T 1954/08 (Marketing simulation/SAP) 06-03-2013
Download und weitere Informationen:
Predicting marketing campaigns using customer-specific response probabilities and response values
Summary of facts and submissions
I. This appeal is against the decision of the examining division to refuse European patent application No. 04727275.2, entitled "Predicting marketing campaigns using customer-specific response probabilities and response values". The application was filed as international application PCT/EP2004/003930 and published as
A2: WO-A2-2004/090765 (21 October 2004).
II. The examining division refused the application for lack of inventive step (Article 56 EPC 1973) over a notorious data processing system comprising a plurality of terminals and computers connected via a communications network. They considered the marketing prediction method as a business scheme, an abstract administrative plan, an intellectual scheme and a mental act for performing mathematical steps. The division did not identify any technical detail beyond the use of commonplace communication and data processing means.
III. The appellant requests that the decision under appeal be set aside and that the case be remitted to the department of first instance with an order to grant a patent on the basis of one of four amended claim sets (main request, 1st to 3rd auxiliary requests) filed with the statement setting out the grounds of appeal (24 September 2008). Oral proceedings have been requested on an auxiliary basis.
(a) Claim 1 according to the main request reads:
"1. A computer-implemented method of predicting outcomes of marketing campaigns, the method comprising:
determining a response probability for each of a plurality of customers, the customers being intended targets of a marketing campaign;
determining a response value for each of the customers that indicates a predicted value of a response to the marketing campaign by the customer; and
predicting an outcome of the marketing campaign using the response probability and the response value,
wherein at least one campaign step (206) in the marketing campaign comprises a plurality of alternative campaign elements (207, 208, 209),
further comprising assigning the customers to the campaign elements (207, 208, 209) using an optimizing algorithm, and
wherein the optimizing algorithm assigns and reassigns the customers to the campaign elements (207, 208, 209) while evaluating the predicted outcome of the marketing campaign, but does not reassign a customer to a campaign element to which the customer has previously been assigned, and where each assignment of a customer to a campaign element is recorded in a binary map, such that the optimizing algorithm provides a best goal value for the marketing campaign."
(b) The first auxiliary request appends the following paragraph to the end of claim 1 of the main request:
"and where the optimizing algorithm is terminated after a user-defined number of customer reassignments does not improve the most recent best goal value."
(c) The second auxiliary request appends the following three paragraphs to the end of claim 1 of the main request:
"the customer reassignments are recorded in a list after finding the most recent best goal value, where
the optimizing algorithm is terminated after a user-defined number of customer reassignments does not improve the most recent best goal value, and where
upon said termination, the assignment of customers to campaign elements corresponding to the best goal value is determined by reversing all the assignments made since finding the most recent best goal value."
(d) As compared to the second auxiliary request, the third auxiliary request inserts four paragraphs in the middle of claim 1 so that the complete claim reads (italics added to highlight the inserted paragraphs):
"1. A computer-implemented method of predicting outcomes of marketing campaigns, the method comprising:
determining a response probability for each of a plurality of customers, the customers being intended targets of a marketing campaign;
determining a response value for each of the customers that indicates a predicted value of a response to the marketing campaign by the customer; and
predicting an outcome of the marketing campaign using the response probability and the response value,
wherein the marketing campaign comprises at least first and second campaign steps, and wherein predicting the outcome of the marketing campaign further comprises:
using the response probabilities for the plurality of customers to predict a number of responses to be received if the first campaign step were performed toward the plurality of customers;
selecting a target group (168) of customers from the plurality of customers using the response probabilities, where a customer is included in the target group if a randomly generated number is less than the customers response probability, the target group (168) being substantially equal to the predicted number of responses; and
predicting an outcome of performing the second campaign step toward the target group (168); and
wherein at least one campaign step (206) in the marketing campaign comprises a plurality of alternative campaign elements (207, 208, 209),
further comprising assigning the customers to the campaign elements (207, 25 208, 209) using an optimizing algorithm, and
wherein the optimizing algorithm assigns and reassigns the customers to the campaign elements (207, 208, 209) while evaluating the predicted outcome of the marketing campaign, but does not reassign a customer to a campaign element to which the customer has previously been assigned, and where each assignment of a customer to a campaign element is recorded in a binary map, such that the optimizing algorithm provides a best goal value for the marketing campaign, where
the customer reassignments are recorded in a list after finding the most recent best goal value, where
the optimizing algorithm is terminated after a user-defined number of customer reassignments does not improve the most recent best goal value, and where
upon said termination, the assignment of customers to campaign elements corresponding to the best goal value is determined by reversing all the assignments made since finding the most recent best goal value."
