Thanks for the help:
The effect of content-based recommendations on advertisement
By:
Joost Dorgelo
Content
› Introduction
› Recommendations
› Music
› Online advertising
› Conceptual model
› Method
› Result
› Discussion
First look
›Research Question: what are the effects of
content-based recommendation system on the amount of
banner impressions and the click trough rate?
Recommendations
› Shift from analog to digital music storage. › Increased choice.
▪ More difficult to find music of the preferred type (Kaminskas & Ricci,
2012; Chung, Rust & Wedel, 2007).
› Consumers get help with finding their preferred music
(Hennig-Thurau, Marchand & Marx, 2012; Fleder & Hosanagar, 2009; Kaminskas & Ricci, 2012).
› The recommendation systems improve sales numbers (Fleder & Hosanagar, 2009).
▪ Not by increasing the amount of options, but by increasing the
Effect on advertisement
› Money is earned indirectly.
- by targeting the user with advertisement.
› Many online platforms do not rely on sales, but are dependent on advertisements.
Why do we use technology in the first place
› TAM Model by Davis (1986).
▪ Usefulness & efficiency.
Music type
› Genre classification
(Silver, Lee & Childress, 2016).-
Grouping of the industry,
-
Listening to the same type of music.
› 25 music genres -> 6 music types
(Shäfer & Sedlmeier, 2009).-
Electronic, Sophisticated, Rock, Pop, Folk, Rap.
-
These 6 music types are related to the fulfillment
Recommendations based on music type
› Recommendations based on music type, therefore
believed increase the usability of the platform for the
user. This is a main driver for the use of the platform
› H1: recommendations based on music preference are more likely used than generic
recommendations.
› H2: recommendations based on the music preference will increase the amount of
visits to the website.
› H3: recommendations based on the music preference will increase the duration of
the visits of the website.
Individual differences
› Different personality’s have different music preferences (Rentfrow & Gosling, 2003).
› H4: People with a preference for Rap, Pop or Electronic music, are more likely to
use recommendations based on music type.
› Rap and Electronic are linked to agreeableness, making them more likely to use the recommendations.
› Pop is negatively linked to openness to new experience.
Online advertising
› For a publisher, every visit is a product sold (Chen & Stallaert, 2014)
- H2: recommendations based on the music preference will increase the amount
of visits to the website.
›
Click trough rate- Increased duration of exposure towards the banner
advertisement, increases the probability to click on the banner
(Ghose & Todri, 2015).
› H5: Visit duration has a positive effect on click trough rate.
› H6: Duration of the website visits mediate the effect that recommendations based
Classification & Clustering
› Music streaming service Songa.
› 301 randomly selected users.
› Artist classified based on music type.
› Recommend music based on their preference.
› Test and control group.
›K-means++.
Recommendations
› Every user gets five
recommendations
based on their music
preference.
› Placed in the
recommendations
Classification
› No Sophisticated and Folk preference.
› Dutch folk music and Pop/Rock preference.
Music type Amount of songs % % cumulative
Clustering
›DB index of 33 at six clusters was the lowest score.
Hypothesis testing
› H1: Music type recommendations increases recommender use.
- Binary logistic regression.
- Not significant
› H2: Recommender use increases the amount of visits.
- Linear regression.
- Not significant
› H3: Recommender use increases the visit duration.
- Linear regression.
Hypothesis testing
› H4: Electronic, Rap or Pop are more likely to use the recommendations based on music type.
- Binary logistic regression.
- Having a preference for Electronic, Rap or Pop music, makes a user
more likely to use the recommendations based on music type. › Further analysis of Pop preference
- Majority of the uses was caused by Pop listeners.
- Further analysis turned out that only Pop listeners are more likely to
use the recommendations.
Hypothesis testing
› H5: visit duration increases the click trough rate.
- Linear regression.
- Not significant
› H6: the effect of recommender use on click trough rate is mediated by visit duration.
- Non parametric bootstrap mediation analysis (Preacher & Hayes,
2004).
Discussion
› Based on the collected data:
- The use of content-based recommendations based on music
type is discouraged, if maximizing advertising revenue is the goal.
- Pop listeners are more likely to use recommendations systems
that are based on music type.
- Pop and Rock music are the most popular genres.
Additional findings
›
Electronic listeners have shorter session (marginally significant).- 27 minutes lower than the average. - Relation with exercise.
- This confirms the importance of context.
› Increased amount of visits make users more likely to adopt the recommendations (marginally significant).
- Increasing odds of 5% per visit. - Relation with boredom.
Limitation & Future Research
› Short observation window.
› Only 1 set of recommendation. › Dutch Folk music.
› Only popular music was classified.
› Different types of recommender systems:
- Collaborative, - Context.
Thank you for your attention!
Literature
Anderson, C. 2009. FREE: The future of a Radical Price. New York: Hyperion Books.
Chen, J., & Stallaert, J. (2014). An economic analysis of online advertising using behavioral targeting. Mis Quarterly, 38(2), 429-449.
Chung, T. S., Rust, R. T., & Wedel, M. (2009). My mobile music: An adaptive personalization system for digital audio players. Marketing Science, 28(1), 52-68.
Davis, F. D. (1986). A technology acceptance model for empirically testing new end-user information systems: Theory and results (Doctoral dissertation, Massachusetts Institute of Technology).
Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: a comparison of two theoretical models. Management science, 35(8), 982-1003.
Fleder, D., & Hosanagar, K. (2009). Blockbuster culture's next rise or fall: The impact of recommender systems on sales diversity. Management science, 55(5), 697-712.
Häubl, G., & Murray, K. B. (2001). Recommending or persuading?: the impact of a shopping agent's algorithm on user behavior. In Proceedings of the 3rd ACM conference on Electronic Commerce (pp. 163-170). ACM.
Hemming, J. (2013). Is there a peak in popular music preference at a certain song-specific age? A replication of Holbrook & Schindler’s 1989 study. Musicae Scientiae, 17(3), 293-304.
Hennig-Thurau, T., Marchand, A., & Marx, P. (2012). Can automated group recommender systems help consumers make better choices?. Journal of Marketing, 76(5), 89-109.
Literature
Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior research methods, 36(4), 717-731.
Rentfrow, P. J., & Gosling, S. D. (2003). The do re mi's of everyday life: the structure and personality correlates of music preferences. Journal of personality and social psychology, 84(6), 1236.
Schäfer, T., & Sedlmeier, P. (2009). From the functions of music to music preference. Psychology of Music, 37(3), 279-300.
Silver, D., Lee, M., & Childress, C. C. (2016). Genre Complexes in Popular Music. PloS one, 11(5).