Since the schemas are vastly different there's probably no alternative to building a custom ETL process. If you already have a workable solution to convert the

Since the schemas are vastly different there's probably no alternative to building a custom ETL process. If you already have a workable solution to convert the data then you should probably just get to work and do what you can to keep your unknowns to a minimum even if it is a slow and tedious process. But if you don't really know if what you're doing will work then be honest about that upfront - you'll need to factor in additional time to compensate for mistakes. Try doing your conversions in a series of small steps with the appropriate testing along the way. I've seen projects that sound like yours go way over budget because of well-meaning attempts to use "silver bullet" tools (often championed by non-stakeholders) that either took too long to figure out how to use properly or didn't end up working as well as expected.