If you already have a team or resources available with database skills and experience and availability of an ETL-tool then by all means it might be beneficial to take advantage of that.
QlikView can very effectively and efficiently be the one and only tool you need for small to mid-size and even larger-size implementations. It all depends on the skill level and ability to adopt and understand the possibilities and techniques that QlikView offers.
QlikView can make use of QVD-files that are single-files that represent a logical in-memory table as a disk-image. So these will be the equivalent of tables in a database. You can even use QVW's with only data and no UI to store ready-made datamodels or half-made self-service datamodels for different datamarts (QlikMarts). It is quite feasible and also documented how to do incremental loads to speed the ETL-process and handle things like history and slowly-changing dimensions. Incremental loads are covered in the QlikView Reference Manual in a separate chapter.
Daily data loads can be handled by having reloads scheduled on the QlikView Server or the QlikView Publisher (if you have a license for that). They can be scheduled according to many criteria and dependencies - especially if you have QlikView Publisher.
Predictive analytics is not as far as I know related to the ETL-part but can be handled first and foremost in the Application layer of QlikView - unless you integrate things like R in the load scripts (QlikView ETL) which is also possible and there are examples on how to do that. Predictive analytics might be better served by using Alteryx which integrate well as a source for QlikView.
