The subject matter relates to a method and a device arranged for creating media content depending on consumption data. Media content, for instance video content, audio content or the like has been used as means for communicating between companies and customers for a long time. To give customers instructions how to use certain services or devices, companies have created media content which is sent to users, explaining these services and or devices. Such media content is static and always the same for each customer, independent of any information about the customer.

As regards utility companies, these companies are in constant contact with their customers. For instance, each time a utility issues a consumption bill, a certain amount of customers will for instance dial-in to a service hotline of the utilities to inquire about their respective invoice. The same applies for new customers, which have been contracted with a new contract. These new customers oftentimes do get in touch with the utility companies by phone, when they have certain questions about this new contract. It has been found that the questions customers pose are usually similar or the same. It moreover has been found that certain types of customers do pose the same questions, and the answers are always similar, however, the relevant consumption data and/or contract data of each customer is different. It moreover has been found that explanations to customers are better understood and better accepted, if the customer is approached on a personalized basis, i.e. if the explanation of the service or the bill is linked to the actual consumption data and/or contract data of the customer. Thus, the problem of approaching customers with media content is that such media content is not personalized. The result is that the user's acceptance and understanding of communication is low, if a communication is done by known media content. For this reason, it is a technical object of the subject matter to provide for automatic media content generation using customer information for each individual customer.

According to the subject matter, media content is generated using consumption data. For this reason, first of all a consumption dataset is obtained. A consumption dataset may comprise information about consumption of resources by a user. Consumption may be, for instance, consumption of electricity, consumption of gas, consumption of water, consumption of prepaid pay TV or the like. Any type of consumption can be user specific, resulting in different consumption datasets for different

consumers/users.

From the obtained consumption datasets, at least one parameter may be extracted. Parameters within consumption datasets may be, for instance, an amount of service used. For instance, an amount of service may be a fixed price for each electricity meter and/or a price per unit for each kWh of electricity used.

With the extracted parameter, an individual first media sequence is created. This first media sequence is individual depending on the extracted consumption parameter. This means that the media sequence is tailored to the exact information of the first parameter. The output of the first media sequence is for instance a video, which is individual for each parameter, e.g. individual for each customer.

In order to automatically generate media content, it is further proposed that the first media sequence is combined with a generic media content to obtain a consumption data enriched output media content. The generic media content is capable of being combined with the media sequence in order to generate one single output media content. Within this output media content, the generic media content may comprise all information which is common for all or a certain number of users, and the added first media sequence is user specific, i.e. relates to the consumption parameter(s) obtained for one specific user. By using the video sequence, each media sequence can be individualized to a certain consumption behavior of a certain user.

It has been found that media content can be overlaid to create one output media content. For instance, video overlaying technologies are well known in the art. These technologies allow merging two individual videos into one output video. Seamless overlaying results in that in the output video it is not possible to detect, which source a certain video frame stems from. Also, in one video frame, two sources may provide different content, which content is then merged into one video frame. Video overlaying technologies are for instance used in on-screen display or subtitling.

Moreover, video overlay technologies are used, for instance, when generating menus for DVD or Blu-ray technology. The overlaid object(s) may, according to the subject matter, be media sequence(s), which are generated depending on the first parameter.

It has been found that most of the media content may be generic and only some parts of the output media content can be consumption data dependent. Thus, the generic media content may have a duration which is longer than the duration of the media sequence. Thus, when overlaying the media sequence on top of the generic media content, only a part of the generic media content is enriched with the media sequence. It may, however, also be the case that the media sequence is merged into the generic media content by adding one or more video frames to the already existing video frames of the generic media content. It should be noted, that not only video frames can be merged, but also audio content or other content and the description relating to video frames may also apply to audio content or any other media content.

According to an embodiment, a second parameter is also extracted from the consumption dataset. The second parameter may relate, for instance, to different consumption data than the first parameter. In order to make the output media content even more customer specific, a second, different media sequence is generated depending on the second parameter. Then, also the second media sequence is combined with the generic media content. The combination technique can be the same as the combination between the first media sequence and the media content. It is preferred that both media sequences are combined with the generic media content at different time instances. That is, the generic media content can be enriched during different time intervals with the first and second video sequences.

According to an embodiment, the consumption dataset comprises at least one parameter, preferably more than one parameter of the set of the following parameters name and/or address of a customer, date and/or type of a contract, amount of resources used, in particular kWh electricity or m ³ gas, and/or a term of a contract, and/or a price per unit of resource.

Using the name and/or the address of a customer can, for instance, be used for generating a personalized first or second media sequence. Therein, the customer may be addressed personally, for instance, by verbalizing his name and/or by printing his name or the like.

The same may apply for the date and/or the type of a contract. The day of a contract may be, for instance, the date when the contract was signed. A type of a contract may contain information about certain contract options. The information may be used for making the video sequence more individual.

