In this decision, the European Patent Office considered a particular multidimensional data structure with a hierarchy of levels for each dimension to provide a technical effect. Here are the practical takeaways of the decision T 1159/15 of 6.4.2020 of Technical Board of Appeal 3.4.03:
This European patent application generally relates to mathematical models that are created based on stored information, variables and assumptions (conditions). The variables and assumptions are modified and several candidate models are generated and evaluated. Based on these evaluations, one of these candidate models is selected as the final model and is used for forecasting purposes.
The application mentions models related to sales of a product as one example. Based on various variables (e. g. price, geographical distribution, advertisement cost) and assumptions (e. g. higher prices decrease sales or increased advertisement costs increase sales) models attempting to estimate future sales are generated and used in order to create a business plan.
Here is how the invention is defined in claim 1:
Claim 1 (sole request)A system configured to determine a final model operable to be used to forecast information for an objective, the system comprising:
a variable determination module (201) determining at least one variable operable to be used for the final model and determining a modification to the at least one variable;
an assumption determination module (202) determining an assumption operable to be used for the final model, wherein the assumption includes a transformation for the at least one variable describing how the at least one variable impacts the objective or how the at least one variable impacts another variable operable to be used in the final model, and the assumption module determines a modification to the assumption;
a model generator (203) generating a candidate model using the at least one variable and the assumption, and generating a new candidate model using at least one of the modified assumption, the new variable and the modification to the at least one variable; and
a model evaluation module (204), executable by a computer, and determining a statistical measure and an indication of relevance for the at least one variable in each of the candidate model and the new candidate model, wherein one of the candidate model and the new candidate model is operable to be selected as the final model based on at least one of the statistical measure and the indication of relevance for the at least one variable in each of the candidate model and the new candidate model;
wherein the at least one variable comprises a multidimensional variable, and each dimension includes a plurality of levels, and the variable determination module identifies the dimension and a level of the plurality of levels for the at least one variable, and wherein each dimension is organized in a hierarchy; the system further comprising:
a multidimensional data storage system storing information for models generated by the model generator, including the candidate model, the new candidate model and the final model, wherein the multidimensional storage system uses a meta data layer (401) and a data layer (402) to store the information,
the meta data layer (401) storing a relationship between the at least one variable and the objective, an indication of the at least one variable, the dimension and the level for the at least one variable, and the assumptions for each of the candidate model, the new candidate model and the final model; and
the data layer (402) including data for the at least one variable in each of the candidate model, the new candidate model and the final model;
wherein the meta data layer (401) stores aggregation rules for the at least one variable, and the storage system is configured to perform multidimensional queries using the aggregation rules stored in the meta data layer (401); and
wherein the data layer (402) includes data that is at the lowest level of each dimension, wherein the aggregation rules determine how to aggregate up from a lower level in a hierarchical dimension to higher levels in the dimension and what transformations to apply for each level.
Is it technical?
The examining division had considered that claim 1 comprised technical and non-technical features, and that the only technical feature of the claim was a general purpose computer (as implied by the feature “a model evaluation module (204) executable by a computer…”). All the other features of the claim related to a business method as such.
According to the examining division, such a general purpose computer was so well-known before the priority date of the application that it did not require written evidence. Thus, No prior art search was carried out during the first instance procedure.
The appellant contested this decision and argued that at least the claimed multidimensional data storage system storing information for the models which uses a meta data layer and a data layer to store the information were technical features. According to the appellant, these features defined a particular way of storing data in the data storage of the claimed system which was not the “notorious” way data would be stored in a general purpose computer. Hence, a prior art search should have been carried out.
Here is what the board of appeal decided:
The defined data storage is multidimensional and comprises a meta data layer and a data layer to store the information. Variables are stored in the data layer. These variables have dimensions (attributes) organised in a hierarchy. The hierarchy may include sub-attributes or levels for each dimension. For example, one dimension may be geography and the levels in the hierarchy may be country, region, city and zip code (see paragraph  of the application).
The meta data layer stores, among others, aggregation rules for the stored data. The aggregation rules describe how to aggregate up from a lower level in a hierarchy to a higher level and what transformation to apply for each level (see column 9, lines 14 to 25 of the application as published).
This configuration enables the system to respond to multidimensional queries across different levels in the hierarchies (see paragraph  of the application).
Moreover, as defined in claim 1, the data storage system stores data at the lowest level of each dimension and uses the aggregation rules to determine how data are to be aggregated up to hierarchically higher levels in the dimension (see also column 9, lines 25 to 33 and paragraph  of the application).
In the board’s view, the defined data storage contains two types of data. Firstly, data encoding cognitive content, such as information related to variables, assumptions etc. These data are used in the generation of the models. Secondly, the aggregation rules, which are not related to any cognitive content but are instructions related to the operation of the system when responding to queries. These data could thus be characterised as “functional data” (see also T 1194/97, OJ EPO 2000, 575, Headnote II and Reasons 3.3 to 3.5; T 425/03, Reasons 6.2 and 6.3).
The features of claim 1 identified above define thus a particular multidimensional data structure with a hierarchy of levels for each dimension, in which data are stored at the lowest level of each dimension. Moreover, the data structure stores instructions on how the stored data are to be aggregated up to higher levels of each dimension (see also paragraphs  and  of the application).
In the board’s view, these features provide for a technical effect that goes beyond the “normal interactions” within a computer executing a business method, because they define a particular way in which data are stored, retrieved and processed, which affects the storage space used and the speed of processing.
This would be a “further technical effect” so that these features are to be regarded as technical features and not as part of the non-technical (business) features of the claim.
The board points out that the assessment of the technical effect(s) obtained by the identified technical features, i. e. whether there is less storage space used or the query processing speed is higher, involves a comparison with the state of art and belongs, hence, to the discussion about inventive step. It is established case law and practice that assessment of the technical character of the claimed-subject matter is to be carried out without any consideration of the state of the art.
It follows from the above that in the assessment of inventive step of the claimed subject-matter, the identified technical features should not be included in the non-technical aim that is given to the skilled person for implementation.
Moreover, the board is also of the opinion that these technical features define a particular way of storing, retrieving and processing data, which does not fall under the generic definition of a general purpose computer with the corresponding data storage. In the board’s view these features cannot be considered as being notoriously well-known technical features for which no documentary prior art evidence is necessary.
Therefore, the case was remitted back to the examining division for further prosecution, including carrying out a prior art search.
You can read the whole decision here: T 1159/15