Stormy Weather The Accuracy of Search Volume Estimation and Forecasting Graeme Benstead-Hume @mrgraeme
Stormy Weather
The Accuracy of Search Volume Estimation and Forecasting
Graeme Benstead-Hume@mrgraeme
Introduction
We wanted to compare the actual traffic of a number of keywords against the figures provided by the keyword traffic volume estimation tools we use.
• Gauge accuracy• Estimate correlation• Find ways to produce more
accurate results
How we might use estimations
We might use estimates to:
• Choose the best performing keywords for a campaign
• To estimate traffic volume at various rankings
• To estimate ROI for these terms
How we have used results in the past
When we have needed more accurate results:
• Segmenting key-terms by type and traffic volume
• Create coefficient based on know traffic volumes
The StudyTook 50 keywords ranking at position 1 on Google UK
Took average monthly traffic of terms over 12 months & updated based on CTR
Gathered monthly search volume estimation data from:• Google Keyword Tool• Wordtracker• SEOBook
Actual results vs. Google Keyword Tool data
Actual results vs. Google Keyword Tool statistics
• Avg. % Accuracy 13403%• St dev. 45623%• Mean accuracy 1066%• Accuracy of total traffic 982%• Correlation of datasets 74%
Actual results vs. Wordtracker data
Actual results vs. Wordtracker statistics
• Avg. % accuracy 1099%• St dev. 2459%• Mean accuracy 156%• Accuracy of total traffic 673%• Correlation of datasets 85%
Actual Results vs. SEOBook data
Actual Results vs. SEOBook statistics
• Avg. % Accuracy 4232%• St dev. 18710%• Mean accuracy 44%• Accuracy of total traffic 313.31%• Correlation 80%
Comparing the toolsAverage % accuracy
1.Wordtracker – 1099%
2.SEOBook - 4232%
3.Google – 13403%
Correlation to actual data
1.Wordtracker – 85%
2.SEOBook – 80%
3.Google – 74%
Correlations between the toolsCorrelation between tools
Google – Wordtracker 0.73
Word Tracker – SEOBook 0.97
SEOBook - Google 0.63
How to fiddle with the results
We were hoping for more consistent results…
Especially if we were going to be presenting our finding in front of 600 clever SEO types…
So we got to fiddling…
Playing with coefficientsGoogle * 0.25
• Avg. % Accuracy 1660%• St dev. 5706%• Mean accuracy 77%• Accuracy of total traffic 119%
Getting rid of the data we didn’t like
Estimations for keywords below ~50 search volume a month seem least accurate
I.e. Our estimation tools do not perform so well with long-tail terms
Disregarding low traffic volume keywords
• Avg. % Accuracy 173%
• St dev. 379%
• Mean accuracy 44%
• Accuracy of total traffic 125%
Did we learn anything?
• Alone these estimations are far from reliable
• Estimations for low volume keywords are especially unreliable
• To get the best results we must use existing real life data
Limitations of study
• Keywords came from a variety of industries
• Not all keywords had a years’ worth of data to be averaged
• All tools sample slightly different data for estimations
Thank you!
@mrgraeme
Presentation available at sitevisibility.co.uk/blog