Analyzing "Brand Ticker" data.

The Brand Ticker provides brand specific marketing data. On a day to day basis they list the brand value and top 3 associations gathered from news mentions and social data. They provide no documentation on how they process the data or which sources they use.

Here is short look at their service. I pulled half a year’s data in irregular intervals from their API using a small r-script. See this GIST

Due to some recent attention we start the analysis by looking at the Volkswagen (VW) data.

##    attr_1 attr_1_flag      attr_2 attr_2_flag   attr_3 attr_3_flag
## 25  liked       green complicated         red advanced       green
## 26  faked         red   polluting         red  secured       green
## 27  liked       green complicated         red advanced       green
## 28  faked         red   polluting         red  secured       green
## 29  faked         red   polluting         red  secured       green
## 30  faked         red   polluting         red  secured       green
##      change    value         brand brandId       Date
## 25 2.957404 5109.233 Volkswagen VW     439 2015-11-25
## 26 3.678028 5389.976 Volkswagen VW     439 2015-11-27
## 27 7.359091 6147.744 Volkswagen VW     439 2015-12-02
## 28 7.359091 6147.744 Volkswagen VW     439 2015-12-02
## 29 2.977252 6574.628 Volkswagen VW     439 2015-12-08
## 30 2.977252 6574.628 Volkswagen VW     439 2015-12-08

The first 6 columns are the brand associations, gathered from news data, or a survey (who knows…?). It’s helpful that the flag attributes already label the associations as positiv/negativ (red or green). The value column is the so-called “Brand Value”.

plot of chunk unnamed-chunk-3

In order to see if the brand value is related to associations, I color-code the brand value with the top (attr_1) association.

plot of chunk unnamed-chunk-4

For the diesel-gate event, we see that the top association is a rather lagging than predicting variable. The brand value massively drops, the association remains “secure”. The association change from “secure” to “polluting” happens with a distinct lag. In order to visualize the second and third association, we can use a simple histogram of mentions in the data set.

plot of chunk unnamed-chunk-5

As before, we can add all associations to the brand value time series.

plot of chunk unnamed-chunk-6

In order to unterstand the brand value variable it bit better, I scraped VW’s stock market data from Yahoo. See this GIST

plot of chunk unnamed-chunk-7

Apparently brand value is just a linear function of stock market data and can be regarded as more or less identical.

Brandtickers Brand value has a direct relation to VW's stock price
Dependent variable:
value
(1)(2)
Close365.498***273.071***
(22.435)(6.069)
Constant-3,057.386***
(724.722)
Observations3030
R20.9050.986
Adjusted R20.9010.985
Residual Std. Error854.555 (df = 28)1,073.895 (df = 29)
F Statistic265.414*** (df = 1; 28)2,024.172*** (df = 1; 29)
Note:*p<0.1; **p<0.05; ***p<0.01

I find it quite misleading to speak of “Brand Value” and taking stock market data as a reference. Plenty of factors (economic condictions, production costs, …) have an impact on stock prices, while the value of a brand is related to subjective perceptions.

Written on December 13, 2015