« previous next »
Pages: [1]

Author Topic: Who knows how to really work Google Trends?  (Read 4243 times)

dogboy

  • Guest
Who knows how to really work Google Trends?
« on: December 04, 2012, 03:20:17 PM »
I have 5 or KW terms that have a strong, obvious relationship with one another.

The problem I'm having is the scale is out of whack because some terms are high frequency and some low frequency, so you see one big line, and the rest are pushed to the bottom. Here is an example:
https://www.google.com/trends/explore#q=blue%2C%20red%2C%20white%2C%20the&cmpt=q

...I need everything in one screen and I don't know how to tweak that screen.  I can download the data, in a CSV, but is there anything else I can do to represent this data?
« Last Edit: December 04, 2012, 04:34:01 PM by dogboy »
Logged

dogboy

  • Guest
Re: Who knows how to really work Google Trends?
« Reply #1 on: December 04, 2012, 04:33:31 PM »
oh, and that example, MickG, is only a random example; I am not trying to infer anything other than it's hard to see anything meaningful, because one variable overshadows the rest...

 ;D
Logged

mick g

  • Inner Core
  • Sr. Member
  • *
  • Posts: 445
    • View Profile
    • Driveways Bolton
Re: Who knows how to really work Google Trends?
« Reply #2 on: December 04, 2012, 06:05:02 PM »
I never opened my mouth  ;D
Logged
I've learned that pleasing everyone is impossible. But, pissing everyone off is a piece of cake!

IrishWonder

  • Inner Core
  • Hero Member
  • *
  • Posts: 561
    • View Profile
    • IrishWonder's SEO Consulting
Re: Who knows how to really work Google Trends?
« Reply #3 on: December 04, 2012, 07:42:32 PM »
The only thing I can think of is build a graph in Excel based on your cvs data and display larger scale data sets as is on the graph while displaying the smaller level data sets as multiplied by whatever, depending on their range compared to the larger sets. That would put them on the same screen/graph but then you still have to keep in mind that the smaller range numbers have been multiplied to make them appear there. Should work if you a trying to see seasonal trends or however they change over time but of course not for visually comparing the sheer volumes - but I suspect that is not what you're trying to do anyways?
Logged

BoL

  • Inner Core
  • Hero Member
  • *
  • Posts: 1209
    • View Profile
Re: Who knows how to really work Google Trends?
« Reply #4 on: December 04, 2012, 10:43:02 PM »
Try using a log scale, that would show the correlation without the large gap in scale.
Logged

Adam C

  • Inner Core
  • Hero Member
  • *
  • Posts: 623
    • View Profile
Re: Who knows how to really work Google Trends?
« Reply #5 on: December 05, 2012, 10:17:23 AM »
Graph them in Excel and put the outliers on a secondary axis.

http://office.microsoft.com/en-gb/excel-help/add-a-second-axis-HP005201690.aspx
Logged

dogboy

  • Guest
Re: Who knows how to really work Google Trends?
« Reply #6 on: December 05, 2012, 10:21:37 AM »
>trying to see seasonal trends or however they change over time
right. I'm trying to show a correlation between 2 variables, and then assuming that pattern exists, jump to proving causality.

>Try using a log scale
No way to do that in Trends' settings?
Logged

dogboy

  • Guest
Re: Who knows how to really work Google Trends?
« Reply #7 on: December 05, 2012, 10:22:45 AM »
ah, Adam, that sounds like actionable advice... much obliged.
Logged

Adam C

  • Inner Core
  • Hero Member
  • *
  • Posts: 623
    • View Profile
Re: Who knows how to really work Google Trends?
« Reply #8 on: December 05, 2012, 10:24:00 AM »
Actually, thats not great advice (mine above) on its own.

You'll be better off making separate data pulls.  Put the keywords into groups with similar volumes, and pull each group separately.  

Then plot them together.  You'll get a clear view of the trends but no (or at least less) clarity on relative volumes.

(Also pull all together so you understand the ratios between the large and small volumes.)

My suggestion above is better for relative volumes, but you may find the trend data lacks precision for the lower volume terms
Logged
Pages: [1]
« previous next »