Pakke.dks Chat Solution
When the digital chat solution handles 80% of customer inquiries
To most people who opened a business school textbook or heard the “Increase revenue and reduce cost” mantra, it might sound like a dream scenario – yet, it is what Ulrik Kjær Larsen, owner of pakke.dk, and his 12 man team who offer shipping services in Denmark, Sweden and Norway experience.
The secret to the IT chat solution lies in Pakke.dk analysing all incoming questions for customer service and determining what questions were support related and what questions were sales related.
After Ulrik and his team structured the historical data, they fed BotXO’s chat robot the questions and configured the answers they would usually give a customer via email, chat and telephone. The chat solution has been set up for both Pakke.dk, Billigpakke.no & Billigtpaket.se – in Danish, Norwegian and Swedish and supporting all three language versions in English.
Results within 24 hours
Big impact with little effort
“Of course, we had great expectations of BotXO’s starter kit – but not in our wildest imagination did we imagine that our new chatbot would halve our number of support tickets and relieve our customer service to the extent it does. Add to that the fact that we can also assert that the traffic coming through the chatbot has a conversion rate which is 4 times higher, and we almost have to pinch ourselves to believe it is true. But it is.” Ulrik Kjær Larsen, founder and CEO of Pakke.dk, says.
Switching from live chat to BotXO’s AI bot
Pakke.dk used to have a chat managed by a customer service agent. However, this solution wasn’t effective enough. The agent could not handle all incoming tickets. Moreover, the chat was available for customers during regular working hours. Pakke.dk was searching for a solution that could run 7/24 and take care of repetitive questions. Thus, customer service agents would be able to spend their time on more complicated inquires.
BotXO’s conversational chatbot was what Pakke.dk was looking for. Firstly, the bot is easy to build, with no coding required. It is fast and provides customer support 7/24. Equally important, Pakke.dk needed a bot speaking Danish. BotXO, unlike other bot companies, was able to create a Danish-language solution tailored according to Pakke.dk’s needs.
At the moment, the chatbot answers 80% of questions! Only 20% of inquires are handled manually. That is an enormous relief for customer service. The bot can handle 1000 customers at the same time, increasing efficiency of ordering process.
Since Pakke.dk receives many inquires on weekends, AI chatbot seems like a perfect solution, which improved customer experience and customer satisfaction.
With a digital chatbot, which can be easily trained for different languages, Pakke.dk considers to scale their business to other countries.
It always starts with data analysis
The data analysis forms the basis of the customer journey itself.
“When we start up a new collaboration with a customer, one of our first tasks is to get a handle on the data. Our task is to analyse and structure existing data collected from email chat support and match it against the FAQ. But in the case of Ulrik and his Pakke.dk team, they had already been keeping track of their data. Therefore, in regards to pakke.dk, our role was more about supporting the actual implementation and advising on the use of our chatbot SaaS chat solution,” says Farnaz Aref, former Chief Conversational Design Officer, BotXO.
Implementation of structural dialogues
The configuration of chatbots has much in common with configuring marketing automation
The implementation of chatbots means looking at the overall dialogue structure, establishing links between text dialogues and adding rules that determine what happens when a user asks a question. When the chatbot has been configured, typically as both a customer service and sales chatbot, it will be implemented in up to three different channels – on the website, in Facebook Messenger and in mobile phone apps.
The measurement of the entire customer journey
The next step for pakke.dk and BotXO is to link data, ensuring that the entire customer journey can be monitored as a whole – from traffic to initialisation of dialogues, to the links the user clicks in the chatbot and on through the sales funnel and out into the website to see which targets users hit.
“Obviously, when we experience this great a success with that little effort, we are interested in finding out what happens if we spend even more time on the chatbot. We can already see how we can strengthen the chatbot by making it smarter and answering more questions, so that is where we will start. Since we are already a data-driven company, we want to link our data with the data we can get from BotXO and work in a more structured manner going forward,” Ulrik Kjær Larsen says.
The chatbot technology is relatively new, why it obviously involves new ways to measure success and limitations.
“We are a young company with a new technology for existing markets. We have a good overview of our data, but did not really put it to use yet. The clients we have have been more interested in seeing what the chatbot technology really is and how it fits into their existing business. Now is the time to go for a more data-driven approach. Several of our current customers are already ready to work in a data-driven manner, and we are pleased that pakke.dk requests data-driven decisions and development,” says Martin Stahl, former CMO/CDO of BotXO.
BotXO has data, the customers have multiple sets of data, both from the customer service platform, web analytics, social media, email and NPS. This is the data to be given priority, used for optimisation and converted into a basis for decision making.
New KPIs and measurement concepts for a new technology
In the holistic view of all data sets, the linking of data and working in an even more data-driven manner, BotXO introduces new KPIs which the generally data savvy marketer does not normally work with. These are e.g. “Fallback” and “Human take-over”
“Of course, we are working with a new technology that involves automated dialogue, and even if we implement NLP now in the second quarter, there will still be areas which a chatbot must first learn to understand. And in the specific cases where the chatbot has its limits, we work with a fallback which refers the chat to a customer service representative (Human take-over). That is, when the chatbot runs out of answers, our fallback is to ask users if they want to speak to a customer service representative. While this of course means that we do not drop the customer, it also tells us where the limitation to each chatbot lies.”
“Once we have data in the form of questions, the context to which the questions relate, time stamps etc., we can transfer new knowledge and learning to the chatbot. This is our foremost area for taking action. Data in the form of number and types of customer service inquiries and the number of sales conversions and traffic to the chatbot will also play a role when more knowledge is to be transferred to the chatbot,” Martin Stahl concludes.
Written by Martin Stahl