(e) Technical problem presented in the statement setting out the grounds of appeal
The skilled person was a computer scientist or a computer programmer who received the task of computing the outcome of a marketing campaign comprising a plurality of campaign elements. The business manager who wanted to use the prediction method might require the optimising algorithm to provide a best goal value for the marketing campaign. However, in order to accomplish that task, the computer scientist had to design and implement an algorithm which computed the optimum result. Thus, the choice of algorithm was not dictated by business considerations but was connected to the particular manner of implementation. The algorithm taught a technical professional how to find an acceptable best goal value in a reasonable amount of time. Thus, the skilled person was confronted with the objective technical problem of computing the best goal value for a marketing campaign efficiently and reliably.
(f) Inventiveness argumentation presented in the statement setting out the grounds of appeal
A general-purpose computer was provided with additional functionality without simply choosing an existing method. Claim 1 taught the skilled person how to design and implement an optimising algorithm which finds the best goal value efficiently and reliably. The innovative algorithm was specifically tailored to avoiding a sub-optimal assignment of customers to elements of a marketing campaign. Neither existing computational algorithms (e.g. iterative improvement or simulated annealing), nor mathematical formulae (e.g. a formula for calculating an absolute maximum) suggested the claimed combination of steps.
IV. The Board summoned the appellant to oral proceedings, scheduled for 6 March 2013. In an annex to the summons, the Board voiced its preliminary opinion that claim 1 (all requests) did not involve an inventive step over a general computerised method for processing data according to any existing mathematical algorithm. Finding a maximum revenue or profit value (A2, page 2, paragraph 2) of a simulated marketing campaign appeared to be a commercial rather than a technical purpose. Therefore, the iterative mathematical algorithm according to claim 1 remained a mere mathematical contribution which did not enter into the examination for an inventive step.
V. In response to the summons, the appellant defined an objective technical problem as how to implement a method of predicting outcomes of marketing campaigns on a computer. The manner of implementing such a method required technical considerations. While the skilled person had various ways to implement such a method on a computer, the solution of claim 1 was influenced by technical limitations of a computer (reference to T 258/03-Auction method/HITACHI, OJ EPO 2004, 575, reasons 5.8).
In particular, a human seeking to build a target group of customers would not consider using random numbers. Surprisingly, that method was more efficient than a human approach, particularly for a large data set (i.e. a large number of customers).
The use of a binary map in the context of an optimisation algorithm made it more likely that an optimal set would be determined, and enabled the determination to be performed efficiently. The binary map was used in consideration of technical limitations of a computer system, since the intuitive approach of a human facing a similar problem was difficult to implement in a computer system.
A list of reassignments to return to a previously found goal value (when subsequent iterations failed to find a higher value) was more efficient than storing each set of customer assignments to campaign elements and showed consideration of the memory limitations of a computer system. A set of customer assignments to campaign elements could require storage of one million values (A2, page 16, lines 16 to 18), whereas the method might terminate if 10,000 reassignments did not improve the most recent best goal value (A2, page 17, lines 4/5).
VI. By a letter received 12 February 2013, the representative informed the Board that neither the appellant nor the representative would attend the oral proceedings. A decision according to the state of the file was requested.
VII. The Board held oral proceedings in the appellant's absence on the appointed date (6 March 2013) based on the above-mentioned four requests (filed on 24 September 2008).
Reasons for the decision
The application
1. According to the introductory portion of the description, companies are often interested in trying to predict the outcome of a marketing campaign before it is carried out (A2, page 2, paragraph 2).
However, target groups of customers can be very large (several hundred thousands, or even millions). Therefore, even a modest number of alternative campaign activities may result in a great number of possible combinations of customer/activity assignments. It would be easy to simply assign the customers to marketing activities at random, but this approach is not very likely to result in optimum results for the marketing campaign (A2, page 2, paragraph 3).
2. The application seeks to improve on iterative algorithms which have been used to find a global maximum of a goal value of an envisaged marketing campaign (A2, page 5, paragraph 2; page 15, line 11 to page 16, line 23; page 22, paragraphs 2 and 3; page 24, paragraph 2; page 29, paragraph 1; original claim 12). The iteration represents a "simulation" of the marketing campaign (A2, page 20, paragraph 2; page 29, last paragraph).
Two conventional optimisation algorithms (Greedy approach; Taboo search) are said to get stuck at local maxima and may not be able to locate an overall, global maximum (A2, page 2, line 20 to page 4, line 12).
