A large DIY retail chain needed to instantly disable specific pickup points across its delivery aggregator without manual coordination across teams, and to get clear reasons for delivery delays and cancellations from third-party carriers. Azati's QA engineer tested both features end to end, including integration testing with external delivery partners.
backend services tested for delivery status reason tracking
features QA-validated end to end, including third-party integration testing
tools used across test management, monitoring, and message brokering
The client operates a large retail & logistics network spanning more than 100 stores, 11 dark stores, and 6 distribution centers, with tens of thousands of employees and a logistics aggregator coordinating delivery across multiple third-party partners. As the pickup point network grew, two operational gaps started costing real time and trust: there was no fast way to pull a problematic pickup point out of circulation, and nobody could get a clear answer, quickly, about why a specific order had been delayed or cancelled by a carrier.
Both problems had the same shape. A simple-sounding business need, turn off this point, tell me why this order is late, was buried under slow internal coordination and undocumented behavior from third-party systems. Azati's QA engineer joined the team responsible for the pickup point and delivery aggregator to validate the fixes the engineering team built for both issues, with a particular focus on how those fixes behaved once real delivery partners were involved.
Removing a point from the list customers see used to require manual coordination across multiple internal teams. That was workable at a small scale, but it broke down as the pickup point network grew and the reasons for needing to act quickly multiplied:
When a customer asked why their order hadn't arrived, the honest answer often required a slow back-and-forth between internal teams and the delivery partner. There was no structured way to capture and surface the carrier's own stated reason:
Both fixes touched backend services that communicate asynchronously and depend on data from external delivery partners. Testing had to cover internal logic and the behavior of the full integration path, not just one service in isolation:
Validating an exclusion rule or a reason code that travels across Kafka topics and RabbitMQ queues before reaching a customer-facing filter is not a simple click-through test. Azati's QA engineer worked directly with the platform's message-driven architecture, using Conduktor and the RabbitMQ management plugin to verify that data moved correctly between services, not just that the end result looked right.
Both features depended on behavior from third-party delivery and pickup point partners outside the client's own systems. Azati's QA engineer planned and executed integration tests that crossed those boundaries, coordinating around partner-specific quirks rather than assuming every external system would behave the same way.
Test cases and results were maintained in AllureTestOps, with task tracking in Yandex Tracker and supporting documentation in Confluence. That discipline matters in a large retail organization where multiple teams touch the same delivery aggregator over time and need a reliable record of what was tested and why.
The engagement ran inside the client's existing sprint-based Agile process. Azati's QA engineer integrated into that cadence directly, picking up backend changes already in flight rather than requiring a separate ramp-up phase before becoming productive.
Whether it's third-party integration testing, message-driven architecture, or fast-moving backend changes, Azati's QA engineers can validate it without slowing your team down.
Talk to a QA engineerAzati tested the new exclusion functionality that lets the operations team remove individual pickup points, or entire groups of points, from what customers see during checkout. This included validating the new database collection storing exclusion attributes, the CRUD methods for managing entries, and the logic for excluding points by shared attributes such as delivery partner or geographic region.
Azati tested changes across three backend services responsible for receiving cancellation and delay reason codes from delivery partners and propagating them to the team that owns the logistics operator dashboard. This included verifying the fixed reference list of reason codes and the history of delivery date changes shown per order.
Both features depended on real behavior from external delivery and pickup point partners, not just internal logic. Azati ran integration tests against connected partner systems to confirm that exclusion rules and reason codes behaved correctly across the full data path, using Postman for API-level verification and Swagger for contract reference.
| Area | Azati contribution |
|---|---|
| Pickup point exclusion | Tested CRUD logic, group-based exclusion rules, and the enriched pickup point filter |
| Delivery reason tracking | Tested reason code propagation across three backend services |
| Third-party integration | Validated behavior across connected delivery and pickup point partners |
| Message-driven architecture | Verified data flow through Kafka and RabbitMQ using Conduktor |
| Test documentation | Maintained structured test cases and results in AllureTestOps |
| Delivery cadence | Worked inside the client's existing Agile sprint process |
The platform handles operational logistics data rather than customer payment information, so testing focused on data integrity and correct access boundaries between internal services and external partners. Service-to-service communication runs through Kafka and RabbitMQ with managed access controls, and monitoring across the stack is handled via Kibana, Jaeger, and Grafana Loki, giving the team visibility into how exclusion rules and reason codes move through the system in production.
Azati provided a single QA engineer who joined the client's existing pickup point and delivery aggregator team, working inside their established sprint cadence rather than as a separate, parallel workstream.
The work was delivered as two distinct testing efforts tied to two separate features:
When a point becomes unprofitable, shows signs of fraud, or a partner reports an operational or technical outage, the team can now remove it from circulation quickly instead of routing the decision through a slow, manual process involving multiple internal teams.
A fixed reference list of carrier-reported reasons, combined with a visible history of delivery date changes, gives internal teams a direct answer instead of a multi-step investigation, reducing back-and-forth between teams and with customers.
Faster pickup point corrections and clearer delay explanations both reduce the moments where a customer is left waiting without a real answer, directly improving the experience around order fulfillment.
The engagement deepened the team's hands-on experience with Apache Kafka for reliably collecting, storing, and asynchronously transmitting high-volume data flows between services in real time, expertise that carries directly into future logistics and event-driven platform work.
What stands out about this engagement beyond the two specific fixes:
Building exclusion logic around shared attributes, partner, region, rather than only individual point IDs, means the same mechanism keeps working as the network grows. A single partner outage can be handled with one rule instead of dozens of manual removals.
A fixed, structured list of cancellation and delay reasons does more than fill a database column. It turns a vague "your order is late" into a specific, defensible answer, which matters as much for internal team friction as it does for customer-facing support.
In a system built on internal logic plus several third-party partners, most of the risk lives at the boundary between systems, not inside any single service. Treating partner integration testing as a first-class part of the QA scope, rather than an afterthought, is what actually catches the failures that matter in production.
B2B automotive parts portal engineering: database architecture unblocked via ADR, multi-region order management, and vendor API integrations delivered on a live platform.
Adobe AEM support and full DAM migration for a global beverage enterprise: 10,000+ assets transferred with 100% metadata integrity in under 2 months.
Offline computer vision system for automated pallet counting using edge AI.
Modernized retail corporate system with optimized warehouse processes and enhanced efficiency.
Modernized warehouse management platform rebuilt on modern tech stack for scalability.
Modernized Shopify app with enhanced stability, performance, and engagement.
Last updated
