product management

Reusable Search Infrastructure

 

At U.S. Bank, I led a team responsible for developing a suite of reusable Adobe Experience Manager (AEM) components for the enterprise. This case study focuses on the development of a reusable Search component and the accompanying infrastructure required to support it. Key elements of the project include:

    • Creating a scalable, reusable platform that enables development teams to efficiently build and deploy digital experiences by leveraging pre-built, standardized AEM components, reducing the need for redundant development efforts.
    • Developer-focused product design, ensuring ease of use, flexibility, and adoption across teams.
    • Seamless integration with multiple technology stacks, ensuring compatibility and scalability within the organization’s diverse ecosystem.
    • Headless CMS architecture and a cohesive data strategy to decouple content management from frontend delivery, enabling more dynamic and flexible digital solutions.

      The problem

      Our audit of site analytics, search logs, and call center data revealed several key issues:

      • High search exit rates and frequent “no results” pages across the platform.
      • An analysis of search terms related to “no results” occurrences and top call drivers indicated a need for more robust help content.
      • The search experience was hard-coded across different channels and indexed various content sources, leading to inconsistent search results for the same queries on different platforms.
      • A separate content database, used by internal bankers and housed in ServiceNow, contained the necessary information but was managed by a different team called Knowledge Center).

      Key challenges:

      • Fragmented content and an inadequate search algorithm resulted in unproductive searches, high bounce rates, and increased call volume.
      • Siloed development led to a fragmented user experience and high maintenance costs, with content spread across three different databases, each managed by separate teams.
      • The lack of a centralized budget or product team meant there was no alignment of content strategy or search functionality across channels. Different teams operated under separate CAPEX budgets, each with distinct objectives.

       

      Call drivers from call center

      A snap shot of the unproductive search data

      Below is a visual map of the fragmented data landscape shows different channels using various data sources. The content migration from Salesforce to ServiceNow required the Knowledge Center team to manage two sources of duplicate data, as Online Banking continued using content from Salesforce.

      Business goals

      • Build a new search component that can be reused across different platforms and optimize workflows for content strategists, authors and developers.
      • Understand user search intent and identify the patterns to fine tune the search algorithm
      • Secure buy-in from leadership and multiple teams to consolidate fragmented data into a single source of truth, supporting the new search infrastructure and ensuring scalability for future integrations with other databases and platforms.

      Metrics

        • Reduce drop-off rate on search results
        • Increase click through rate among top results (relevancy) 
        • Reduce no result instances
        • Reduce support requests and call volume
        • Increase conversion rate for various product funnels
        • Acceptable performance of API load time
        • Increase speed in configuring the component (devs) and managing content (authors/content strategists)

        Getting buy-ins

        U.S. Bank followed a CAPEX model, meaning newly-funded projects would not begin until the next calendar year at the earliest. Although my team was funded to build the component suite, the integration of different data sources was considered out of scope.

        To address this, I was determined to persuade various product teams to collaborate and create a unified experience that could benefit everyone. By understanding each team’s motivations and pain points, I developed a plan to secure their support. I presented a proposal that outlined the redesign of the search experience, re-engineering data integration, and enabling a headless CMS.

         

        In addition to outlining the specific benefits for each team, the presentation emphasized an estimated $5M in annual losses attributed to lost revenues and calls driven by unproductive search terms. The financial forecast projected increased conversion rates across product verticals and significant cost savings from consolidating siloed databases and reducing call center support. This forecast was strategically designed to resonate with leadership.

        The various teams involved and their motivations

        Research

        In this phase, I led the collaboration among the different teams, including my Platform team, Enterprise Search team, the marketing content strategists, and product managers/business analysts from three different product teams. Additionally, we engaged with various business groups across the bank throughout the process.

        • Conducted audits of search logs, analytics, and call center data.
        • Reviewed the search experience across channels and conducted usability testing with customers.
        • Identified high-value queries that resulted in unproductive searches, leading to targeted content mapping and creation efforts.
        • Audited the Knowledge Center’s content repository.
        • Interviewed development teams to gather requirements for database connectivity, indexing, access control, log monitoring, and deployment workflows.
        • Engaged with authors, content strategists, and SEO specialists to understand their needs for custom tagging, metadata control, and other content management strategies.

        Snapshots of some of the audits and research we did

        Defining the experience and organizational alignment

        In this phase, I led multiple teams and work streams to ensure successful delivery of various aspects of the initiative.

