Research on discoverability at LATICCE point to potential weaknesses of music recommendation in niche market
Using a premium subscriber profile as a reference, our laboratory affiliated to UQAM political science department and to the Centre for research on integration and globalization (CEIM) has conducted tests on music streaming services from March to August 2019. Our singular experiment targeted measuring the quality of service of transnational streaming platforms in a substantial niche market based on a repertoire of over 40 years of independent music production.
The Canadian province of Quebec has a mature recording industry from which the major labels and distributors have been largely absent since the 1980’s. Despite this, it has enabled international careers to develop and strive, notably the likes of Celine Dion, Coeur de Pirate, Grimes, Chromeo or Patrick Watson. Another aspect of this niche market is its majority proportion of french expression artists which can be isolated in terms of research material and studied as a specific cultural phenomenon. Very few western cultural environments have such a circumscribed popular music repertoire largely adopted and consumed by its locals. A recent article in daily newspaper Le Devoir1 cites sources who warns federal political parties in the context of the upcoming Canadian fall elections about the risk of erosion of this specific culture. Solange Drouin, the spokesperson for ADISQ, a Quebec indie label association, says musical albums have dropped from 13 millions units sold in 2005 to less than 3 millions last year.
LATICCE is a constituency of the University du Québec à Montreal (UQAM) and its name stands for Research Laboratory on Discoverability of Cultural Content in a Digital Transformation Context (Laboratoire de recherche sur la découvrabilité et les transformations des industries culturelles à l’ère du commerce électronique). Our small team of around 10 people has been working for the last 2 years on the creation of a discoverability index allowing to measure levels of inventory and quality of recommendations of niche repertoires like the one of Quebec’s french expression indie artists. We believe our raw findings can foster deeper investigations into how taste profiling algorithms can generate proper music playlist construction in the interest of the diversity of cultural expressions. We also believe streaming platforms need to improve targeting subscribers in local markets to retain proper levels of adoption for their products, reflect users preferences and their “discoverability profiles”. Our recent findings could do just that.
Our method is based on some six data point levels : Presence, Visibility and Recommendation which itself lies upon Concordance, Relevance and Freshness (levels of new releases). We have used sample lists of official new releases and historical winners of 40 years of Felix Gala, Quebec’s equivalent of the Grammy’s or Juno’s. We have built persona and user stories and done API queries, web scrapping and human observations on three major music streaming services : Apple Music, Google Play Music and Spotify.
Our most interesting findings are with regards to our reference persona. Designed as an early thirties women, our “niche profiled” subscriber (we will not disclose her name) has opened premium accounts on streaming services. She has bought, added to her library and listened to her favorite music on curated playlists and her own personalized playlist for 21 weeks.
Despite strong emphasis by music services on the quality of the personalized music experience they offer their subscribers, none of the 21 weeks of listening offered nearly what our persona expected. The services offered very varied Cold Start issue reaction (the ability of recommending content with very little data history), the worst being Apple who could not generate any automatic playlist after six weeks of paying subscriptions, the most intriguing being Google with propositions the first day. Overall, the services took very large liberties with the taste profile of our virtual guinea-pig. Our young women has been voluntarily listening to a very consistent and narrow scope of music : she only listens to Quebec based artists signing in French. She listens to this type of content whether we are talking legacy performers or fresh new young Quebec hip-hop artists. Nevertheless, she ONLY listens to french Quebecois stuff. She was given an average content well below 50/100 discoverability index. Overall the services performance can be characterized by the following : Apple New Music Mix has shown little discrimination with regards to artist notoriety, digging too often in new releases without any traction or history and mixing genres and languages of expression in a very chaotic way; Google J’ai de la chance (I’m feeling lucky Dice App) has on the other extreme offered a very high level of concordance with our initial samples list and been constantly coming back to very expected local mainstream artists; finally Spotify Radar des sorties (Release Radar) had the best mixed qualities, but it offered very many propositions outside of our narrow casted persona, showing way too much random decisions.
All premium streaming services failed to match the very precisely tailored expectations of our niche listener.
To find out more about LATICCE get your french dictionary out and surf to our Wiki space at https://wiki.uqam.ca/display/CEIMLAT/Accueil+LATICCE or write me at firstname.lastname@example.org