3. To prevent the search algorithm from revisiting local maxima, the application proposes to memorise (in a binary map 164) parameter combinations (i.e. assignments of marketing activities to customers) which have been considered and should not be reconsidered (A2, page 8, lines 16 to 21; page 16, paragraphs 2 and 3; page 22, lines 9 to 13; page 24, lines 13 to 16).
Third Auxiliary Request
4. Claim 1 of the third auxiliary request contains all the features of claim 1 of the other requests.
Article 56 EPC 1973 - Inventive step
5. In the light of Article 52(1)(2)(3) EPC, an inventive step according to Article 56 EPC 1973 requires a non-obvious technical contribution (T 641/00-Two identities/COMVIK, Headnote 1, OJ EPO 2003, 352; T 1784/06-Classification method/COMPTEL).
The use of computers for automation purposes is technical but commonplace.
A mathematical algorithm may become a technical means, i.e. it may go beyond a mere mathematical contribution, if it serves a technical purpose (T 1227/05-Circuit simulation/INFINEON, points 3.1, 3.2, OJ EPO 2007, 574).
6. However, anticipating a maximum revenue or profit value (A2, page 2, paragraph 2) of a marketing campaign is a commercial rather than a technical purpose. Therefore, the iterative mathematical algorithm of claim 1 remains a mere mathematical contribution which does not enter into the examination for an inventive step.
6.1 The appellant argues that the choice of algorithm is based on technical considerations as it takes account of technical (e.g. memory) limitations of computers and diverges from a human approach.
6.2 In decision T 1227/05, point 3.2.5 (supra) it was held that (the sole) processing speed was not a suitable criterion for distinguishing between technical and non-technical method steps since it was always possible to conceive of a slower algorithm than the one claimed. Similarly, the sole amount of memory a computer-implemented algorithm requires is equally unsuitable for determining whether or not a method step contributes to the solution of a technical problem since it is always possible to imagine an algorithm demanding more memory. Furthermore, whether or not an algorithm is similar to what a human being would do may play a role for the examination for inventive step, but this examination presupposes that the technicality of the feature has been established.
6.3 The appellant further argues that the choice of algorithm is not part of the requirements supplied to the skilled person by the business manager, and concludes that the choice of algorithm is tied to a particular manner of (computer-)implementation.
The Board agrees that the mathematical algorithm is not provided by the business manager who is only interested in an economic forecast on which he can base his decision for a marketing campaign.
However, the Board does not agree with the appellant's conclusion that the algorithm is provided by the implementing programmer. In the absence of a technical overall effect and purpose, the algorithm is provided by a mathematician for mathematical and ultimately commercial purposes. Mathematical definitions do not become technical by defining commercial relationships. For example, response probabilities and response values of customers are based on past customer behaviour (A2, page 26, last paragraph; page 27).
6.4 As to the implementation of the algorithm, no internal function of the computer requires a non-obvious consideration to track and reverse incremental changes in the form of reassignments.
6.4.1 The random number mentioned in claim 1 solves no problem other than splitting a large list of customers into two partial lists, without achieving any external technical effect or implying any technical consideration of the internal functioning of a computer.
On the implementation level, random number generators are well-known. The application implicitly confirms that finding as it is silent on any technical detail of generating random numbers.
6.4.2 A binary map of flags settable for each possible customer-activity assignment (A2, page 8, lines 20/21) does not diverge fundamentally from a human's approach when testing a multiplicity of assignments: a human would obviously mark (i.e. flag) tested assignments so as to test other assignments next.
Nor does the mathematical or commercial meaning of the flagged information imply any non-obvious technical modification of general computer functions.
7. The step that "the optimising algorithm is terminated after a user-defined number of customer reassignments does not improve the most recent best goal value" is considered next.
As mentioned above, the innovative potential of the algorithmic scheme can be left aside since it does not serve any technical purpose and, thus, does not contribute to the technical character of the claimed method and cannot enter into the examination for an inventive step.
Said lack of a technical purpose is not altered by defining a mathematical criterion for terminating the algorithm.
8. The Board concludes that the method according to claim 1 of the third auxiliary request does not involve an inventive step over a general computerised method for processing data according to any existing mathematical algorithm and, thus, does not meet the requirements of Article 56 EPC 1973.
9. A fortiori, the broader versions of claim 1 (main request, first and second auxiliary requests) also lack an inventive step.
ORDER
For these reasons, it is decided that:
The appeal is dismissed.