Building a Product Advisor with Pimcore and Perspectives

The Challenges of Building an (Intelligent) Product Advisor

I would like to demonstrate the efficiency and the intuitiveness of Pimcore based on the use case of a product advisor, which we recently had to build for a big market leader within the sports industry.  We utilized and combined several Pimcore features for the integration of complex product data. Learnability, i.e., discoverability of features and intuitiveness, become vivid with the application of a quite new Pimcore feature, viz. perspectives. 

The requirements for building the product advisor were as follows:

Act like a real-world shop assistant: The product advisor has to operate like a real-world shop assistant: it would ask the customer a number of questions, and provide a predefined
set of answers and recommendations for each choice. Like a real-world assistant, the advisor should consider the user needs and avoid irrelevant or redundant questions.  For instance, if a potential customer wants to order a tent for her family, the product advisor might automatically exclude tents for only one or two people, and trying to ask more specific questions about the environment in which the family wants to camp, or about the size of the family.

Integration with eCommerce and cross-selling: At the end of the advisor, the best-suited products have to be displayed, and a set of cross-selling products will be recommended. For instance, for a family guy, a fire cooker and a head torch may be recommended, while for an alpinist lightweight sleeping mat and a low-temperature resistant sleeping bag may be offered. At this point, a high conversion rate (purchasing of products) is expected. 

Low amount of manual integration work: The manual effort for the configuration of the product advisor should be at a minimum. Already existing product data and product attributes should be reasonably reused.  Other requirements were intuitive use, support of multi-media assets (images, videos) for visualization, responsive design and SEO.

The Solution Approach

We use Pimcore for the design of the data model, the maintenance of the data structure (e.g., creation and linking of product advisor steps and options), and the administration of content (creation of text, assets, object relations). As the maintenance of the data structure includes various concepts and data types (product advisor steps, products, product attributes, images) that are partly imported from the customer’s product information system (PIM) daily, Pimcore perspectives are leveraged to make the maintenance of the data structure manageable. 

The solution approach is summarized in figure 2, the product advisor perspective.

1) A product advisor consists of steps, whereby several sequential steps lead to a result page, in which the recommended products and cross-selling products are visualized. 

For each step options, which will be either product categories (e.g., tents, backpacks, jackets), or product attributes (e.g. number of doors) can be assigned. 

To avoid that the customer has to maintain redundant branches (leading to an exponentially growing tree), only the minimum set of branching possibilities need to be defined. The selection of the "optimal next step" is left to the Pimcore eCommerce framework, which filters the available options (cf. figure 1), and the "available next steps" based on the choice of the user, and whether there are recommendable products available left for the customer's combination. 

2) Because the client already provides various product attributes (e.g., size of tent, number of entries, number of people suitable) via an external interface, these attributes, that are already assigned to specific products, are extendable and reusable as options within the product advisor steps. Typically, the client adds a specific attribute image, a short description, and expert knowledge about a specific choice.

3) Because some required product advisor options (e.g., camping type: family, alpinist, beach type, traveller) are not available in the client's interface, the client can define additional options directly in Pimcore. These are stored as product attribute objects within the Pimcore structure and must be assigned to the relevant products (4). 

4) Assets are imported via the client's interface, but to show optimized products the client can add product advisor images for options in a single place (5).

Result

As one can see, by leveraging an integrated perspective, the configuration of the product advisor has become much more manageable for the client.   For instance, manually created attributes can be directly assigned to products (4) and product advisor steps (1), via drag-and-drop. The entire configuration process becomes much more transparent and the client saves essential time.

From the developer’s view of point, most important thing is that custom views are configurable within minutes. A perspective can be activated by only adding two configuration files in the Pimcore configuration directory (customviews.php and perspectives.php). Each subtree is represented by a single configuration block in the configuration file. Titles and icons are customizable and perspectives are applicable Pimcore user-specific.

Conclusion

With its strongly object-oriented and scalable approach, Pimcore easily tackles complex problems and never fails to produce practical solutions.  The process of data structure definition, integration, enrichment and visualization is very efficient, intuitive and thus pleasurable. 
With Pimcore, developers have the right tools to build a solid data (access) layer and backend with minimal effort, but also provide clients with advanced logical components and vigor frontend services. I could not imagine using any other solution than Pimcore, and I would highly recommend this product to all developers around the world.

Sources

• Pimcore Perspectives: https://pimcore.com/docs/5.x/Development_Documentation/Tools_and_Features/Perspectives.html
• Learnability: 5 Principles of Interaction Design To Supercharge Your UI: http://www.givegoodux.com/learnability-5-principles-interaction-design-supercharge-ui-3-5/
• How to Measure Learnability: http://www.measuringu.com/blog/measure-learnability.php 
• Easy to Learn: https://www.over-yonder.net/~fullermd/rants/userfriendly/3
• Pimcore Perspectives: https://pimcore.com/docs/5.x/Development_Documentation/Tools_and_Features/Perspectives.html