But the trick is, you probably couldn't do a better job. There is little technical challenge in a product like this. Building crud apps that scale to low tens

But the trick is, you probably couldn't do a better job. There is little technical challenge in a product like this. Building crud apps that scale to low tens of thousands of users is not that challenging. Since (I believe) we all share that belief, then why do projects like this keep failing? I've worked on large medical information systems. The problems those encounter are (1) you essentially never get to bulk replace entire systems; you will have to support ongoing data ingestion and export to other running systems, often systems not designed to support this or with poor apis; (1a) if / when you fuck up, critical processes are disrupted. You have to engineer your systems and create backup processes to make this very unlikely to happen. There is a case I can't really share too many details of where a database was accidentally dropped in a datacenter of a hospital chain you've heard of. Said hospital chain was unable to access medical records for 20+ hours; this was in the 90s running on giant hpux machines and it took a long time to reimport data off tape. My employer was on the hook for $30M iirc, and I wouldn't be surprised if this actually contributed to serious patient injuries or deaths. Now obviously law enforcement isn't quite as critical as medical records, but it's not far behind. (2) over the life of such systems, lots of workflows / processes have developed around using them inside hospitals or large doctors offices. You have to discover these processes, mostly stored in the heads of line level workers and their immediate supervisors, and figure out how to enable them in your shiny new system. To be concrete: the medical record system didn't support some requirement. That didn't stop users from needing it. So employees or doctors decided to use a field in a certain way, and they know how to interpret the presence and/or data in that field. You don't even know a field is being repurposed, let alone how they actually use it. This will be a large problem importing all the old data and in recreating the necessary functionality. (3) no one person, and probably no small set of people, know or understand all the ways the system is used. Somehow, you have to capture that knowledge and use it to create a functioning new system. At least the medical record company I worked for hired lots of doctors and nurses out of hospitals to help them understand how users used the system in practice. You essentially don't understand at all how prosecutors, police, secretaries, and all the other people involved in law enforcement use this system. There are clear parallels to replacing systems like this. Particularly when you have to do a live transition.