Google
Trends is an online search tool that allows user to see how often particular
keywords, subjects or phrases have been searched and asked for over a specific
period of time. Just like that, Google trends graphs often show a term that is
used over time and where geographically most people are searching for a given term.
Significantly,
Google trend enable us to perform advanced keyword analysis, examine seasonal
search patterns, discover a new untapped keyword, increase relevant traffic
sales and save individual months of time and energy (Lyons, 2010).
Particularly, Google
trends also provide daily and weekly reports on the volume of queries, which
related to various industries and companies. As a matter of fact, have you ever
noticed that the search share for coupon increases during the holiday shopping
season and the summer vacation season? The query (car tire) would be assigned
to category vehicle tires, which is a subcategory of auto parts that is a
subcategory of automotive. Nevertheless, research showed that they
are not claiming that Google trends data can help in predicting the future but
they are claiming that Google trends might help in predicting the present. For
instance, the volume
of queries on automobile sales during the second week in June may be helpful in predicting the June
auto sales report, which is released several weeks later in July. It may also be true
that June queries help to predict July sales, but researcher leave that
question for future research, as this depends very much on the particular time
series in question (Hyunyoung Choi, Predicting the Present
with Google Trends, 2011). As a
result, researcher found out
that queries can be quite useful in leading indicators for subsequent consumer purchases
in situations where consumers start planning purchases significantly in advance of their
actual purchase decision.
Basically,
there are many forecasting methods that can be used with Google trend data but
we will point out four potential methods that usually implement from time to
time. Firstly, linear regression is one of
the widely use method for forecasting method since it could analyze the use of
an independent variable to predict a dependent variable. To bear in mind, we
can use changes in an independent variable to forecast changes in dependent
variable if there is a strong correlation between a dependent variable and an
independent variable (theweeklytrade, 2010). Namely, the
author gave intensive information of search frequencies as an independent
variable, which used housing price and sale volume as an independent variable.
The research also had shown that there is a high correlation between search
frequencies and housing sales while there is a moderate relationship between
search frequencies and housing price. Secondly, talking about Google trends, it has characterized the
predictability of a trend series based on its historical performance (Weiss, 2009). According to
the research, in order to do so they compared the difference of forecasted
trends, which applied at some point in the past, to the trend actual performance.
Once the difference between the forecasted trends and the actual trends is
smaller than a predefined level, Google represents the trends query as
predictable. On the other hand, the article also stated that Google trends
provide a time series index of the volume of queries users enter into Google in
a given geographic area. The query index is based on query share, which equal
to the total query volume for the search term in question within a particular
geographic region divided by the total number of queries in that region during
the time period being examined (Hyunyoung Choi, Predicting the Present with Google Trends, 2011).
As a matter of fact, Google
trend index value or formula = {search queries volume at period (relative
value) / Total search volume (highest relative value)} *100
The study had shown that the
scale is presented in the range of 0-100, where 100 represent the search peak
or the highest frequency and intensity of searching activity for the specific
query. Trends
Index values for every period are calculated by dividing the relative value by
the highest relative value (Hyunyoung Choi, Predicting the
Present with Google Trends, 2011).
Table 3 numerical example of
Google Trends computation.
1
|
2
|
3
|
4
|
5
|
6
|
7
|
|
A number of new search
quires (A)
|
100
|
200
|
300
|
400
|
500
|
600
|
1200
|
Total volume of search queries
(B)
|
500
|
700
|
1000
|
1400
|
1900
|
2500
|
3700
|
Relative value (=A/B)
|
0.20
|
0.29
|
0.30
|
0.29
|
0.23
|
0.24
|
0.32
|
Google Trends Index value1
|
62
|
88
|
93
|
88
|
81
|
74
|
100
|
Source: Compiled by the
authors. (Gerard Chmyznikov, 2013)
Thirdly, another model or
method would follow the following model, which use for analyzing and
forecasting with Google trends data. The author assumed that model predicts and
forecast about the sale of a company (Hyunyoung Choi, Predicting the
Present with Google Trends, 2011).
+Model 0:
-log(yt) ~ log(yt-1) + log(yt-12)
+ et
This model predicts the sales
of this month using the sales of last month and 12 months ago
+Model 1
-log(yt) ~ log(yt-1) + log(yt-12)
+ xt(1) + et
This model uses an extra predictor,
i.e. Google query index to predict the sales of the present.
-log(yt) = 2.312 + 0.114*log(yt-1)
+ 0.709*log(yt-12) + 0.006*xt(1)
Sales of present month is
positively correlated with the sales of last month, the month 12 months before and
the Google query
-log(yt) = 2.007 + 0.105*log(yt-1)
+ 0.735*log(yt-12)0.005*xt(1)+0.324*I(July 2005)
Remarkably, there was a
special promotion week in July 2005, so they have added a dummy variable to control
for that observation and re-estimated the model
By the same token, it is
essential to use prediction error and mean absolute error along with the model
above in order to produce the best result and come out effectively.
-Prediction error: Predicted
value – observed value
Thus, it would come out with
graph below:
Last but definitely not least,
another forecasting method that could be used with Google trends is R, which
would be able to convert the data to a time series with a 52-week frequency
(Google Trends reports weekly search data) (Morrison, 2012). It seems fantastically useful.
Below are the codes (including all the Google data):
Which result in
Hence, converted
the search index time series into a Holt Winters object. The Holt Winters
function comes as standard with R, so nothing to do here. But this is just a
proof-of-concept run:
Which produced a
nice looking fit, based on the trend and 52 individual coefficients:
Forecast package,
and run the forecast. Holt Winters function for the next year.
Which, in turn,
gives this:
Forecast, the
orange area gives the 80% confidence interval and the yellow the 95% confidence
interval (Morrison, 2012).
In
conclusion, Google trends
give user with the consumption of choices, check competitors especially would
be pivotal for blogger or website which can help drive more traffic to their
own site. Undoubtedly, Google trends is not only useful for business sector or
marketer but also in many other aspect of life since people can use it as a
path of better decision making and self awareness for their own living.
Bibliography
Weiss, D. T. (2009, September 07). On Google’s new
forecasting capabilities and their importance to Market Research.
Retrieved from trendsspotting.com: http://www.trendsspotting.com/blog/?p=1536
Morrison, M. (2012, November 26). Forecasting Google
search volume using R. Retrieved from blog.magicbeanlab.com:
http://blog.magicbeanlab.com/data-viz/forecasting-search-volume/
Lyons, K. (2010, April 21). Case Study: Using Google
Trends to Discover New PPC. Retrieved from ppchero.com:
http://www.ppchero.com/case-study-using-google-trends-to-discover-new-ppc-opportunities/
Gerard Chmyznikov, L. G. (2013, November). FORECASTING
ECONOMIC ACTIVITY IN THE BALTICS: LET US GOOGLE IT.
Hyunyoung Choi, H. V. (2011, December 18). Predicting the
Present with Google Trends.
theweeklytrade. (2010, October 08). Can Google Predict
the Future? Retrieved from hotstockmarket.com:
http://www.hotstockmarket.com/t/80890/can-google-predict-the-future
No comments:
Post a Comment