Conversion rate optimization is a relatively new field with many new terms to understand.
Conversion Rate Optimization Glossary: |
|
| Conversion rate optimization | |
| Control group | |
| Campaign formats | |
Conversion Rate Optimization
Conversion rate optimization is art and science of creating an improvement in the percentage of site visitors who complete a key action. Conversion = visitors who complete an action / total visitors.
What would you do with an increase in conversion? The bottom line is that increased conversion gives marketers options, and these options lend themselves to both increased control and profitability for their business. Here are three strategies as to what you can do with a successful conversion rate optimization campaign:
- “Security Maximization” - An increase in conversion is used to reduce total spending while holding total acquisition constant. That is, without affecting acquisition, a marketer can choose to “bank” the cost of acquisition savings.
- “Profit Maximization” - An increase in conversion is used to drive additional acquisition while holding spending constant. In this case, the marketer chooses to “bank” the incremental total margin created by the additional sales.
- “Acquisition Maximization” - An increase in conversion is used to leverage the new lower cost of acquisition with the marketer choosing to spend all they can at the new lower point. As long as the lower cost of acquisition represents a self-funding proposition, the marketer can choose to spend all they can at the new reduced cost, thus driving two key actions: 1.) reducing their competitors ability to acquire customers, and 2.) increasing their total profitability. It’s important to keep in mind that when a marketer is on the other side of this strategy, it can appear as though the competitor is making poor business decisions, when the reality is that they’re successfully driving their business forward.
Control group
Similar to when a drug company gives a portion of the test subjects a sugar pill, they receive nothing new and are used to develop a baseline measurement against which an improvement or reduction in performance can be measured. It’s good testing form to provide for a control group. As well, it gives us a reference point to show how well the optimization campaign performed.Content Targeting and Website Personalization Campaign Formats:
Website Testing Methodology:
- Timing of tests:
- Simultaneous tests – two or more versions are tested against each other at the same time.
- Consecutive tests – one page is tested after another. This type is easiest to do but is fraught with a large problem: traffic varies over time so the results are clouded by the uncertainty of how the traffic variances from one period to the next effected the results of the test.
- Segmentation strategies:
- Random - Test participants are treated randomly and are not differentiated based upon any behavioral or other segmentation data. Optimization systems that segment visitors randomly work under the pretense that an “average” visitor exists and that a site can be tailored for that average. However, that is rarely true. Because of this, without segmentation, an optimization effort will necessarily produce a reduced effectiveness in the new design for many visitors at the expense of the success for others. The larger of the groups will typically out weigh the success or failure of smaller groups. Obviously, if a marketer were also able to effectively sell the visitors who were sub-optimized in test, their metrics would be positively effected. We see the affect of this “average” concept in changing site conversion rates. Changes that are largely due to a shifting mix in the visitor profile.
- Behaviorally segmented - To truly optimize a site, one would need to identify differences between visitors and trigger experiential changes for them. Once a visitor is identified within a behavioral segment, they are randomly split into testing channels. The homogeneity of the attitudinal mindset present in behaviorally segmented optimization tests allows marketers to get much closer to their customers and optimize much more successfully.
Example - Imagine that a Google AdGroup drive traffic for the Widget Company. If the marketer is using keywords that are more expansive than exact match, visitors who query for “cheap widgets” will be delivered to the exact same page and test as those searching for “quality widgets”. It’s not realistic to believe that these two visitors are coming to the site with similar expectations or goals. If one were to optimize randomly.- Online channel options:
- Multi-channel personalization - Through on personalization campaign, control the brand impression and experience across numerous visitor touch points. Online marketers who use this sophisticated approach commonly use: websites, email, 3rd party banner ads, and layers/popups.
- Website path - Individual pages within a longer process can lead to a conflicted visitor experience and reduced throughput if those pages are tested or personalized in separate campaigns. That is, if a marketer runs optimization campaigns on individual pages within a critical path on a web site (i.e. landing page to thank you page) the results of an individual page could change the success of pages that were previously tested and updated. Some optimization services can run multiple concurrent tests on the pages within the path (reporting not linked) but few if any can test the affect of sequential multivariate tests. In all, most optimization efforts require the successful completion of multiple pages and comprehensive path testing is a likely next step in web site optimization.
A|B tests
“black and white” comparison– Sometimes referred to as “split tests”, these tests attempt to determine the impact that one design has over another (others). This type of test is best suited for understanding how a very small number (2 usually) of dramatically different variables are perceived by visitors – i.e. test a landing page where you include a long text version to a short text & graphical version. These are easy to do and can provide strong initial results but are subject to the following problems:
- Rapidly diminishing returns – marketers can find a few large mistakes up front but struggle to gain the next step.
- Results come slower than multivariate tests – You can only effectively test a couple of variables and permutations at a time.
- Results are less conclusive than multivariate tests – marketers struggle to understand what portion of the changes between the alternatives created the changing results. "Did the new call to action or the new H1 headline create the improved results?"
Multivariate Tests:
Where A|B tests compare black and white differences, multivariate test look at the varying “shades of gray”. Typical tests are designed to look at a larger number of placeholders (~5) on a page and test variants of each. The potential number of permutations to be tested can grow very fast – for example, a test with 5 placeholders and 3 elements per placeholder will create 243 permutations, and a test with 5 placeholders and 5 elements per placeholder will create 2,187 permutations.The Taguchi Method is a variant methodology of multivariate tests. To allow for a large number of test permutations while minimizing the time necessary to complete a statistically sound test, mathematicians have created sophisticated models, such as Taguchi. Here, the model will study the relationship of placeholders and elements and suggest a limited number of tested permutations. That is, of the 2,187 permutations in the above example, the Taguchi method may select 40 specific permutations for one to test. From these 40 permutations, the model can predict the winning combination of placeholders and elements, even if that exact permutation was not actually tested. It draws statistical inference of the strength of the relationships exhibited by visitors and calculates the best case scenario.
