Tips for attributing specific rises in rank to increases in traffic

[Oren.]

I'm curious how people attribute a rise or fall in a specific keyword to increases/decreased in traffic on a large website that ranks for a wide variety of keywords. I'm specifically interested tracking how small moves on the front page, especially the top positions and in competitive keywords affect (or don't affect) traffic/CTR.  I'd like to be able to attribute specific changes in rank to their impact on traffic.  I'm thinking about this more for internal testing and optimization rather than for reporting as a way to try to hone resources on what has the largest impact.

Things I'd like to know more about:

• How to measure the difference in traffic/CTR for a term that moves from 2nd to 3rd position in SERPs.  3rd to 1st.  4th to 5th.  1st to 2nd. etc. etc.
• The same fluctuations, but in a local listing.
• The effect of moving above or below a local listing, or a listing with some other kind of extended result (sitelinks, news article, etc.)
• How to measure the impact of an appearance or disappearance of authorship on click through rate, especially for positions 2-5 in the SERPs.  The same would go for a video snippet.
• Measuring changes resulting from any other type of change

I'd love for as many people as possible to answer, with as specific or broad of an answer suits your experience and expertise.  Feel free to answer only some of the questions, or provide related information.  Links to articles, anecdotes, case studies, hunches, and hard numbers are all appreciated and welcomed. I look forward to hearing from people!

[EGOL]

Go into webmaster tools, click on your site, click on "search traffic", click on "search queries", click on a keyword.... there you should be able to see position in search results, impressions, clicks and CTR for a period of time.  You can download that table for future comparisons.

I have great curiosity about the same questions as you.  However, the amount of time that could be spent on this is enormous.  Also, this data is distorted because rankings for you might be different than rankings for me, and all of your pages could be pulling traffic for a wide range of keywords.

So, for millions of visitors per month I am not watching any of this.  Instead, my measure of content value is simply the number of visitors who enter the site through a specific content page.  And, for a few of those I have adsense channels or trails to retail confirmation pages to get some dollar value of production.

I rarely look at these once that it is confirmed that conversions are happening - maybe a couple times per year to see where the money is coming from and making plans for new content development.   Otherwise, I am off onto making more content.

[Travis_Bailey]

My best input here is; "Don't do that."

Instead, worry about real things. Drowning yourself or your client in data doesn't help. What does help is calls, emails and sales. Maybe you can think of a few more things that keep the lights on, and how your campaign contributed to that.

bukowski_1-Custom.jpg

[randfish]

I have to agree with the general sentiment of both Travis & EGOL. While it is intellectually interesting to discover the answers to questions like the ones you've asked, it serves almost no practical purpose. I've thought about it often, wondering if the ability to get better about predicting how movement in SERPs could affect traffic and conversions, but the real answer is that the variability is too high to make predictions accurately even with a lot of data (you'd need data across literally millions of SERPs to have a true model, and even then, it would likely require machine learning + inputs we don't have access to).

The simple models are "good enough" - i.e. rankings go up, traffic related to those keywords will, too. Keywords that drive particularly valuable traffic that rank poorly and have less competitive SERPs should be the focus of marketing activities

Someday, I hope to build software that can get better at this (and we're working on some of the more simplistic models with ranges today at Moz), but it's likely not a great use of time to try and get insanely detailed on this.

[Oren.]

Alright, thanks for the input guys.  I guess it's good to know that everyone is shooting into the dark, at least a bit.