The Unbundling of Industry Mind-Set Metrics:
Investigating the Effect of Own and Competitor
Mind-Set Metrics in an Advertising-Sales Relationship using
Dynamic Hierarchical Factor Models
Peter van Voornveld
S2785978
MSc Marketing Intelligence & Management Supervisor: dr. K. Dehmamy
Second reader: prof. dr. J.E. Wieringa
Introduction
Data about the value of a brand in the “hearts and minds” of
consumers is increasing.1
Mind-set metrics represent complex intermediate mind-set states, which evolve in conjunction with competitor mind-set metrics. 2 • Industry movement 3
• Batch dependent measurement errors 4
Incorporating mind-set metrics of the focal brand and its competitor into modelling efforts is challenging.
• Collinearity & Overparameterization 5,6
Through the application of a dynamic hierarchical factor model, the interconnected nature of industry mind-set metrics can be unbundled.
Incorporating this latent knowledge could improve the predictive
performance of sales in the advertisement-sales relationship. 7
Research Question
Do own and competitor intermediate mind-set states
extracted using a dynamic hierarchical factor model provide a
useful basis to improve the ability to predict sales in the
Data Description
• The dataset is provided by MeMo² and ***.
• Motor insurance division
• 117 weeks of data: January 01, 2016, to March 26, 2018.
• Car insurance sales, advertising channels: television, radio and online.
• Competitors: COMPETITOR 1, COMPETITOR 2, COMPETITOR 3, COMPETITOR 4, COMPETITOR 5,
COMPETITOR 6, COMPETITOR 7 and COMPETITOR 8.
• Mind-metrics: Top of mind awareness, spontaneous brand awareness, aided brand awareness, aided
campaign awareness, brand consideration and brand preference.
Model Choice
Dynamic Hierarchical Factor Model (DHFM)
Interpretation: Variance Decomposition• Captures movement on different hierarchy levels: 1. Market dynamics (3)
2. Mind-set states dynamics (2,2) 3. Competitor dynamics (1)
4. Mind-set metric dynamics (1)
Factor Augmented Vector AutoRegressive model with
Exogenous variables (FAVARX)
Interpretation: Impulse Response Functions, Generalized Forecast Error Variance Decomposition and Forecasting
• FAVARX takes into account the dynamics and feedback
effects of the intermediate mind-set states and the
competitive actions. 7
• FAVARX allows to incorporate endogenous and exogenous variables. 8
Results
• Variance Decomposition
• The mind-set state level mostly explains spontaneous brand awareness, aided campaign awareness and brand preference of ***. • The mind-set state level mostly explains variance the competitors COMPETITOR 3, COMPETITOR 1, COMPETITOR 2 and
COMPETITOR 8.
• Comparing the predictive performance of the FAVARX vs. the benchmark model
• The FAVARX model outperforms the VARX model based on the average squared prediction error: FAVARX 0.303, VARX 0.379 • Both models outperform a naïve based on the Theil’s U-statistic: FAVARX 0.842, VARX 0.942
108 109 110 111 112 113 114 115 116 117 -1,25 -1,00 -0,75 -0,50 -0,25 0,00 0,25 0,50 0,75 1,00 1,25
Results
Hypothesis Expected relationship Findings
H1: Effect of advertising on own mind-set metrics + Television on own factor 1: Negative effect
Television on own factor 2 & radio on own factor 1 and 2: No effect H2: Effect of advertising on sales + Television: Positive effect
Radio: Positive and negative effects Online: No effect
H3: Effect of own mind-set metrics on sales + Own mind-set factor 1 and 2: No effect H4: Effect of competitor mind-set metrics on sales - Competitor mind-set factor 2: Negative effect
Competitor mind-set factor 1: No effect
Sales Television Radio Mind-set factors
59,9% 13,9% 6,4% 3,8% 5,4% 5,1% 5,5% 19,8%
FAVARX
75,9% 17,7% 6,4%VARX
*** mind-set factor 1 *** mind-set factor 2Competitor mind-set factor 1 Competitor mind-set factor 2
Generalized Forecast Error Variance Decomposition
Discussion and Conclusion
Do own and competitor intermediate mind-set states extracted using a dynamic hierarchical factor model
provide a useful basis to improve the ability to predict sales in the advertising-sales relationship?
The performance did not improve significantly. Although non-significant the performance did improve by 25,1% based on
the average squared prediction error statistic.
• Television show to have a positive effect, where radio has a mixed effect and online has no effect.
• Type of industry
• Composition campaign
Managerial Implications
Data Collection Sales Growth Competitors
Tracking own and competitor
mind-set metric could be very
useful.
- Short term sales predictions
- Signalling proxy of competitor
branding efforts
*** should reconsider the
usefulness of radio for
growing sales.
To analyse the branding
performance of a company, the
relevance of competitor
branding efforts should not be
underestimated.
Limitation and Future Research
• Dataset
• Sales are based on registrations.
• Different type of industry to test the relevance of the research framework.
• FAVARX estimation
• Non-normality was found in the residuals.
• Not all draws are checked for the model assumptions.
• Not accounted for industry relevant factors, like car sales.
• Interpretation
Thanks for your attention!
References
1. Bowman, D., & Gatignon, H. (2009). Market Response and Marketing Mix Models: Trends and Research Opportunities.
Foundations and Trends in Marketing, 4(3), 129–207.
2. Park, J., Dillon, W. R., Lee, S., & Chaiy, S. (2015). Getting the Most Out of Customer Satisfaction Tracking: A Dynamic Multilevel
Structural Equation Model of Key Performance Metrics (Working Paper).
3. Rajagopal. (2008). Measuring brand performance through metrics application. Measuring Business Excellence, 12(4), 29–38.
4. Sridhar, S., Naik, P. A., & Kelkar, A. (2017). Metrics unreliability and marketing overspending. International Journal of Research
in Marketing, 34(4), 761–779.
5. Farley, J. U., Lehmann, D. R., & Mann, L. H. (1998). Designing the Next Study for Maximum Impact. Journal of Marketing
Research, 35(4), 496–501.
6. Gelper, S., Wilms, I., & Croux, C. (2016). Identifying Demand Effects in a Large Network of Product Categories. Journal of
Retailing, 92(1), 25–39.
7. Srinivasan, S., Vanhuele, M., & Pauwels, K. (2010). Mind-Set Metrics in Market Response Models: An Integrative Approach.
Journal of Marketing Research, 47(4), 672–684.