IBM App Connect includes AI-powered Mapping Assist to dramatically accelerate the speed of development, shorten time to value, and improve your overall user experience. Mapping Assist uses a pre-trained AI algorithm to securely provide users with intelligent, customized data mapping suggestions.
The need for digital transformation is more important than ever before, with 56% of CEOs reporting that their digital improvements have led to revenue growth. Businesses are recognizing the need to focus on creating convenient, digital solutions both internally for employees and externally for their customers to stay ahead of competition. Integration is key to digital transformation efforts – by harnessing the power of your existing data you can extend the reach of your digital IT assets, whether you’re providing an existing service through a new digital channel or tapping into an entirely new market via APIs surfaced through partner applications.
When building integrations, data mapping is the process by which a user specifies how data elements from a source application correspond to elements in a target application. For example, mapping the “Post Code” field from Application A to the “Zip Code” field in Application B. Data mapping is the first step of a wide variety of integration tasks, including data transformation or data mediation between data sources and a destination. But data mapping has long been a point of frustration for developers since mapping fields between different endpoints is complex and time-consuming, slowing the speed at which businesses can respond to changes and remain competitive.
What makes mapping complex?
- An increasing number and diversity of endpoints
There is a steep learning curve when integrating a new application, particularly in understanding the application’s metadata. This is especially challenging when integration development is not a primary role, for example a marketing professional wanting to use a new SaaS application to simplify event management will need to map existing customer data from a CRM application. And the need for integration continues to grow as businesses are using an ever-increasing range of endpoints, from home grown apps to off-the-shelf applications, databases, files, SaaS applications, Webservices, and more. In this environment, understanding how to map elements between many different applications and delivering at speed becomes a challenge for even the most experienced integration developers.
- No field naming standards
Applications are built by different vendors with no universally agreed upon conventions for field naming. For instance, a PIN may refer to a pin-code (personal identification number) in one application while it may refer to a product identification number in another. This lack of standardization makes it very difficult for users to reuse knowledge when working on different applications. Users are thereby forced to spend valuable time learning every new application or endpoint before they can start to deliver business value through integration.
Many application vendors also allow administrators to develop custom objects or to make schema changes in existing business objects. This customization generates additional overheads on the user to learn and understand field naming outside of the publicized documentation.
Additionally, making changes to integrations created by someone else poses a challenge, especially if design related mapping decisions are not documented properly. Reviewing and understanding this additional documentation, where available, adds time and complexity.
- Field structures further compounding complexity
There are multiple examples of applications, for instance MS Dynamics 365 and IBM Maximo, which have huge numbers of fields to map, in some cases more than one thousand fields. With so many fields to map, building integration flows that contain these applications can become especially time intensive.
Furthermore, applications can also provide data via nested fields. For example, the address field may be made up of multiple fields such as ‘house number’ and ‘street’. As a result, field hierarchy must be a consideration while mapping objects. The user may need to repeatedly map the source flat structure to the target nested structure and vice versa – a time-consuming activity with greater risk for errors.
Simplify integration with IBM App Connect’s Mapping Assist
Mapping Assist provides intelligent, customized data map suggestions when building a flow to simplify integration and alleviate the complexities mentioned above. Mapping Assist dramatically accelerates the speed of development, shortens time to value and improves your overall user experience.
Figure 1: By Selecting ‘Insert suggestions’, all suggestions with the highest match rate for a given node are automatically populated in the target application’s fields.
Mapping Assist utilizes a pre-trained Artificial Intelligence (AI) algorithm which uses Natural Language Processing schema matching to find the closest match between source and target mappings. No pre-training is required to get started – Mapping Assist suggestions are provided by the algorithm as soon as you start building a flow. The algorithm assigns a percentage of confidence to each mapping suggestion based on semantic proximity to a target field, using business object metadata such as the field name, description, and display name to find the correct match. This means that Mapping Assist will also facilitate mapping fields from custom objects and schemas where there are no field naming standards.
Users can choose to auto-populate all fields with the suggested highest confidence matches or manually inspect all suggestions to insert specific mappings (see figure 1 and 2). Mapping Assist will not override any user defined mappings and it will also remember any of your previously accepted mappings. Additionally, Mapping Assist factors in all applications within a flow when providing suggestions for a target application to offer especially relevant, customized suggestions.
Figure 2: Alternatively, you can view individual field mapping suggestions with their match percentage before selecting all mappings or only specific mappings.
Mapping Assist is trained to provide map suggestions for complex nested mapping fields, not just for flat field structures. For data sources with a large number of fields or cases where there are a significant number of nested fields, Mapping Assist will run in the background enabling users to continue their work on other tasks.
Figure 3: Suggestions are generated considering all previous source nodes in the flow. In this image you can see that there are suggested mappings from multiple applications to Salesforce fields
IBM Mapping Assist is secure
Mapping Assist runs where your integrations run so there are no connections to a cloud service, further securing your privacy and increasing suggestion generation speed. And at IBM we recognize the importance of protecting your data. Your Mapping Assist AI model and runtime data is your intellectual property – IBM will not share it with others.
IBM App Connect Mapping Assist is available in IBM Cloud Pak for Integration 2020.2.1 and IBM App Connect Enterprise for Certified Containers.
Co-Author Amy McCormick