Polyphonic HMI: Mixing Music and Math
This paper is an analysis of a case study originally conducted by the Harvard Business School in August of 2005 and is based on the challenges of introducing a new technology into a market place that for decades been based on “gut feelings and intuition”. The new technology was initially designed to assist consumers in music stores find music that met a certain criteria. Later this was changed because of a sharp decline in music sales. The new revision of the technology was designed to assist music producers, record companies, and artists in the selection of music that could be successful. Faced with a very small marketing budget the challenge of the marketing team was to decide what marketing plan would give the best results. I. Background
The company introducing the new technology was Polyphonic HMI. Polyphonic was a subdivision of Grupo AIA whose core competency was the use of artificial intelligence coupled with natural sciences to provide complex business solutions for their customers. They were a small company of approximately 50 people but had a wide portfolio of business interests that included energy, finance and ebusiness. In 2002 AIA decided to venture into the world of entertainment and introduce their tools into the industry. They did this by forming a new company called Polyphonic HMI. Polyphonic’s team consisted of a relatively small number of staff members and scientists but had access to the AIA’s data and scientist staff and was given an annual operating budget of around $500,000. The product being introduced was based on the science of analyzing music by its mathematical characteristics. The new technology named Music Recommendation System and used a database compiled of millions of songs to isolate features like melody, tempo, pitch, rhythm, and cord progression.  Initially the company decided to target the consumer market segment in which to introduce the technology. The concept was centered on a customer going into a big box music store not knowing which music to select or what genre they may like. By entering some basic information the new technology could assist them in the selection of particular songs that met their particular set of criteria. Thus enhancing the consumer’s experience by getting them the music they wanted to hear and increasing sales for the retailers. By the time the product was ready to release there had been a sharp decline in music sales in the big box stores. (Figure 1) It was believed that this was due to several factors: first, increasing pressure from online competition and the ability to download music through the internet. Second, the belief of the consumers that online music should be free.  Third, was the consumers move from vinyl, cassettes and CD’s to a more portable digital format.  The New Product
Faced with this Polyphonic and AIA revisited the design intent of the technology and created a new program called Hit Song Science (HSS). The way HSS worked was it looked at the same information as the previous version but instead of looking for music with certain criteria entered by an individual consumer it looked at the similarities of songs that had made the “Top 40” hit list and the melody characteristics of each. What the Polyphonic’s team found was that there were clusters each hit fell into and Polyphonic’s scientists felt that their system could accurately predict which songs would be hits and which ones would not, based on the analysis of HSS. After continued analysis and refinements in the program Polyphonic’s found that they could predict whether or not a song would be a hit with an 80% success rate. The calculations were made based on the weighted score found by HSS. If the songs weighted score was 7.0 or greater there was a good probability that the song would be a hit. According to CEO Mike McCready there were still limitations to HSS  but if those...
Please join StudyMode to read the full document