Image recognition is the last approach to automating applications if nothing else works to identify UI elements on the screen (like selectors or keyboard shortcuts). Because image matching requires elements to be fully visible on the screen and that all visible details are the same at runtime as during development, when resorting to image automation extra care should be taken to ensure the reliability of the process. Selecting more/less of an image than needed might lead to an image not found or a false positive match.
Image matching is sensitive to environment variations like desktop theme or screen resolution. When the application runs in Citrix, the resolution should be kept greater or equal than when recording the workflows. Otherwise, small image distortions can be compensated by slightly lowering the captured image Accuracy factor.
Check how the application layout adjusts itself to different resolutions to ensure visual elements proximity, especially in the case of coordinate based techniques like relative click and relative scrape.
If the automation supports different resolutions, parallel recordings can be placed inside a PickBranch activity and Robot will use either match.
If OCR returns good results for the application, text automation is a good alternative to minimize environment influence. Google Tesseract engine works better for smaller areas and Microsoft MODI for larger ones.
Using the MODI engine in loop automations can sometimes create memory leaks. This is why it is recommended that scraping done with MODI be invoked via a separate workflow, using the Isolated property.