Since we're on the subject of streaming vs. traditional TV, here are some n=1 observations.
Even in the case of free (or, more generally, no marginal cost) programming, I'll wager we've all felt cheated after watching a disappointing show. "There's an hour I'll never get back" sums the feeling up nicely. Therefore, part of getting potential viewers to sample a show is creating a low perceived viewing cost and risk of wasted time.
Here's where the n=1 comes in. I've noticed when I channel surf or I realize that a favorite show is about to start, I tend to underestimate the commitment associated with checking it out while simultaneously overestimating the scarcity. Since I don't know when my next chance to see it will be, I'll just watch until Columbo comes in or the Magnificent Seven talk to the old man or Ray Liota takes Lorraine Bracco to the Copa. Invariably, though, when those moments pass, I keep watching.
From a marketing standpoint, the combination of low perceived cost and scarcity is extraordinarily powerful (check out any marketing psychology book for examples), By comparison, when I sit down to watch something on-demand there's no sense of scarcity, I'm more aware of opportunity costs and, for some reason, I feel more like I'm making a commitment to watch the whole show -- opening credits to closing credits.
What a Hulu or a Netflix does have is a selling point often inversely related to scarcity, convenience. "Watch what you want to watch it." The key piece there is "what you want to watch." Once you've made your selection, convenience becomes a big plus.
It is often useful to think of marketing as three stages of persuasion:
1. Convince customers they want a product;
2. Convince them they want to buy it from you;
3. Convince them they want it now.
Recommendation algorithms do a good job with the third and, if you're seeing the recommendation, the second probably isn't a problem. I do not, however, see them doing a good job with the first, at least not without a great deal of help.