The landing page on Songza.com declared that it was Tuesday afternoon and displayed five options: "Boosting Energy," "Brand New Music," "Working to a Beat," Working (Without Lyrics)," or "Working in an Office." Click through the last option and five more appear — for Old Hits, Current Hits, Country, "Indie Music That's Not Too Weird," and "Easy, Breezy Summer Songs." Clicking the last category again, I arrived at a selection of three playlists and selected one called "Lazy Summer." A vibraphone sample rang out and Sublime's "Doin' Time" began.
Earlier this month, Google bought Songza for a rumored $39 million (Google never disclosed the terms of the deal). In May, Apple paid $3 billion for Beats, a high-end headphone company with streaming services on the side. Digital downloads are down, and the tech giants' acquisitions acknowledged consumers' growing preference for online streaming. To market observers, Google's acquisition added to the mounting competition posed to Pandora.com, the Oakland-based company currently dominating the streaming market with more than 75 million users.
When Google elected Songza to combat Pandora, the company wagered on its particular approach to delivering free streaming content. Songza relies on human curators to create playlists, then a few questions guide users toward a set of songs suited to given situations. In contrast, Pandora relies on the Music Genome Project. The system breaks songs into more than 450 musical traits. Elusive qualities — such as emotional intensity — are quantified by analysts on a scale of one to five. Users input a song or artist to begin, and Pandora populates a playlist of similar tracks.
At stake is which service delivers likable songs more consistently. The choice, for consumers and for the investors upon whom companies like Pandora depend (the company is not currently profitable), is between the Music Genome Project's meticulous categorization and algorithms, and Songza's reliance on human curators. Underlying the comparison, however, are assumptions about how people listen to music and why. Do listeners respond positively or negatively to songs based strictly on identifiable sonic elements? Can a few questions and some data analysis guide a listener toward the perfect playlist? What privacy issues arise?
Music carries different associations for everybody. Some of these are informed by cultural input — what our friends like, what critics recommend — but others reflect memories, as anyone who's painfully reminded of past traumas by particular songs knows. Obviously, neither service accounts for users' personal baggage, but Songza's ethos asserts that it doesn't matter.
In a Forbes interview, Songza CEO Elias Roman claimed that musical preference mostly reflects what he calls "context," the where and what of a listener's activities at the moment of listening. People don't usually have time to listen to music for the experience's inherent value, he claimed, so the mission of Songza is "to facilitate or enhance situations — in other words, to make right now better," Roman said. "Music is the means. Lifestyle enhancement is the end."
"Lifestyle enhancement" entails data mining, a feature of the service that likely appealed to Google. To arrive at Sublime's "Doin' Time," I provided at least five data points to the website: day of the week, time of day, location, device type, and climate (Weather.com and Songza are partners). Earlier this year, The New York Times reported that Pandora's free version has begun tailoring advertisements to individual users based on listening history. It's nearly impossible to avoid giving up personal information online, but Songza's interface unabashedly requires users to provide personal data up front.
Music can certainly enhance experiences, as anyone who returns to the same genre or artist for a certain activity knows, but listeners seek different mood music in the same situations. On a bright and warm Sunday afternoon in a crowded public park, no single playlist accommodates the preferences of separate groups, even if they're all, say, having picnics. Tunes tailored to please everyone in a given situation do exist, though: They're called muzak, or elevator music, and they're designed to be ignored. Taken to its logical conclusion, perhaps Songza's reverence for context would deliver strictly innocuous instrumentals.
Pandora's exacting musicologists and Songza's playlist curators are styled as technicians of taste, but musical preference resists prediction and prescription. Listeners do look to individuals for recommendations, but those are specific critics, friends, and record store clerks with whom they have rapport and trust. The richness of musical discovery is largely extracted from those relationships. The depersonalized equivalent touted by streaming services is an inadequate replacement for recommendations by real people.
Nevertheless, Pandora, Songza, Beats, and the market's other major player, Spotify, are scaled to serve growing throngs of users. As long as the competing companies jockey for pieces of a growing market, benefactors and speculators will continue to neglect how streaming technology struggles to accommodate all of the cultural bias and intuitive emotional responses that comprise individual taste.
After all, I hate Sublime. However, Herbie Mann's version of "Summertime," from which "Doin' Time" samples its vibraphone vamp, is quite suited to an afternoon at the office.