The amount of resources used can also be a parameter. Such a parameter may, for instance, state the amount of kWh a consumer has used during the last billing period. The same may apply for the volume of gas, for instance measured in m ³ . Such an amount may be printed and/or verbalized in the media sequence. For instance, a video sequence may have in spoken words information about the amount of resources used. For instance, within the video there may be an off comment, stating, for instance, "you have used 3625 kWh of electricity during the last 12 months". Another information may be an end of term of a contract. Such an information may be also verbalized in the media sequence by stating, when the contract ends.

Moreover, also depended on the type of a contract, the price per unit may be used as a parameter. The price per unit can also be verbalized during the media sequence.

The generic media content may be, however, not generic for all customers, but may be generic only for a certain type of customer. In such a case, the generic media content first is created based on generic sequences put together along as so called "story line". Which sequences are put together to create the generic media content may be dependent on at least one parameter of the dataset, e.g. at least one parameter defining a type of a customer. When the generic media content is generated, different media sequences, each being depended on a value of a parameter, can be appended to each other. Thus, for instance, for customers having contract type "A", a first sequence "A" can be used and for customers having contract type "B", a second sequence "B" can be used.

Then, if a customer uses for instance more than 2.000 kWh during a billing period (as an example of a parameter, a sequence "C" can be appended and if a customer uses less than 2.000 kWh during a billing period, a sequence "D" can be appended. This may be done until the whole generic media content is generated based on the different media sequences each being selected depending on the value of a parameter.

As has stated above, within the generic media content, for instance video or audio data may be used. A video content and/or an audio content may be used within the generic media content. At least one of the media sequences can also be at least video content or audio content. It may be possible to merge a generic media content being a video content with a media sequence only being audio content. Moreover, it may be advantageous if the generic media content and the media sequences are of the same content type, i.e. video content. According to an embodiment, a content of at least one of the media sequences contains at least information relating to the parameter and/or the value of the parameter. In particular, information relating to the parameter and/or the value of the parameter is reflected in the content using clear text. Moreover, audio can be used to verbalize the parameter or the value of the parameter. Also a combination of textual, graphical or audio content can be used for expressing the parameter or the value of the parameter.

According to an embodiment, the consumption dataset is linked to the customer dataset. It has been found that not all customers need to be addressed with output media content. Moreover, it has been found that certain types of customers are more inclined to inquire the service provider than others. Based on statistical data, it can be calculated which customer is likely to inquire the service provider, e.g. by phone. A selection of which customer is about to inquire the service provider can be done using the customer dataset. At least two variables can be extracted from the customer dataset for each customer. These two variables can be compared with statistical data of these variables. From this, similar customers may be detected. For these similar customers, the likelihood of an inquiry to the service provider can be provided based on statistical data. If a certain percentage of similar customers have inquired the service provider, then it can be decided that for the current customer output media content is generated. Thus, by comparing a certain customer with statistical data of customers it is decided, whether or not output media content is generated.

It has been found that only few variables are necessary to decide, whether a customer is about to inquire the service provider or not. These variables can for instance be billing amount, positive or negative bill shock, a period of being a customer, demographics, language of a customer or the like.

As has been explained above, a certain customer is compared with similar customers from a historical customer database. Similar customers are found by comparing customer variables with similar variables in the historical database. Among the variables, which are compared, there are for instance the name and address of a customer, the age of a customer, a type of a household of a customer, a type of a contract of a customer, a martial status of a customer and/or a creditworthiness of a customer. The customer dataset is used for extracting at least one of these variables, preferably more than one of these variables. If the variables are extracted, the values of the variables are used to compare the variables with historical data. Then, with similar customers having similar values of the variables, it is checked, if these customers have inquired the service provider in the past, in particular in response to a certain event. If a likelihood exists that a customer will inquire the service provider based upon a current event, the decision is made that the output media content is generated. Historical data can be updated with new information about user behavior. As will be described below, customer reactions can be tracked and used to update historical user data. Thus, there is a closed loop, allowing improving the historical database constantly.

For the output media content that has been generated, customer contact information can be obtained from the customer dataset. The customer is contacted using the output media content by transmitting the output media content as such, i.e. as a file, or as a link, e.g. as a hyperlink, e.g. an URL to the customer. The customer may then consume the media content. During consumption of the media content, usage information of the output media content by the customer can be monitored. It can be checked, whether the customer consumes the media content at all, whether the customer only consumes part of the media content and the time and/or the number of times the customer is consuming the media content. All this information may be fed back to the historical database.

