By Bryan Olshock and Justin Gabbert of Red Door Interactive
There has never been more data available to digital marketers. With this abundance of information, however, comes the significant challenge of understanding and leveraging the data.
The goal of any SEO campaign is to look at a project and quantify its impact. Was the project a success? What was the lifetime value of a project? What types of projects work best for each objective? Once these facts are known, it is possible to gain a better understanding of how and why certain projects were successful and then predict the impact of future projects and initiatives. More specifically, if the expected return on a project is known, it is possible to prioritize budget and timing accordingly.
Most marketing experts measure unique users, page views, time on site, bounce rate and search rankings to determine if a campaign is working. These metrics don't determine, however, why it worked. Fortunately, there are lesser-used statistical analysis processes that can be leveraged to make these determinations and get to the root causes of poor performance.
1. Simple Linear Regression
A simple linear regression is a basic input versus output equation. In relation to SEO terms, analyze organic search trends over a 12-month period and adjust for seasonality. To do this, first find the seasonal index, a breakdown that shows the relationship of one month to the entire year.
Once the index is complete, "de-seasonalize" the data by dividing each month by its seasonal index, run a simple linear regression to find a true forecast and then re-apply the seasonal index to each corresponding month. Once this simple linear regression is complete, it is then possible to create yearly forecasts for organic search traffic, set accurate SEO program goals and generate monthly expectations for a project that takes seasonality into account.
2. Multiple Linear Regressions
A lot more variables exist that can impact organic search traffic than just seasonal trends, and that's where the use of multiple linear regression becomes effective. This is an approach for modeling the relationship between a dependent variable ("Y") and multiple explanatory variables (denoted "X"). From an SEO standpoint, a multiple linear regression takes external factors such as paid search, display advertising, market fluctuations and seasonality into consideration. With these factors included in the equation, it is possible to determine how various inputs impact search traffic, accurately predict future organic search traffic and provide a historical indication of where SEO may have led to incremental traffic.
Grow Your SEO
See what gardening and SEO have in common in this infographic at wsm.co/gardenseo
3. Impact Analysis Model
Multiple linear regression relies heavily on historical organic search traffic, but the impact analysis model (another form of analysis) allows us to establish a baseline, incorporate multiple inputs and uncover the incremental value of SEO projects. A baseline for organic search traffic is essentially what the numbers would be without the use of SEO. That being said, traffic should fluctuate with industry search trends. In order to develop a baseline, first determine the search volume trends within the industry or vertical. Once that is complete, determine a year-over-year growth/decline rate for each month within that sector and then apply (monthly) the growth/decline rate to the previous year's organic search visits and get this year's baseline. Then, using a multiple linear regression, determine the impact of media spend contributions including online media click and impression contributions. This process takes into account both online and offline advertisements.
Incremental visits are then determined using the following equation:
Organic visits - (Organic Baseline + Online Media Click Contribution + Online Media Impression Contribution) = Incremental Visits.
With these factors considered, digital marketers can determine how many incremental visits and conversions the SEO project drove over the course of a year, how the various SEO projects impacted traffic and what the impact of future projects will be. Using linear regressions to forecast the future, multiple linear regressions to understand the external factors and the impact analysis model to determine expected return on a project, it becomes possible to prioritize budget and timing moving forward.
Applying this level of advanced statistics to SEO projects can seem to be a relatively novel concept, but its effectiveness demands a closer look.
