It’s a testament to SEO becoming mainstream that more and more marketing folks are asking, and often demanding, that the results of an SEO effort be projected beforehand similar to how other marketing activities are handled. And it’s a testament to the immaturity of the SEO space that such projections often are overly simple, full of caveats, and off by margins so large that guessing wouldn’t have been any less accurate.
Projections remain difficult because of some basic and mostly insurmountable obstacles. First and foremost is that no one really knows all of the intricacies of the search engine ranking algorithms. Second, those algorithms continue to change regularly, especially at Google, such that even if reverse engineering were used to establish hard and fast guidelines today, those same rules could be wrong tomorrow. And the third big reason is that it’s impossible to know if competitors – not just brands like yours, but any site that is competing for visibility in the organic search engine results – are about to kick off their own expanded SEO initiative, which would make your job even harder.
So how’s an SEO strategist supposed to respond to requests for a projection model? I’d start with the simple model that many use and then add some refinements.
Simple Projection Model
Start by identifying keywords of interest, extracting search volumes from somewhere like Google’s Keyword Tool, totaling the volumes, and then multiplying by some percentage that represents expected clicks for the target position in the SERPs, e.g. 40 percent CTR for a number one ranking.
Conceptually this model is fine, but in practice it tends to be inaccurate because of the assumption that you will actually achieve the desired ranking.
Refining the Projection Model
You’ve probably got one or more competitors that are similar enough to you that their current rankings represent reasonable targets. Or to put it another way, since a competitor in the same space with similar resources can rank number two for a keyword, it’s reasonable to assume that with the right SEO strategy you could claim that number two spot.
This sort of comparison eliminates the impact from not being able to take advantage of tactics that could tarnish your brand (your competitor also wants to protect their brand) as well as the constraints that regulatory restrictions impose on your industry (your competitor has to deal with the same ones). This approach also means you’ll automatically remove implied expectations that you’ll somehow outperform organic competitors like Wikipedia.
Adding Upper and Lower Bounds
A range is expected with projections and it’s also good to establish one tracking purposes. If you’re going to use search volume data from Google, one way to determine a range is to use exact match for the lower bound and phrase match for the upper bound of potential clicks to your site. Using exact match lets you look at the potential benefit of targeting specific keywords via things like specific title updates and anchor text use. Using phrase match allows you to account for the side-benefits of SEO, which typically results in traffic for keywords you didn’t explicitly target. I would, however, stay away from broad match because the data often double or triple counts keywords, which effectively guarantees your model will be wrong.
With a refined and more realistic model in hand, you may be tempted to stop the analysis, but that’s not ideal. Instead, as you proceed with your SEO effort and you plot actual performance against projected performance, you have the opportunity to further refine your model by revisiting and updating your assumptions. For example, are you getting the click-through rate you expected from the SERPs? If not, adjust the model. Was your model skewed by a small number of high volume keywords for which you’re not making the progress expected? Then remove them from the list or split the calculations into short-term vs. long-term projections.
With the current data available, projecting traffic from a planned SEO program is going to be just an approximation, but that doesn’t mean the effort isn’t worth doing. You’ll end up with a keyword list you can continue to reference; a list of assumptions and risks you can continue to monitor; and a potentially useful way to summarize your efforts to upper management. And if the model happens to be accurate, even better!