It all depends on the granularity. As Scrapperdog said, if it's per hand, it's a huge advantage -- you lose just over half your hands in BJ, so a 10% refund is worth over 5% overall. A 5% edge at BJ is unheard of.

But they probably award it per month, or something like that. If you play 100,000 hands per month at $10 each, and the house advantage at BJ is 1.0%, then you will have lost, on average, 1.0% x 100,000 x $10, or $10,000. Getting back 10% of that is a refund of $1000, reducing the house edge from 1.0% to .9%. That's not worth a whole lot.

Somewhere between the two extremes is reality. If you chose to play just a few hands per month, ten or twenty, you're effectively playing just one tiny session. Odds are pretty high that the session will have some wild variance. If you win big, great. If you lose big, you get a big 10% refund. The advantage of 10% back on the loser can actually make it a net positive game.

To have the 10% effect big enough to make it a winning game, you have to play so few hands (per refund) that you're not experiencing a long term average at all. You have to be experiencing mostly variance. . . This can be fun if you enjoy individual gambles and you want to place some bigger wagers. If you can't afford that, or don't get a thrill from playing 10 or 25 hands per month, you can't pull off the +EV trick. If you play enough hands per refund, you are guaranteed a refund every month simply because you are guaranteed a loss!

It would be interesting for someone to calculate how many hands per month can be played with a 10% loss refund to equate to a real house advantage of 0%.