If there’s one thing that’s universally accepted, it’s that the results of SEO are unpredictable.
However, as more investment is made into the channel, we’re being put under increasing pressure to be performance-driven (and show the value we can drive “up front”) in the same way, our paid counterparts have been for some time.
Invariably, this means setting targets and being held accountable for whether we hit them or not. But this shouldn’t be an issue and in fact, I believe it’s completely possible, with the right caveats and education, for SEO to become a results-led channel. And ultimately, if we’re going to continue to grow in popularity (and see further investment), we need to be.
Measure the right things
For a long time, SEOs have talked about “improving keyword rankings”, with the coveted number one spot on a high traffic term, being the primary focus for many marketers. In 2019, however, this approach is myopic. I’m not ruling out rankings entirely, they still have value, but promoting the idea that a small, well-backed site, can outperform a titan like Amazon, Compare The Market, or RightMove for a competitive term, is unrealistic. Moreover, the resources required to do this could be better spent driving change elsewhere and this is why it’s important to measure the right things.
For forecasts, I typically look at three metrics: traffic, conversions, and revenue, with the former holding the most importance.
My reasoning – While you may receive a significant amount of new traffic from a high-value term, but (more often than not), the traffic will not be as qualified, lowering the “true value”. Used as a vehicle to drive wider site change, SEO. However, it should build results across your businesses search landscape as a whole and this can only really be measured through increased traffic (and from this, we can work out conversion and revenue increases).
But traffic isn’t a “clean” metric either and can be influenced through changes in the market, wider media activity, competitors, or even the weather. For this reason, many look to derive a “growth factor” or absolute increase figure from keyword gains, basing the output of expected traffic from changes in position (and the clicks received from this). This is an approach we’ll discuss shortly and certainly can give an indication of any potential performance gains, but its reliance on keyword data still provides a few flaws.
Put this into context
A key point to make with any forecast is that it is a projection, based on factors that we believe we can predict with varying degrees of certainty. Apple cannot provide an exact figure to shareholders, for the total number of new iPhones they will sell; nor can Disney foresee the specific number of tickets the next Star Wars film will sell. This context is important, given the unpredictable nature of search marketing.
I think we often lose ourselves in trying too hard to be accurate, that we miss the point of a forecast altogether, it’s an indication, not a definitive statement of growth that could be achieved, linked to the investment that is made. So it’s perfectly fine to caveat as such and represent any figures as non-binding. I personally use a “percentage certainty” score depending on my experience of rolling out specific activities (and their impact), which is often well received.
Importantly and as a key takeaway for this section, you need to put your data into perspective.
If you’re required to project the impact of a minor change perhaps to help and IT prioritize their development queue it may be more realistic to provide a percentage range instead of a specific number. So long as the methodology and your explanation of how this could come to be are logical and data-driven. Ultimately, a forecast is only as reliable as the data and the experiences of the person making it.
Projecting growth (or not)
In terms of working out the specific figures, each SEO has their own methodology, but broadly speaking, these can be grouped into three:
1. “Experience-based gut feel”
This kind of forecast draws on the expertise of the individual and can be a useful, “quick and simple” guideline for making snap decisions. The reliability of the projection is intrinsically linked to the knowledge of the person making it and shouldn’t be utilized in any official capacity or to make business-critical decisions. But this is not to say that this type of prediction isn’t valuable, it just depends on the use case and context the forecast is being made under.
2. “Ranking-focused click curves”
This uses ranking data to project an increase in keyword position over time (and the value from this). In my experience, many “off the shelf” forecasting tools use this method, which relies on modeling traffic based on click-through rates from ranking positions. More advanced studies might include competition or seasonality factors (perhaps from Google’s Keyword Planner) and may provide the output as either a total growth figure or shown by month using a compound growth rate or more complex increase curve.
3. “Traffic-based growth modeling”
This is, in my view, the most accurate way to forecast and frequently uses historic traffic data to predict the impact of not carrying out SEO activity, taking the market impact, seasonal changes, and other factors into account, and estimating projected increases from this. It’s often incredibly difficult to perfect and typically requires the help of a data scientist. But, if executed correctly, this methodology can provide a good representation of the impact your proposed activity can make.
A side note: In many scenarios, you may expect to see an increase in traffic predicted over your time period. However, SEO takes time and doesn’t often conform to the pre-set timelines of a campaign or contract. As such, it may be that a true forecast may not show immediate growth or even (if your site is following a downward trajectory) no increase at all.
The key takeaway here is that without the activity, you may be in a much worse place that you are predicted, and no growth doesn’t necessarily mean no value.
However, at their heart, each of the above methods still requires an expert to assess the types of changes you are intending to make and their impact. This must be someone with enough experience of the channel to understand how the different projects will move the needle, in the context of your site’s situation.
There’s no “official view” of how search activities will impact your performance and opinions may differ between experts. Moreover, the value of each work-stream will likely change by industry (or even by the query) and this is an important context to keep in mind when discussing targets and key performance indicators.
So far, we’ve discussed context, measurement, and methodology.
As such, in this whistle-stop tour of forecasting, it would be a missing piece not to touch on evaluating performance and targets. As I mentioned in the last section, SEO takes time and we’ve already established that it’s highly unpredictable. So, holding it to the same degree of accountability as other channels may be unhelpful. Importantly, performance needs to be a two-way and open conversation between your expert and your business (as it’s impacted by many different factors); in my experience, continuing with this approach throughout the completion of a strategy and into the evaluation process is the best way ensure you’re driving the most value from the channel.
There’s no official blueprint or road-map to SEO success and it’s highly likely you’ll try things that do not work and others that exceed expectations. Not every fluctuation in traffic or missed target (or over-achievement) for that matter is due to performance and your expert should be able to explain why this is the case. Deviations do not always mean the model is off either; remember, in an SEO forecast, you’re ultimately asking your expert to predict not only the direction of a constantly evolving algorithm but changes to market demand and the world you operate within. It’s not a coincidence that many of us become futurists!
To sum up, predicting SEO results can be very hard but is a challenge we must seek to solve if we are to take the channel to the next level within digital marketing teams. A prediction may use a data-driven methodology, but is still essentially the view of a specialist, based on their experience of the search landscape. Not all forecasts will show growth and not all targets will be hit, this shouldn’t be seen as an issue as long as the reasoning behind this can be explained.
It’s been nearly 3½ years since Google announced RankBrain, but we have little detail about how it works. Here’s what we actually know–and what is myth.
It’s tempting to go after broad, high-volume keywords with huge audiences (and equally huge competition). But there’s an equally large — albeit more distributed audience — to be reached by targeting lower-competition, long tail keywords.
Scaling a brand one link at a time can consume a great deal of time. Using PR can change this and grow your audience and links much more quickly.
The data-driven attribution model is exclusive to customers of Google Analytics 360 and offers an easy way to personalize experiences to your customers.