AI-Based Image Anomaly Detection in Foam Manufacturing
Traditionally, manufacturing companies have relied heavily on manual supervision and monitoring of their production lines to ensure quality and detect anomalies.
Our client is large global automotive supplier with hundreds of manufacturing facilities and sales centres spread across 25+ countries. Their global sales team has in place a global Sales Database for tracking sales and generating revenue forecasts across all primary operating groups of the business. The executive leadership team relies on the sales volumes and forecasts for future planning and management of the business.
The Sales Database solution is hosted on on-premise servers and, initially, both the infrastructure and application were managed by their global IT team. At the front-end of the solution, business units input their sales data, which is saved into the Sales database. This database feeds a sophisticated analytics cube, which conducts complex calculations (such as currency, US GAAP, CAGR, etc.), and feeds a series of Power BI dashboards and Excel reports. This cuts down on the report generation time as the calculations are not dynamically done when the report is refreshed but are instead pre-populated in the cube.
However, the cubes within their architecture were built in a manner that did not allow their Global IT team to make changes and updates in accordance with the business model. Occasionally, when the team was working on their month-end or quarterly forecasting activities and there were too many users at once, the reports would not complete. This lowered their performance and adversely impacted the stability and reliability of the Sales Database forecasts.
Moreover, internal resolution of issues that cropped up was becoming difficult. Their client’s global IT team is based in Central Europe and coordination was a challenge due to the time zone difference. The IT team also lacked sufficient resources to support the cubes and Power BI reports internally.
Adastra had initially been engaged by the client to build several business-centric Power BI reports on the front-end of the Sales Database solution.
As we started working closely with their team, they sought our support to redesign and optimize their failing cubes in order to improve their performance and speed up the reporting process. We also improved the availability and flexibility of the cubes and the team was now able to make changes to them as the business calculations and requirements evolved.
In the process of upgrading their cubes, we realized that the solution was sitting on undersized infrastructure which was being shared by Sales Database and four other business-critical applications. Adastra’s experts recommended migrating the solution to dedicated servers, and subsequently worked with their business infrastructure team to migrate the solution to improve the platform stability, scalability, availability, and responsiveness by distributing the workload.
The successful completion of the development work convinced both the business and IT groups that Adastra would be the right partner to provide them with ongoing application and end-user support for their Sales Database platform, and in early 2019, Adastra was brought on-board by the client as a managed service provider for the solution.
The Managed Services we offered our client initially started as a staff augmentation and in June 2020, it graduated into a performance-based service model. Our team is responsible for providing ongoing support for the cubes and reports and the scope of our work is divided into three distinct categories:
Monitoring and alerting: Adastra’s team is responsible for proactively monitoring the Sales Database platform to ensure adherence to the availability, performance and capacity metrics for the database, cube, cube refresh jobs and reports (Power BI, SSRS and Excel). Monitoring dashboards have been set up to view the system health indicators, such as cube performance and report response time.
Our service agreement sets out strict availability and performance targets and Adastra also commits to a pre-defined response time based on the task criticality. In the event of performance degradation or availability issues, we have a process in place to send alerts out to the respective parties (to Adastra for application management or the client’s global IT for infrastructure issues).
Issue resolution and support: We provide Level 3 Application support related to the Sales Database cube and reports via an automated alerting system and tickets submitted by the functional team that are routed through their Global IT help desk. The client maintains level 1 and 2 support internally, so the Sales Database functional team (Level 1) acts as the first point of contact for business users. They raise a ticket for issues they are unable to solve that their end and sends it to their Global IT help desk (Level 2). From there, all application related issues are routed to Adastra for resolution.
Once the severity level is assessed, Adastra triggers the Incident Management process and applies a fix. Each step of the process is documented, and a constant channel of communication is maintained with the client’s functional team.
Solution improvement: Adastra supports the client in planning, designing, and developing enhancements for the Sales Database cube and reports. On an ongoing basis, our team provides architectural and design recommendations in alignment with industry best practices, and upon approval, we implement those structural changes. We also support the functional team in managing the intake of new tasks and projects.
Our client now has a high-performing, reliable Sales Database solution that they can count on to deliver timely and accurate forecasts to drive their business planning. A stable platform that meets or exceeds expectations is mission-critical and essential for their reporting and decision-making processes globally.
Adastra’s team ensures that the cube and reports meet or exceed a target of 99.9% availability, and that the data is refreshed successfully every ten minutes The team has also greatly enhanced the cube by implementing dynamic measures and report performance by simplifying the rendering process and it now takes 10 seconds or less for the cube to respond and for reports to be rendered.
The platform is continuously monitored to ensure all issues are immediately identified and addressed to avoid any business user impact, which is especially critical during the Business Planning period. For critical issue resolution, the Adastra team provides a response time of under 31 minutes while low priority issues are resolved in under 1 day. Keeping with the client’s service and support needs and to provide a cost-effective managed service solution, we offer best-shoring, with an architecture and development team based in Toronto, a client service manager in Detroit and a scalable managed services team in Bulgaria. The service manager ensures that all system and support performance targets are met each month and provides a monthly performance report to functional team. The client’s functional team now works with the confidence that in the scenario of something going wrong, they are guaranteed quick response times and have a 24x7 point of contact for escalation.
Moreover, with Adastra’s managed services team taking care of all application related service and support, the client does not need to scale their internal IT team. The enhanced speed and performance of the cubes also resulted saved time and effort, especially during monthly and quarterly forecasting activities.
As the organization’s needs have evolved, Adastra’s team has been involved in ongoing development and enhancements to add new features to make the solution more robust.
Over the course of our engagement, Adastra has established itself as an extension of the client’s in-house functional team and they look to us for recommendations on not only present improvement, but also on future-proofing their solution to support the organization’s ongoing growth and management.
We will contact you as soon as possible.