Article Preview
TopIntroduction
In current B2B integration scenarios, companies have to face the great overhead in B2B message templates. In order to exchange messages, they need to customize these standard message templates published by standardization organizations like the UN/CEFACT (UN/CEFACT, 2003). These templates contain thousands of fields in order to cover all the business needs of one industry domain or even across industries. For the creation of a new message (also called message guide) only a small amount of fields from the standard template is used. Creating such a message guide means customizing the standard template: redundant fields have to be removed and missing fields have to be added. This manual process is time consuming and error prone. Another big issue is the different semantic understanding or the miss-use of a field: business partners might use the same syntactical field for different purposes. Additionally, market leaders might ignore given template structures and force partners to adapt to their interpretation. Over the last decades, a great number of various EDI standards emerged. Standardization organizations tried to cover industry domain-specific requirements on the one hand, as well as industry-independent demands on the other hand. Furthermore, if companies need to support various standards, integration cost increase even more. For each standard, software has to be adapted or additional modules need to be bought. A step towards the reduction of the standard heterogeneity is the creation of subsets for different industries within one standard. However, in this case the same compatibility problems mentioned above occur as well: business partners might not use the identical subset or misuse fields.
In case two business partners want to exchange messages using different standards or subsets of standards, they require a mapping between these messages. The mapping assigns each field of the source message to a corresponding field of the target message guide. The creation of a message mapping between two standards is also a time consuming process. Usually, consultants are hired by the companies in order to create these mappings as expert knowledge of the involved standards is required.
In total, integration cost still make up around 40% of companies’ IT budgets (Kastner and Saia, 2006). In terms of guide creation and mapping creation, there is a lot of potential for effort reduction.
As a pre-study, we examined concrete standard business document templates of 7 different message types from 15 different e-business standards of different industry domains, and their interpretations from 50 different companies. Our analysis revealed that on average more than 60% of the structure and the elements of each schema are semantically comparable. However, only 5% are syntactically similar. Consequently, a precise and commonly understandable lingua franca with consistent and semantically unambiguous meaning of structure and elements is feasible and the key solution. Our approach incorporates a canonical data model (CDM) as the single view of data for multi-enterprises, enterprises, divisions, or processes and can be independently used by any system or business partner. In this article, we enrich the approach presented in (Dietrich et al., 2013) by refining the notion of the knowledge base to the concept of the Unified Data Model (UDM).
There are several related research approaches, like the MOMIS tool (Beneventano et al., 2001), the Porsche approach (Saleem et al., 2008), the Xyleme project (Delobel et al., 2003), and the BInXS approach (Santos Mello & Heuser, 2005). Besides the existing research, several solutions have found their way into the industry, for example the Contivo solutio from Liasion (2012) or Boomi AtomSphere (Boomi, 2012). However, all related works only partly tackle the problems presented in this article. The central data model is mostly manually modeled and a common structure is never computed. This means that especially cross-standard communication remains a big issue for companies. In addition to our previous research, a comprehensive evaluation of the existing work regarding the individual features of our approach is provided.