> with no source control This is bad. > It's all require_once. > it doesn't use any framework This is not necessarily bad. > no code has ever been deleted

> with no source control This is bad. > It's all require_once. > it doesn't use any framework This is not necessarily bad. > no code has ever been deleted This is bad > Multiple versions of jQuery This is bad > a full rewrite is necessary, but how to balance it? You never need to fully rewrite something. You can always take an incremental approach. If the code and 3 people generate 20 million dollars of revenue (on their own? or with massive sales support? what is the cost of goods?) then it's got to be doing something right. I'd start with the source control. Just check everything in, so it's easy to go back. Do that on the server they develop on, even. (But have a script that pushes the checked-in code to offsite.) Second, make it possible to spin up a second instance of the same application, in some automated fashion, out of bare source control. This may mean dumping schemas and checking them in, and probably figuring out what data in the database are "necessary configuration" versus "user payload data." Then, you can initiate integration testing on top of the second cluster. You can also turn this into some kind of local sandbox development setup. Once that is done, you may be able to change the code quicker, because you can do and test it locally, and perhaps have some acceptance tests on top of the mudball. At that point, you can switch over to doing development locally and deploying (and, ideally, having the ability to un-deploy.) After that, starting to clean up should at least be possible with less risk, because you can test it in isolation. You can then start pulling on threads in the code, such as standardizing library versions, detecting and deleting unused code, putting like modules together, and so on. You don't need fancy tools for managing this, shell scripts and command-line git are probably plenty enough. Resist the temptation to spend six months engineering the build system of the future! Of course a lot will depend on details, but from your brief description, this sounds like the path forward -- focus on making it possible and safe and cheap to iterate, and then you can get on with actual iteration. Don't waste time on big rewrites; instead do things incrementally. Don't believe that any one tool will save the day, because it won't.