Get Free Quote
« Analytics RX:... | Blog home | SEO Monthly Update... »
March 22, 2011

Confidently Predicting ROI for SEO — SES New York 2011

Print Friendly

Another session, another solo presentation. Now that we’ve learned to set up our analytics correctly Chris Boggs (Director, SEO, Rosetta – on Twitter @boggles) will teach us all about determining ROI for SEO. You want that, right? Or maybe you just like throwing money around. (In which case, throw some my way.)

SES New York 2011

This room is much darker than the others. Must resist the need to nap. Don’t mock, in my estimation, I got up at 4:30 a.m.

So Chris was invited to speak at eMetrics on this topic four years ago but didn’t have many solutions to the questions he was asking. Now, he thinks he has a few answers.

There will be a prose version of this presentation on SearchEngineWatch, if for some reason you don’t like my coverage.

Introduction

SEO has three areas of focus – Technical, On-site optimzation, off-site promotion. They need to build on each other, starting with a solid technical base.

SES: Chris Boggs Presentation

[Briefly covers the JC Penney and Overstock news, I know you're savvy so I won't recap it.]

Links become more diverse as you diversify your content.

Obstacles to Measurement of ROI from SEO

Since SEO has made its way into marketing calendars, savvy marketers have been trying to show more than just “improved rankings”. Predicted ROI = (anticipated revenue from SEO efforts) – (Proposed cost of SEO project). Obviously both those elements are hard for SEO because there is no hard and fast rule, for determining that.

Are you measuring rankings or are you measuring performance? You have to measure performance.

Study: If sponsored listing are not labeled, what percentage of searchers click on Organic listings versus Paid Listing? Who doesn’t click at all?

4.2 million searches – excluding the people who didn’t click at all, 15.8 percent clicked on the sponsored listings. If you include the people who didn’t click anything (35%) then 10% clicked on the sponsored listings.

Which would you choose? (mock up SERP, not actual result)

Given three worthy choices, you may choose something based on your passion rather than on the order of the results. That makes clickthrough impossible to predict because you don’t know what the message that will resonate most is.  Statistically you can make a guess but in reality, each user has a unique story.

The description could influence people differently. If someone was interested in price points and there were prices in the description, that might cause a click. If you like a particular site, that may cause a click. If you want variety, that may cause a different click.

Universal and personalized search make the biggest difference in 2011. Slightly different keywords could return very different looking pages. Additions of News, Shopping, Images. The refinements offered in the navigation may be different. Even if you’re number one, you don’t know what the SERP will look like when the search is done.

Unknown Variables:

  • Total number of keyword phrases – It is very difficult to predict the number of unique phrases that a Web site can rank for
  • Number of searches for each keyword phrase – Predictive tools cannot provide highly confident estimates
  • Ranking for each keyword phrase – Rankings may take time to gain, and may be disrupted by Personalization and Universal Results
  • Average CTR for each top ten position – Very difficult to predict accurately across different stages of purchase cycle
  • Traffic for each keyword phrase – Based on fluctuating CTR
  • Assigned average value or arbitrary value for each visit (Conversion %, AOV)

All of these must be ignored to some degree when you’re projecting results from ROI.

Developing ROI model Using Past Data

How do you do predict traffic and conversion growth?

Version 1: They used a very basic model based on past traffic to the domain and using average monthly traffic as a baseline to measure against. They had a conservative projection and an aggressive projection. Their model was accurate but too conservative

Version 2: Updated the model to include average traffic over two years and existing variance (loss of traffic). Given conversion percentage and average order value, predicted growth in actual income. (Allows for including seasonality as well as having a baseline in the future.)

Use multipliers to adjust for the unknown. Limiters and parameters are based on tools and experience. That will tell you what you’ll be able to achieve by what time. Average month by month predictions, to account for seasonality is the best way.

Additional variables to refine into “aggressive” vs. “moderate” SEO:

  • Total number of links required per keyword phrase – estimations based on looking at top ranking pages for high priority keywords, and analyzing number of inbound links pointed to them
    • (Faulty because assumes “a link is a link, and more is better”)
  • Predicting increased or decreased conversion rate – based on site or platform upgrades, and cost analysis versus competitors, a number of tweaks could be made to this KPI
  • Assigning value of opportunity based on cost to drive traffic through Paid Search (PPC Listings) – very difficult to accurately project this since the goal is not for SEO to necessarily replace Paid Search (SpyFu has a good tool for this but paid and organic are not the same and the value is not the same)
  • Average CTR for each top ten position – Very difficult to predict accurately across different stages of purchase cycle

More complex doesn’t mean more accurate. It just makes your charts more confusing.

ROI Models Based on Competitive Analytics

If you don’t have history or volume (new sites, niche sites, sites without properly configures analytics, new domains) you can use multiple competitors’ sites to predict how much traffic they get and how much you will be able to get based on that.

Look at competitive intelligence tools as relative measuring tools rather than direct competitors. You want to have them graded all against the same baseline.

Keep in mind that you may need to use multiple sites, since competition can vary from keyword to keyword.

Simple Formula:

Identify non-brand keywords that competitor sites are ranking for and define a limited set.

Apply a CTR for each keyword based on visibility (#1 = 40%, #2 = 20%, etc.)

Multiple that by the search volume.
SES: Chris Boggs Presentation

The Future of SEO ROI Projections

SES: Chris Boggs Presentation

This is a self correcting model in a tool Rosetta has built. This will help you find one-time events that can be filtered out again to continue adjusting. This tool is still in Beta.

Wrapping up:
You’ll never be 100 percent confident in your ROI predictions. There’s a lot of factors involved.

  • Actual implementation rates of SEO recommendations
  • Actual number of inbound links gained (Social Media participation level)
  • Level of competition, and their SEO sophistication
  • Ongoing algorithmic updates – Farmer/Panda changed a lot of people’s predictions
  • Ability to update projections
  • Leverage successes to gain more SEO budget and further improve performance

Q&A

[skipping this. The deck will be available online to attendees and his article will be out soon on SEW.com]





Comments are closed.



Learn SEO
Content Marketing Book
Free Executives Guide To SEO
By continuing to use the site, you agree to the use of cookies. AcceptDo Not Accept
css.php