The first phase covered the main actions such as validating inputs, checking tracking information, and preparing replies. Once the system was used at scale, a few weaknesses became clear.
Failure in USPS web scraping
Occasional failure in USPS web scraping caused by slow loading pages or shifting page structures
Limited visibility into workflow failures
Very little visibility into workflow breaks because there was no visual context to understand what went wrong
No retry logic for transient UI issues or missing elements
No built-in retry approach when the interface lagged or when an element failed to appear
Lack of screenshots/logs
No screenshots or detailed logs to help teams understand why certain emails could not be processed
Risk of resource leakage
Risk of leftover browser sessions that stayed open when an error stopped the flow midway and created unnecessary resource use
These gaps made it clear that the system needed to grow into a more resilient and easier-to-diagnose setup, which led to the start of a second phase.
We focused on strengthening the existing workflow so it could handle real world variations with more clarity, stability, and the ability to recover on its own. The goal was to make the automation dependable even when no one was watching.
Intelligent Retry Mechanism
The system now makes repeated attempts to interact with USPS pages, including locating fields and buttons when they do not appear on the first try.
It waits for elements to settle using timed checks so the workflow is not disrupted by slow-loading pages.
These steps reduce false failures that happen simply because the page is delayed or still rendering.
Visual Logging for Debugging
Screenshots are captured at important points, especially when something goes wrong.
This includes moments when tracking information cannot be retrieved, when the USPS interface changes, or when expected elements are missing.
These visuals help support and testing teams quickly understand the issue without needing to read through technical logs.
Deep Debug Logs
Every main stage is recorded, from initial validation to USPS interaction to labeling and drafting.
The logs also note the type of errors encountered, the number of retry attempts, and the final outcome of each run.
This level of detail makes it easier to trace the root cause and understand the system’s behavior under different conditions.
Browser and Resource Management
The browser is closed reliably even if an unexpected error interrupts the flow.
This prevents leftover sessions or memory issues during long or busy periods.
Centralized handling ensures the workflow exits cleanly every time.
Fail Safe Labeling and Drafting
Emails with tracking problems are automatically marked for human review so nothing slips through.
A contextual draft is still created even when the tracking lookup fails, ensuring the customer receives a timely response.
This approach guarantees that every incoming email is acknowledged and handled with care.
This round of improvements turned a basic workflow into a steady, dependable system that can support high-volume operations with confidence.
Retry Logic Layer
The workflow now uses conditional retry loops built with Playwright waitFor methods, allowing it to try again when the page or an element is slow to load.
Timeout settings were adjusted to match typical USPS loading patterns so the system waits only as long as needed.
Each retry and fallback is recorded in the logs, giving clear insight into why the workflow had to make another attempt.
Visual Log System
Screenshots are taken at important points, including successful and unsuccessful USPS status checks, moments when elements are missing or delayed, and the draft view right before the Gmail API is called.
These images are saved in an automatically created screenshots folder, complete with timestamps, so teams can easily review what happened during each run.
Debug Logging
Logs are stored in a clear JSON or text structure with markers for each stage of the workflow.
They include timestamps, the number of elements found at key points, how many retries were made, and any errors or labels applied.
This structure helps teams follow the full path of the automation and understand every action it took.
Browser Lifecycle Handling
Dedicated cleanup steps ensure the browser closes properly even when an unexpected error interrupts the process.
Memory usage is observed throughout the run, so the system stays stable during long periods of activity.
Each workflow cycle includes a confirmation in the logs that the browser closed as expected.
Reduced Undigonised Failures
60 % reduction in undiagnosed failures with clear visual proof of what went wrong, support teams could resolve issues quickly without waiting for developer help.
Faster debug time
The mix of detailed logs and screenshots made it easy for non-technical staff to trace a problem and fix it in 3x less time.
Zero resource leaks
Strong browser lifecycle controls kept sessions clean and prevented the buildup of unused processes during long periods of activity.
Near complete labeling coverage
Even when USPS tracking did not return a result, each email was still marked correctly and given a 100% relevant draft, ensuring no conversation was left unattended.
Conclusion
With this second phase, DataToBiz turned a basic workflow into a fully reliable USPS automation system that can handle real-world variations and give full visibility into its operations.
The addition of retry logic, visual debugging, detailed logs, and careful resource management allows the system to run independently and consistently at scale.
This improvement reduced the need for manual work while giving the client practical tools to oversee, manage, and enhance their support workflow with confidence.
Briefly describe the challenges you’re facing, and we’ll offer relevant insights, resources, or a quote.
Business Development Head
Discussing Tailored Business Solutions
