an effective alternative to writing coding tests is to automate the comparison of output data between old code -vs- new code using same set of input data ... th

an effective alternative to writing coding tests is to automate the comparison of output data between old code -vs- new code using same set of input data ... this is valid when the initial few prod deployments had people manually confirm the output data was as expected I worked in a global investment bank for many years on very large teams building systems responsible for processing hundreds of billions of $ in real time trading transactions where this approach not only worked it facilitates extremely productive use of software engineering talent We also published prod data tapped and published from various points in our data processing pipeline which was subscribed to by QA prod-parallel which did the burn in of all new code ... it was fed live prod data and output of QA was compared to prod output in real time I created a fuzzy match process which gracefully handled gradations of data match when faults where encountered when comparing data feeds from prod -vs- QA ... this allowed people to focus on chasing down the data diffs and not wasting their time writing coding tests which would have been very quickly obsolete