It’s fairly widely recognised now that the traffic figures quoted by Google in their Adwords tool are pretty much bunk. This is a bit of an issue for exponents of SEM and SEO both, as we all need to have some idea of traffic levels in order to guide the direction of a campaign, choose the best keywords, and so forth.
Fortunately, this isn’t the only information that Google releases. Their Google Insights and Trends tools both information about relative search trends that you can download into your spreadsheet programme of choice (ahem Excel ahem) and manipulate as you wish. Here comes the caveat emptor – be careful with this raw data. The numbers quoted are normalised to fit a scale, and as such, very depend on exactly what readout you’ve specified, what keywords you’ve entered, the timescale and so forth.
Even when you’ve narrowed a range that you want, you’re left with a choice of ‘fixed’ and ‘relative’ scaling for your data dump. The basic difference between this is that ‘relative’ scaling normalises the data to an average based on the timescale, whereas ‘fixed’ scaling measures everything against the average of a fixed period of time, normally 2004, assuming that data exists for that time, the moon is in the right part of the sky, and so on and so forth. Of course, as you might imagine, not all ‘fixed scalings’ are created equal; the normalisation factor will shift, depending on which particular keywords you have used as your input. The upshot of this is that you can get different values for the same search term.
All this being the case, it’s clearly quite a complex set of numbers to handle. But are not difficulties just opportunities in sheep’s clothing? I may be getting that confused with something else. Anyway, here are some pointers to guide you through this mathematical minefield and help you to produce some results that are in some way meaningful.
1. Relativism is everything. In a world with moving goalposts, you need to try to establish a fixed point. Chose one keyword that you know has pretty good traffic (I normally chose the one with the most volume I can find.
2. Use fixed scaling. Always do this. Always.
3. Include the chosen search term in all of the searches you perform. I generally use this phrase and four others in all of the searches I do.
4. Chose a point in time, and stick with it. Except when you need to compare seasons. Then don’t.
5. Create a normalisation factor by scaling all your data to your fixed point. Remember that it has to be the original fixed point, so you might have a few factors to include in your equation.
This should help you along the way. It’s worth bearing in mind that even this is a pretty simplistic look at things – it doesn’t take into account the different ways that Google displays search returns, for example.
Anyone have any thoughts on improving search volume estimates?
Related posts:
