In an interview with Robert A. Bjork this article suggests interleaving skills is better than focusing exclusively on one skill at a time. How much of a difference does the research indicate that it makes?
I think it would be unfair to reduce this effect to a single number-- most certainly, it depends on the task at hand, as well as a number of other factors.
To give a rough idea, though, you may want to look at Richland et al. (2005) who found roughly a 25% increase in retention on a knowledge test when using interleaving. Also noted in this article,
Interleaving has been shown to be more effective for producing learning than blocking or massing learning (A1A2A3A4A5B1B2B3B4B5) on many diverse sets of stimuli including word pairs (Battig, 1979), motor patterns (Shea & Morgan, 1979), and word translations (Richland, Bjork & Finley, 2004)
Robert Bjork claims that interleaving is only one intervention among a class of "desirable difficulties", or things that might make your task seem harder but will improve it in the long run. Other such desirable difficulties include generation (e.g., fill in the blank rather than multiple choice; see Karpicke & Roediger (2008)) and the spacing effect (studying at regular spaced intervals is more effective than studying everything all at once). Bjork has shown that interleaving is not just an artifact of the spacing effect--after all, interleaving requires spacing--but that its contribution to retention is in addition to spacing.
If you head on over to Bjork's lab webpage you'll see that he has a decades of research publications available free for download.
Richland, Bjork, Finley, & Linn (2005). Linking Cognitive Science to Education: Generation and Interleaving Effects. FREE PDF
Karpicke, & Roediger (2008). The critical importance of retrieval for learning. Science, 319, 966-968. Retreived from https://www.msu.edu/course/psy/101/altmann/karpickeRoediger08.pdf