        • From a search experience perspective, I guided the UX and content teams in creating various search experience concepts for customer testing, including a robust “no results” concept. I facilitated discussions to align all channels and platforms toward a unified search experience.
        • From a component development perspective, I conducted cross-functional workshops involving UX designers, developers, and authors to achieve an intuitive design for the component’s authoring experience. This design and the component itself were socialized and tested with various internal teams to ensure usability.
        • From an enterprise reusability and alignment perspective, I coordinated efforts between the product designer and the Design System team to align design patterns and interactions effectively. I also facilitated collaboration between the platform development team and the Design System development team to integrate React front-end components with AEM components and ensure alignment with the style-agnostic design token architecture established by the Design System team.
        • From a data quality perspective, I led collaboration between the Enterprise Search and SEO teams to ensure successful content mapping, including the identification of acronyms and synonyms, as well as the integration of content from various subdomains and microsites into search results. Additionally, I worked with the Knowledge Center to fine-tune the content in ServiceNow, ensuring the integrity of the single source of truth.

          ServiceNow-AEM integration & Headless CMS

          This phase was crucial for building the data infrastructure necessary for the search experience and component to function effectively, during which I led multiple engineering teams to consolidate content into a single source of truth.

          We faced two conflicting choices:

          • On one hand, as AEM is the enterprise’s preferred content management system, Platform leadership wanted to migrate content from ServiceNow to AEM.
          • On the other hand, the Knowledge Center team strongly preferred to retain ServiceNow due to its various workflows, which are essential for managing the large volume and ever-changing nature of the content.

          The decision hinged on a technical and cost analysis comparing:

          1. Migrating Knowledge Center content into AEM and building out all necessary workflows,
          2. Integrating ServiceNow content with AEM through APIs.

          To facilitate this decision, I coordinated between my development team and the ServiceNow team to create a feasibility prototype for integrating ServiceNow data into AEM’s content fragments and designing a headless CMS architecture. This prototype aimed to assess the effort required for integration. After the cost analysis, we opted for the second option.

          Once the performance and feasibility of the integration were confirmed through the prototype, I coordinated efforts between the ServiceNow and Platform teams to develop various APIs. I worked with the development teams to standardize the data structure, ensuring that the merged data could scale effectively. Additionally, I established new workflows among different teams to maintain data integrity and consistency following the integration.

           

          A snapshot of the delta API developed to detect changes to the data in ServiceNow

          A snapshot of the requirements

          Fine-tuning the search engine

          This phase is critical for enhancing the search functionality and overall user experience by optimizing content indexing and search capabilities across various platforms.

          I partnered with the Enterprise Search admin team to identify necessary improvements and worked closely to enhance indexing capabilities for various content sources, including redirects and content mapping within the component. 

          As part of this initiative, we implemented several enhancements to the search engine, such as adding features like autocomplete and relevant suggestions to improve user experience, and new search analytics capabilities. Additionally, we optimized the system for future integration for different teams, from enabling voice-command search to indexing documents on Sitepoint.

            Testing and deployment

            Finally, the testing phase! The search component was first deployed to select internal teams to gather early feedback from both devs and super-authors. These teams helped identify usability issues and provided valuable input on performance and feature improvements. The components then were made available to a wider group of authors and content strategists.

            Features like ServiceNow integration, and relevance adjustments were deployed first to only a subset of users to gather insights and ensure performance. Subsequently, unit, end-to-end, cross-platform and performance testing were then conducted for production rollout.

            To facilitate the launch, I also coordinated training sessions and the creation of documentations for authors and developers to help them effectively leverage the new search component.

              The outcome

              • A single source of truth, enabling both users and contact center to search and get the information without needing to manage multiple systems independently.
              • Headless CMS enabled a path for future content reusability across the enterprise, promoting reusability decreasing the efforts for maintenance and inconsistent content
              • Improved findability by 155%, decreased unproductive search by 65%, reduced call volume by 30% and save $5M of the annual cost 
              • Authors and content strategists can easily manage settings for different websites, such as what section/sites to include in the search query, what no result page looks like, tagging relevant content
              • Developers can easily integrate multiple data sources (e.g., intranet, articles, SharePoint) with ease while allowing control over relevancy settings, allowing reusability across various dev teams including voice banking, mobile apps, intranet, and subsidiaries—each with unique content sources and user bases.Developers across teams saved significant time, as they no longer needed to build individual search solutions for each platform. The component could be reused with minimal customization.
              • Regular work flows among the teams to establish recurring communication, ensuring we optimize reusability and minimize rework in the future (eg. a cadence for Enterprise Search and SEO to audit search log, Enterprise Search, Platform and ServiceNow to evaluate performance)

              You can see the live experience built on this reusable search infrastructure at the following public websites:

              https://www.usbank.com/search/search-results.html

              https://www.usbank.com/customer-service.html

              https://www.elanfinancialservices.com/credit-card/search-results.html

              https://www.elavon.com/search-results.html

              Example of the same content fragment being integrated and reused for both the public website and online banking

              Before & After of the search experience

              other Projects

              hapybara

              Identity authentication and social platform

              safe rate

              Seed-staged mortgage startup go-to-market