Mediamarkt logo

Smart general and task-based chatbots

In 2022, we set up four chatbots for MediaMarkt to support the electronics retailer’s customer service. At the beginning of this year, we started by setting up a chatbot for the VAT campaign and a general chatbot for the Belgian website. In autumn, another chatbot for Black Friday and a general chatbot for the Dutch website followed.

Wanted: customer service support

As one of the largest chains in consumer electronics in Belgium and the Netherlands, MediaMarkt gets hundreds (even thousands) of customer questions daily. Many of its customers call or email customer service. Others look for answers in the extensive FAQ.

The better part of customers eventually finds what they’re looking for. But, in general, the search journey lacked efficiency.

That’s why MediaMarkt was on the lookout for a partner to support its customer service. In order to do that, they had general chatbots and campaign-based chatbots in mind. This way, they could: 

  • help customers find an answer more easily; 

  • offer more when it comes to customer service;

  • lighten the workload for the customer service department.

Want to know more about this case? Let's talk
Challenge

Lightening customer service workload

Solution

Two general chatbots and two campaign-based chatbots

About MediaMarkt

Since its inception in 1979, MediaMarkt has grown to become the largest electrical specialist retailer in Europe. The originally German chain now has more than 800 shops (30 of which are in Belgium) and continues to grow.

Chatbots in numbers

125 000

messages received (and counting)

48 000

conversations

80%

of conversations on the Belgian and Dutch websites were successful

Hi, how can I make your job easier?

In 2022, we used Chatlayer, an AI platform, to set up four different chatbots for MediaMarkt:

  • A general chatbot communicating in Flemish and French for the Belgian website

  • A general chatbot communicating in Dutch for the Dutch website

  • A chatbot for the VAT campaign (early March) in Belgium

  • A chatbot for Black Friday (late November) in Belgium and the Netherlands

Set-up and implementation in phases

All bots were set up and implemented in the same way. We started with a workshop to determine the workflow, scope, and technical validation of requirements for all chatbots. 

In the second phase, we focused on the chatbots' conversational design and made sure they sounded human. MediaMarkt’s target audience was our main priority. Keeping the tone of voice consistent and using the right visuals to give the bot a natural feel was a close second. 

Once the conversational design was completed, we started programming the chatbots. For instance, we taught the chatbot algorithms to understand conversation flows, the meaning of certain phrases and the underlying context of a conversation. 

After programming, we tested the bots extensively by allowing a test audience to use them. Based on the feedback they gave us, we tweaked the chatbots and algorithms.

In the final phase, after finetuning the chatbots, we launched and promoted them to make sure the target audience would know about them. We also gave the bots a name for the launch. MediaMarkt website users, meet Marie and others.

Questions and answers found

The general chatbot for the Belgian website and the chatbot for Black Friday are both live at the moment of writing (late November 2022). The chatbot for the VAT campaign was taken offline after the campaign ended. The general chatbot for the Dutch website hasn’t been launched yet.

When we launched, the chatbots were able to handle questions on the MediaMarkt website independently. If necessary, they were able to get customers in touch with a customer service employee (human handover).

After launching, iO monitored the chatbots closely. For example, we analysed user data and used it to improve the chatbot funnel. In addition to that, we also had a look at the most common questions and search intent. These insights, too, were used to improve the chatbots.

Chatbot stats

VAT campaign

During the Belgian VAT campaign, the chatbot had over 500 conversations, of which 57% were in Dutch and 43% were in French. Overall, the VAT chatbot received over 1,000 messages.

  • The bot was able to answer 46% of questions immediately.

  • For 27% of messages, users found an answer using chatbot buttons.

  • For (another) 27% of messages, the bot needed clarification or got the customer in touch with an employee.

The bot users were generally active on Wednesday at 7 pm and Thursday from 9 to 10 am.

General chatbot Belgian website

The general chatbot for the Belgian website has been live since May of 2022 and has since then had over 2,500 conversations, of which 65% were in Dutch and 35% were in French. By now, over 9,000 messages have been sent.

  • The bot was able to answer 36% of the messages immediately.

  • For 42% of the messages, users found an answer using chatbot buttons.

  • For 22% of the messages, the bot needed clarification or got the customer in touch with an employee.

After a chat, the bot gets customers in touch with customer service – if necessary. Bot users are usually active during the day, with clear peaks on Monday at 11 am and on Tuesday from 9 to 11 am.

Black Friday campaign Belgium

For the Black Friday campaign, the Belgian and Dutch websites got their own bots. The Belgian Black Friday bot had almost 10,000 conversations, of which 55.5% were in Dutch and 44.5% in French. Around 20,000 messages were sent in total.

  • The bot was able to answer 33% of messages immediately.

  • For 53% of the messages, users found an answer using chatbot buttons.

  • For 14% of the messages, the bot needed clarification or got the customer in touch with an employee. 

The bot users were active both during the day and at night, with peaks on Friday 25 November from 7 am to 5 pm and Monday 28 November from 9 am to 3 pm.

Black Friday campaign in the Netherlands

The Dutch Black Friday bot had over 35,000 conversations and received nearly 95,000 messages.

  • The bot was able to answer 30% of the messages immediately.

  • For 53% of the messages, users found an answer using chatbot buttons.

  • For 17% of the messages, the bot needed clarification or got the users in touch with an employee.

The bot users were usually active during the day, with peaks on Tuesday 22 November from 9 am to 1 pm, Friday 25 November from 8 to 10 am, Sunday 27 November from midday to 7 pm and Monday 28 November from 9 am to 1 pm.

Want to talk about your project?

We're excited to see how we can help you.