It can be checked, whether customers addressed with the output media content actually did watch this media content. Moreover, it can be checked, whether such customers do inquire thereafter the service provider for instance by telephone or not. Thus, it can be checked, if the output media content did answer all the customers' questions or not. All this can be used within the historical database to improve the decision about whether and when to generate output media content for a certain customer, and to thus establish a closed loop, or a feedback loop to improve accuracy.

In order to enable comparison of statistical data with actual customer data, the usage data obtained is fed back into the customer dataset. Costumer dataset can then be added to the historical database for a further comparison with new customer datasets.

The statistical data is generated, according to embodiments based on the historical database. In particular, the statistical data reflect information about a number of customer inquires and/or a type of a customer inquiry and/or a time of a customer inquiry and/or a type of an event causing the inquiry. The statistical data could also have information about other types of services sold to the customer - other then gas and electricity, e.g. complaints or inquiries about solar panels, meters or any other asset, so that the datasets will become grow over time.

A type of a customer inquiry may be, for instance, a complaint about a service, a question about an invoice or the like. A time of a customer inquiry may be, for instance, a time span, which has lapsed between when the customer has received its bill and when he inquired. Moreover, a time of a customer inquiry can be the duration of his inquiry. Also an event can be the reception of a bill, a new contract, a

termination of a contract or the like.

According to embodiments, it is proposed that a customer is classified by comparing the customer dataset with the historical data. For the customer dataset, at least two variables may be extracted. The values of these variables may be used to look for customers, which had similar values. If these customers are found, it is checked, what is the likelihood of these customers to inquire the service provider, for instance by phone. Depending on the classification of a current customer, statistical information about customer inquiries of customers having a same classification is extracted. Thus, customers are classified depending on the value of their variables and historic inquiry information. Based on this information about customer inquiries, the output media content is generated or not.

In order to enrich the output media content with consumption data, it is proposed that consumption data is received from a smart meter metering the consumption data. Thus, the consumption data may be fed automatically into the consumption dataset. Still, it may be possible to obtain the consumption data from a billing center. The consumption data may be fed manually with meter readings. These and other aspects will become apparent from and elucidated with reference to the following Figures. In the Figures show:

Fig. 1 an evaluation of historic user data; Fig. 2 response probabilities for different user groups;

Fig. 3 a comparison of a user dataset with historic user data; generation of generic media content based on user data;

Fig. 5a a consumption dataset;

Fig. 5b generation of media sequences depending on parameters of the

consumption dataset;

Fig. 5c an enriched output media content;

Fig. 6 a feedback loop for distributing enriched output media content. In Fig. 1, there is depicted a plurality of user datasets 2. Within the user datasets 2, there are stored numerous variables describing consumers. Among these variables, there are, for instance, information about at least one of the name and address of a consumer, a type of a consumer contract, an age of a customer, a type of a household of the customer, creditworthiness of the customer and others. Within a user dataset 2, there may also be provided information about an inquiry of the user. In addition, this information about an inquiry by the user there may be stored a reason (event) for this inquiry. There may be certain customized reasons, for instance "new contract", or "bill", or "change of contract", or the like. These types of reasons together with information whether the user contacted the service provider based on this reason may be stored.

All user datasets 2 thus contain historic data about users and their behavior, in particular inquiries of the users to the service provider, in particular in response to certain events.

The user datasets 2 are used to create user categories 4a-4c and to use these to categorize users. Each user category 4a-c is defined by certain characteristics of the users. All user datasets, which fall within this characteristic, are grouped in the respective user category 4a-c.

Having categorized the users into the user categories 4a-c, it is possible to decide, what is the likelihood of an inquiry by a user within a certain user category 4a-c for a certain event. Such likelihoods are shown in Fig. 2. Fig. 2 shows the likelihood of an inquiry, for instance, in percent for different user categories 4a-c. For the sake of completeness, Fig. 2 shows more user categories 4d-f.

For each of the user categories 4a-f, and in particular for each event, it is evaluated, what is the percentage of users of this user category 4a-c that have inquired the service provider. This likelihood is expressed in percent and shown as a bar in the graph of Fig. 2. It may be possible to define a threshold, above which the likelihood of an inquiry is so high that the service provider wants to approach such kind of customer proactively. Such a proactive approach may be, for instance, by using the enriched media content.

Upon an event, a user dataset 6 is checked for each user as depicted in Fig. 3. A user dataset 6 is first compared with the user categories 4a-c and the user of the user dataset 6 is categorized into one of the user categories 4a-c. This may be done by checking, which user category is most similar to the actual user dataset 6. For each user category 4a-c, for instance a median value of the respective variables may be calculated. It may also be possible to use a logistic regression model which makes correlations to identify the main parameters of the model and to calculate the variables. Then, each variable may be compared with the variable in the user dataset and a difference in values may be calculated. By summing the differences and for instance weighting the differences in the sum, it may be found out which user category 4a-c suits best to the current user dataset 6.

