“Every listener on Pandora has been giving us an immense amount of data,” he told me last week. “Users teach us about their taste. When they come to “premium” it will be imbedded with their personal preferences. We have a big advantage on personalization.”

He’s referring to Pandora’s soon-to-be launched premium on-demand service, a costly upgrade from the company’s bread-and-butter business, an ad-supported online radio service currently used by some 80 million people. This constitutes a head start no other entrant to on-demand music would have. “We can market to them,” insists Westergren, the company’s CEO for the last year.

On the one hand, it’s hard not to admire Westergren’s pluck. His business is eroding, his stock price is down, and an activist investor called Corvex Management wants him to sell the company. On the other hand, it’s hard to believe Pandora can survive the onslaught. As Westergren himself says: “Amazon, Apple, Google. It’s not who you would choose for your competition.”

Get Data Sheet, Fortune’s technology newsletter, where this essay originally appeared.

Pandora all but invented the genre. It wasn’t the first streaming music company. AOL and Yahoo were much bigger. But Pandora offered something much cooler, a “genome” that mapped musical tastes and made highly relevant suggestions about what music users might like to hear next. But it was late to the I-want-what-I-want-and-I-want-it-now game, and it is furiously playing catch up. Westergren says some 450 attributes go into crafting a profile of what listeners like, a picture that can’t easily be replicated by others.

Common sense suggests Pandora is a goner. But music is emotional, and Pandora’s product is sometimes magical. Who knows, maybe quality might just win out.