Pop-song paradox: What it takes to be a one-hit wonder is the opposite of what it takes to have a long career.

Science News examines the science of the Top 40, with a survey of the charts that found what goes into that first, breakthrough hit is the opposite of what makes for an ongoing musical career:

Artists with more variety in their catalog have a better chance to land repeat hits, says Justin Berg, a social scientist who researches creativity and innovation at Stanford University’s business school. But there’s a dilemma for artists who want to be popular over the long-term. Variety isn’t what helps artists land that first hit, Berg reports March 24 in Administrative Science Quarterly. It’s the similarity of a new song to recent hits.

“There actually isn’t a way to thread the needle,” Berg says. “You face a … trade-off as a new creator, between a likelihood of initial [or] sustained success based on the novelty of your portfolio.”

Berg used a database of about 3 million songs from 1959–2010 released by record labels that had produced at least one hit in the United States over that time. Of those songs, nearly 25,000 landed on the weekly Billboard Hot 100, which tracks the most popular songs based on sales data, radio play and now online streaming. That provided Berg with a list of nearly 4,900 artists who had one or more songs that made the list, his yardstick for defining a hit.

Berg then turned to a Spotify system that rates songs on 11 variables, including danceability, energy and key. This system provided metrics on most of the hits and nonhits from the 1959–2010 time window. Berg then noted how closely related hit songs were to the hits from the previous calendar year. He also compiled portfolios for most of the artists who had at least one song on the Hot 100, so he could quantify the variety and novelty of the songs they had released at the time of their first hit. These portfolios also allowed him to compare one-hit wonders to mega-hitmakers and to those who never made it big.

Hits are rare, the data show. Of the 69,000 artists in the original database, 93 percent never had a hit, 3 percent had one and 1 percent had two hits. The success rate for additional hits drops from there.

Berg found that musical artists with what he termed low-novelty portfolios that closely resembled other already existing music were about twice as likely to have initial success. But those who built a more innovative and varied catalog before fame hit were more likely to generate a series of hits.

You can read Berg’s research here, in Administrative Science Quarterly.