As soon as you understand their pains, you are able to please your customers.
You mentioned intercom what’s the difference between your tool and intercom? What’s the distinction between your service and theirs?
On the knowledge base, there’s no other between us and them. We use their knowledge base. Intercom has live chat, we have our own web chat. But intercom is a tool to manage and have a conversation with users and distribute conversations between an agent or sales agent.
Intercom is basically trying to automate their system so the live chat can answer in chatbot style way. Our tool gives you functionalities and features to build a very complex bot. We have easy to use yet sophisticated scenario builder that we can truly build like a tree but more complex. In intercom, you can automate answers but you cannot build a complex bot.
For example in our platform, we can build a bot within an app like some other customers doing today so you get into the app and the bot would be inside the app. You cannot do that with intercom or Zendesk.
One of the best ways to connect with prospects is to connect stories with existing customers but unfortunately, it’s not that simple to ask for a testimonial. How to ask for a testimonial and what makes a good testimonial?
Before we start asking for testimonials, we’d like to build a base of customers and get to know their trends. Ours is pretty narrow right now with only 22 clients using our platform.
Once established, you have to find a client that is willing to explain the value that you are providing to their business and explain it in a way that will help others understand why they should choose to work with your brand.
We don’t have testimonials yet but are working on this. In the meantime, we write long articles that try to solve the problems we see in our target market in order to provide insight and knowledge through helpful content.
What would be the number one advice you would give to a support junior manager before starting position?
I always remind my team to think of every situation as if they were the client. They should be proactive and ask before she or he has a problem. This requires really stepping into your client’s shoes and understand their services.
Even if we’re overdoing it in the beginning, we work to pamper all of our clients. If it’s something we need to do, do it. Even if it consumes resources. Don’t ask, just get it done.Subscribe to blog via Email