How can you optimize RPA testing and debugging for different use cases?

I worked on a project involving extracting data from various invoice formats. Initially, the RPA process faced frequent errors due to format inconsistencies. To tackle this, I implemented a rigorous testing phase, using a diverse set of sample invoices to validate the data extraction logic. This step helped identify and rectify format-specific issues. Additionally, I used debug tools like step-by-step execution and watch variables to pinpoint exact failure points. By incorporating assertions and checkpoints, I ensured the accuracy and integrity of extracted data. This structured approach to testing and debugging significantly reduced errors and made the data extraction process more efficient and reliable.

To optimize RPA testing and debugging for diverse use cases, employ modular testing frameworks to ensure scalability and reusability. Implement version control to track changes and facilitate collaboration. Leverage exception-handling mechanisms to address unforeseen scenarios. Utilize logging and monitoring tools for real-time insights. Conduct thorough regression testing after updates. Employ testing environments that mimic production setups. Implement a robust error-handling strategy and continuously refine it based on evolving use cases. Regularly update test cases to accommodate system changes and enhance overall reliability.

Rpa testing has been always been underestimated task. I have seen and experienced during my RPA journey that testing is done very minimal. Many a times during unit testing we just go for happy path scenarios and do not think of negative scenarios which then breaks the bot in UAT and Prod environment. Data extraction is part of every bot. When we are engaged is such a projects where data extraction is required we need to fix the format of the data that we are going to extract.Source data and target data has to be in sync.Data quality will help to resolve the issues. Building a template to keep the consistency is another way. Using error handling logs while loading the data to target system will help to see if any errors with data format.

Web scrapping will become easy if application supports API, backend resolves issues intact with cross browsing and elements identification issues. Testing proper APIs for web scrapping will be helpful. If APIs are not available then we can have a template where we can keep the html or xpath of the elements which are dynamic in nature than keeping inside code. I have seen this with Salesforce application a lot. Before we deploy to prod one can see if elements still have same xpath as in lower environment.

In a project for a client, I faced a challenge with their internal application that displayed data in a format resembling a table, but it was actually constructed using div tags, not standard HTML table elements. This unconventional structure made data extraction across multiple pages complex. To address this, I implemented robust error handling in my RPA solution. Whenever the bot encountered unexpected layouts or missing elements, the error handling routines kicked in, ensuring the process didn't fail. I also used detailed logs to track these occurrences, allowing me to refine my extraction logic for better handling of these pseudo-tables. This approach significantly improved the reliability and efficiency of the data extraction process.

In a project with a client using the IBM Lotus email system, a challenging older platform where standard protocols like SMTP and POP3 were failing more often than not, I leveraged a UiPath native package that I found on manage packages. The key difficulty was automating email sending and receiving processes with attachements and html body. To overcome this, I meticulously tested and debugged the process using UiPath's robust error handling to manage anomalies and exceptions. I implemented strategic breakpoints and logs to pinpoint issues, ensuring the bot could handle unique Lotus quirks.

Using DLL or a utility task to ensure that organisation email configuration are properly set than using each email command in a task. Error handling helps to find out where exactly the problem lies and logging it can help to identify which line got break. Using credentials vault to keep the global variables can mitigate such risks. Email automation is very important as it sends important notifications to buisness as well RPA admin teams to keep a check on not monitoring. Packages should be up to date. Format of email body should be used same across all the bots.

In a UI automation project for a bank's Citrix environment, I faced the challenge of automating a green screen application whose title changed with certain commands. To ensure reliable automation with UiPath, I analyzed these title variations and implemented dynamic selectors. These selectors, crafted with wildcard characters, adapted to title changes, allowing consistent identification and interaction with the application. I also refined individual UiPath activity selectors for robustness. Through rigorous testing and debugging, including step-by-step execution and error handling, I ensured smooth, adaptable automation. This approach was crucial in managing the application's dynamic nature.

UI automation requires a thorough error handling. Each window has to be captured properly and make use of wildcard in case same browser is not present in upper environment. Selecting the property of elements which is unique always help. Resolution of browser has to be checked properly. How bot going to work in attended and unattended mode for UI application needs to be tested properly. If any unwanted pop up comes that needs to be tested before deploying the bot in production.

The integration of vision-based RPA bots, like the ones that can be created using GPT-4v or UiPath's computer vision module, marks a significant shift in the automation landscape. Moving away from traditional selector and API-based approaches, these bots rely on visual interpretation of screens. That can be a future where we should relearn how to create RPA bots with a totally new approach.

Using AI based window capturing can help in case of any frames present in UI or legacy application automation. A proper logging helps support team to understand on their own where bot went wrong and what was expected from bot. This removes the dependency from developer and support team can work on their own. Using Comments, steps beneficial for support team to understand each step that bot will perform. Testing bot out of the box scenarios, sometimes help to understand how bot is going to behave if any unexpected scenarios come. Each click has to be interrupted during testing phase so that we can anticipate negative scenarios handling capabilities of bot execution.