Safe Sex with a Keyword Tool

Safe Sex with a Keyword Tool

The Problem with JavaScript Logging Systems

Nothing beats "exact actuals" like your own web server logs. (In my experience Google Analytics is always 25% low on actuals and Omniture is always 40-60% low... nothing beats having your own log files and a dependable way to determine actual visitor sessions and filter out bots from the data) I'm not saying GA and Omniture are worthless. Omniture might be. But GA has saved my butt many times by being able to fill in gaps when my log systems failed... My websites get more traffic than Oprah according to ComScore. We log several million points of data per minute. Sometimes we break things as all developers do. Google's consistency of always being 25% below actuals has let us backfill approximations in our timeline. Of course I flag those regions as backfilled estimations from GA.

There are numerous reasons these logging javascripts fail seemingly randomly but consistently over time. In the case of google it is intermittent network latency between google and the client browser and older slower computers that render pages slowly. These bottle necks allow the possibility for a visitor to click a link, bookmark, do a different search before the logging javascript downloads and executes.

The reason Omniture's data falls so short (from my own personal experience) is that their system frequently executes on page javascript before their JS library loads in asynchronous browsers (so basically modern browsers)... They need to rework the script to queue logging events and redirects until the library loads then process the queue. Instead they call functions from a script that half the time hasn't loaded on slower computers and you get those ever so helpful Undefined Object errors... You can click links and see this in the browser's javascript console, just find a site using Omniture and try it on a slow computer. it will look like this in the error console of your browser by page load... error.. no error... no error ... err... err no err...err ... err... no err... no err... no err... err

In short... nothing beats actuals from your own logs. 3rd party JS tracking systems always fall short of actuals unless something breaks on your website.


Why actuals matter with keyword tools


The safest way to use 3rd party keyword tools is as relative performance indicators. If Keyword A is 4X bigger than Keyword B in search volume then tune for Keyword A if you want to be in the results for more searches or tune for Keyword B if you want an easier time getting to the top. Hypothetical ROIs are a big waste of time and are nearly always wrong. Costing is only accurate from hindsight and experience so you will always produce better financials with "actuals" and "after the fact" until you have 3 or 4 years of history to start producing realistic goals and models. Pretty much any other approach is wasted time and bad data. Both of which leads to wasted money.

Note: The best question to ask yourself with analytics tool data is "What beneficial action can I take right now with this data?" If data or "marketing advice" in general isn't actionable in the short term then its not where your time is best spent at the moment. Complete the other business milestones on the "to do" list and earn your fist dollar and make marketing actionable.

For ballpark estimates based on actuals that target the same market and have the same seasonality you can use your actuals from one niche in ratios for all other similar niches. Here is an example ratio estimation using keyword tool data as relative performance indicators vs. actual traffic.

X = ( A / B ) * C

X = Estimated Traffic to the Untested Niche

A = Keyword Tool Search Volume For Untested Niche
B = Keyword Tool Search Volume For Tested Niche you already have
C = Actual Traffic Volume for the tested niche you already have

Using actuals and treating 3rd party numbers as relative indicators is how you safely have sex with a keyword tool. That and condoms.

Another reason script-based tools don't show true stats is that some people have script support turned off in their browsers or use plugins like NoScript that do that for them. That's a small minority, but nevertheless....

Question is what tool(s) do people use if they choose the log-based approach? Is awstats the defacto choice?

And what kinds of statistics and graphs and tracking do those tools support? Last I looked at something like awstats, it was quite primitive.... But maybe it or another tool like it has caught up to the likes of GA and Piwik?

(btw ipopbb, all your posts have made for excellent reading the last few days!!!!)