In Fig. 3, the user category 4c fits best to the user dataset 6. Thus, user dataset 6 is put into user category 4c for further use. Then, the event is checked and the likelihood of an inquiry for this event for this user category 4c is checked, for instance within a graph according to Fig. 2. A can be seen in Fig. 2, the likelihood of a response for user in category 4c is above threshold 5. For instance a user fits into a top decile. Thus, it is likely that the user will inquire in response to the event. Thus, the service provider receives a signal to generate the output content. In the following, the output content is a video.

Fig. 4 depicts a flowchart for generating generic media content. Generic media content may, for instance, already be to some extent tailored to a certain customer type based on customer information. For this reason, certain variables from the actual user dataset 6 are extracted. These variables may, for instance a martial status and/or age of the customer. With this information, the video sequence may be generated as described below.

The video sequence has five intervals 8a-d. For video interval 8a, there is a video sequence 8a', which is a generic for all customers. Appended to this video sequence 8a', any one of video sequences 8b'-b"' can be appended, which selection is based on customer data. For instance, the age of a customer is analyzed from the user dataset 6 and three different age categories may exist. Depending on the category of the age, in the current case the video sequence 8b' is selected and appended to video sequence 8a'.

Then, the martial status of the user is extracted from user dataset 6 and depending on the martial status, one of the video sequences 8c'-c"' is appended to video sequence 8b'. In the current case, this is video sequence 8c'.

At the end of video sequence 8c', there is appended a generic video sequence 8d', which may be generic for all possible video sequences generated.

Thus, by using customer data, the generic media content may already reflect user information.

Then, this generic video content is enriched with consumption data, as is illustrated in Figs. 5a-c. A consumption dataset 10 may be obtained for the current user by comparing user data in user dataset 6 with user data in consumption dataset 10. Each user can be identified by a user identification, which can be stored on consumption dataset 10 and in user dataset 6. By this, consumption data can be linked to user data. Within the consumption dataset 10, not only the user identification is stored, but for instance also a type of a contract "123" and an amount of resources used "ABC". More information may be stored in a consumption dataset 10, however, for the sake of brevity, only two consumption data information are used here. Based on a extracted type of contract "123", a first video sequence 12 is generated. Within this first video sequence 12, the contract type "123" may be illustrated as graphics, text, symbols or using audio or the like. Then, the amount of resources consumed "ABC" is obtained from consumption dataset 10 and a second video sequence 14 is generated. Within this second video sequence 14, the amount of resources use "ABC" can be played back either as text, graphic, symbols, audio or any other type. Both video sequences 12 and 14, which contain consumption data "123" and "ABC" are then used to be overlaid over the video sequence 8a'-d'. This is illustrated in Fig. 5b.

It may be defined with which video sequence 8a'-d' video sequence 12 and video sequence 14 are respectively combined. In Fig. 5b it is illustrated that video sequence 12 is overlaid, combined, input, overlaid or in any other way merged to video sequence 8b'. Video sequence 14 is combined, input, overlaid or in any other way merged with video sequence 8c'. A resulting output media content 20 is generated as illustrated in Fig. 5c. As can be seen, video sequences 8a' and 8d' is enriched with consumption data resulting in video sequence 16 being a combination of video sequence 8b' and 12 and video sequence 18 being a combination of video sequence 8c" and 14. This output content 20 may then be forwarded to the customer, which is identified from the user dataset 6, as illustrated in Fig. 6.

The user datasets 2 storing historic data and being categorized in user categories 4a-c are stored in database 24. The consumption data and the actual user data 6, 10, are stored in a service provider database 26. From service provider database 26, user dataset 6 and consumption dataset 10 are obtained and output content 20 is created as has been described above. Moreover, for creating the output content 20, as has been described above, the user categories 4a-c need to be known from the user datasets 2, which are obtained from a data base 24.

From the user dataset 6, an address, for instance an electronic address is known for the respective customer. Output content 20 is sent to this address, either in the content of an electronic communication or as a link, for instance a hyperlink. A user receives at his device 22, e.g. a smartphone, tablet, computer, smart glass or the like, the communication. Once the customer consumes output content 20, he activates the respective link or opens the file storing the output content 20. This interaction of the user on his computer 22 with the output content 20 may be fed back to database 24. Therein, it can be stored that a user actually has consumed the output content 20. This consumption is an indication that the user was about to inquire the service provider and may be used in the statistical data for creating a graph as illustrated in Fig. 2. By feeding back the user reaction to the output content 20, it is possible to improve the user datasets 20 and database 24 and moreover to improve the preciseness of predicting whether a user is going to inquire a service provider based on a certain event or not. By means of the described method for creating media content, it is possible to automatically and directly contact customers based on their consumption data and their user behavior. This reduces the need for communication resources at the service provider.