SAP Commerce search enhancement

SAP Commerce search enhancement

Itransition boosted the search relevance for the online store with 150,000+ items, ensured a 25% conversion growth, and enabled conversion from no-results pages, along with a 12x reduction of admins’ efforts.


Our customer is a leading European supplier of stationery and office goods, including office equipment and furniture, paper, workwear, household goods, and tools. The company owns a chain of 100+ shops and an online SAP Commerce-based store serving 100,000+ corporate and individual clients.

During the online store UX research, the customer discovered flaws in the website search engine, which is a business-critical functionality for a store with a 150,000+ product line. End users reportedly faced irrelevant search results and failed to find necessary items at all in some cases. This led to user dissatisfaction and loss of potential clients (4% of all sales). The basic packaged search for SAP Commerce couldn’t handle the common queries of B2B and B2C users (e.g. entering product SKUs and product characteristics). Additionally, manual administration of the search engine (managing keywords and their variations, uploading item attributes, and user behavior tracking) was time- and effort-consuming.

The customer needed to improve search relevance and administration. They entrusted optimization of the website search algorithm to Itransition, a certified SAP Silver Partner, considering dozens of implemented SAP optimizations and customizations and our proven ability to resolve non-trivial challenges.


Relying on the customer’s search analytics, Itransition implemented custom search scenarios, personalized and optimized search mechanics, and automated time-consuming tasks of search administrators. Itransition’s modifications boosted search relevance and led to a 20% conversion growth compared to the basic search conversions. As a result, the back-office process automation brought a 12x reduction in admin efforts.

Despite the complex business logic and multiple steps of search processing, the system provides search results immediately without any delays.

How the search works now

The marketplace search is easy to use for potential buyers: once they enter a query, the system generates search results immediately. The search is based on a complicated algorithm and includes the following processes:

  • Search type definition
    The system defines whether the search is broad or exact by analyzing how detailed and complete the entered request is. Then it generates results that meet the user’s expectations in terms of volume and exactness.
  • Search scenario definition 
    Following the popular search patterns (e.g. entering product SKUs or entering the desired product characteristics), the search engine generates the most accurate results. To define the pattern, the solution conducts a lexemic analysis of the entered phrase.
  • Handling typos and synonyms
    The solution recognizes and corrects typos and interprets abbreviations and synonyms to provide relevant search results.
  • Ranking the results based on business needs
    Applying the implemented modificators, marketplace admins align search results with the actual business requirements, e.g. bring some items to the top or hide categories or products.
  • Continuous search analytics
    Thanks to search queries and patterns analytics, the customer gains valuable data related to search trends, keyword performance, typo and synonym effectiveness, popular and no-results searches, etc. This information highlights areas for search experience improvement.

Introducing search scenarios

Before Itransition’s engagement, the search engine was based on two standard algorithms to process requests:

  1. Exact – searching for items that match exactly all the entered keywords and characteristics.
  2. Broad – searching for items that partially match the entered keywords and characteristics.

To meet the broader requirements of the customer’s target audience, Itransition extended the algorithms with redirects and search scenarios:

Redirects and search scenarios algorithm

If the search scenario is not applicable, the system delegates the query to the next scenario until the relevant results are returned.

Itransition introduced the following search scenarios:

  • Search by keyword + item SKUs
    Now end users can search items by SKUs, which is a widespread case among B2B users. The system combs not only through the internal database but also through the competitors’ catalogs. The customer buys/parses and uploads to SAP Commerce the lists of SKUs for identical products from competitors. If a user mistakenly searches for a product by its competitor’s SKU, the system matches items and returns the relevant identical product from the customer’s online catalog.
Search by a keyword and item SKU
  • Search by keyword + item attribute
    Since searchers tend to enter product characteristics, we implemented the functionality to process discrete and range item attributes. Discrete attributes are textual characteristics (color, material, purpose, origin), while range ones have numeric values (price, dimensions).
Search by a keyword and attribute

Admins add synonyms and word forms to discrete attributes.





Cover color



2518 - Color cover










Navy, purple

To process range attributes correctly, Itransition developed functionality to configure units of measure, their unit conversions (e.g. 1m=100cm), and a range of possible deviations, as well as to link units of measure with the corresponding range attributes.


Unit of measure 1

Unit of measure 2

Unit of measure 3

















Conversion rate





Meter, metre



No-results search optimization

One of the customer’s challenges was losing potential clients who entered low-frequency queries. Because the system returned the no-results page, in most cases they abandoned the store, so the conversion was close to zero.

Itransition’s team developed components that make the system return some results and range them in line with the user’s intent:

  • Recommended products. The component generates a list of products if there are no results for certain queries. The list is manually configured by the customer’s content managers based on search analytics data.
  • Alternative requests. To avoid zero results, this component provides links to alternative requests which are also configured by the content managers.

These components are combined to achieve more relevance, while the results are ranged based on keywords’ lexemic analysis and items with similar attributes or related to the keywords get to the top.

Now the no-results search makes 4% of all online sales, in contrast to no conversions before the optimization.

Search personalization

Aiming to improve end users’ search experience, we implemented search personalization features. Currently, the system allows for fine-tuning search results to consider end-user personal preferences, search history, and order history, and apply personal recommendations.

These features come up in the form of drop-down search suggestions and include autocompletion, query history, popular queries, and more.

Personal search suggestions

Administration automation

The search engine requires an ongoing actualization of items' attributes and statuses. Before Itransition, admins used to fill up to 1500+ strings at once manually, which was time-consuming for employees and quite expensive for the customer.

Addressing this problem, Itransition automated massive data upload/export in .csv format by means of the hot folder component in SAP Commerce.

With several clicks of the mouse, admins are empowered to:

  • Upload/change/delete product SKUs, discrete and range attributes (both internal and competitors’ ones)
  • Download user search scenarios for further analysis in the external BI system

Thanks to this feature, the search admins avoided human-factor errors, while the customer saved hundreds of person-hours.

Click tracking development

Search analytics is an essential part of the customer’s operation. Leveraging third-party systems, they tracked user clicks but didn’t have a complete picture of user decisions, intentions, and triggers. To provide the customer with the ability to trace and analyze all user actions within a single search request, from entering a request to product selection to cart actions, we implemented the fixation of search query identifier.

Every search session is assigned with a unique query identifier, that is further assigned to every event triggered by a user. Thus, the user session is logged for further analysis and insight into user behavior.

Here are some examples of the tracked actions:

  • Applying/canceling facet filters
  • Resorting search results
  • Changing results listing view
  • Jumping from search results to a product page 
  • Adding products for comparison
  • Adding products to cart, etc.

The customer admitted the improved visibility and transparency of user behavior, which provided them with a ground for experiments in terms of search UX improvement, and helped them take information-based decisions.


The marketplace is deployed on SAP Commerce with Oracle database, while the search engine is built with the open-source Apache Solr. Well-versed in both, Itransition delivered search functionality optimizations using out-of-the-box features and implementing custom ones. The solution is self-hosted and built with out-of-the-box SAP Commerce integrations only, so it doesn’t require any additional resources and infrastructure support.

For custom backend implementations, we used Java, and for the frontend tasks (such as search bar and drop-down redesign) we leveraged VueJs.

Technology architecture


Due to 5000+ distributed employees and a complex organizational structure with multiple divisions, the customer’s company has complicated and strictly regulated business processes. The marketplace development was complicated as well, since there were standalone streams and project teams focused on particular directions (UI team, integration team, user’s personal account team, order checkout team, and many more).

Itransition has become an integral part of the search engine team. The project work was organized by the Scrum methodology in two-week sprints with backlog elaboration, planning, and retrospectives. The project spanned the following development and approval stages:

  • Business requirements gathering, technical and functional analysis 
  • Creation of a specification and discussing it with the customer
  • Task presentation to the architecture committee (the committee unites experts from other development teams)
  • Task implementation
  • Demo and UAT
  • Testing and QA
  • Release and support

Thanks to a multistep code review (internal cross-review, external review by other customer’s teams, demo for stakeholders), Itransition ensured the highest quality and compatibility of the delivered code with the implementations by other development teams.


Itransition delivered multiple original extensions of the marketplace search functionality in line with the client’s business specifics. With every release, the customer gets instruments that simplify their processes and increase search KPIs. Here are some of the improvements we managed to achieve:

  • + 25% overall conversion growth
  • + 20% growth of conversion from search scenarios compared to the conversion from the basic search
  • 12x reduction of admins’ efforts due to data upload/download automation

Since the customer is satisfied with our achievements, we working with them, delivering regular updates that keep the system stable and high-performing.