How to Research and Test for Conversion Optimization

In the previous post, I discussed what conversion optimization is, why it is so important for growth marketing professionals, and how to begin with conducting the right research and tests that provide significant measurable results.

To continue from the previous post, I will dive deeper into how a growth marketer or conversion optimization expert should begin the entire process of optimizing a website or any brand’s digital assets for optimization.

Conversion Research for Brand New Business

Just as much as the success of a growth marketing framework relies on experimentation, so does it rely heavily on the amount of traffic (no of users participating in the experimentation framework).

Developing a research and conversion optimization strategy for an already established website with a considerable amount of monthly traffic can be somewhat easy unlike optimizing a brand new business for conversion.

A brand new business often has no significant amount of traffic to test conversion strategies with and this makes it hard for conversion optimization.

How do you optimize a brand new website with zero to a little amount of traffic for conversion?

If you are a new business with little to no customer data, the best advice is usually to forget conversion optimization until you have sufficient customer data that you can begin testing, analyzing, and optimizing for.

Thus, as a brand new business, what you need and should focus on is figuring out what product to build and for whom. The need at this stage is to attain a product-market fit by implementing the treatments you have hypothetically generated for the problems you presumed or identified based on assumptions.

Common Conversion Optimization Research & Testing Mistakes

In order to get the optimal results from your growth experimentation, below are some mistakes to be aware of and avoid.

1. Running tests that make no sense, that is, testing insignificant factors or elements that have no important impact or benefit on your business.

2. Thinking you know what works without testing to validate your thoughts or assumptions. We often want to assume that we know where a problem is coming from or what solution should be provided when conducting some analysis simply because we have done turns of tasks that may be similar to the present task.

However, the truth is that you may not know what we work or not until you test because users’ behavior isn’t static, it changes erratically with time and thus cannot be relied upon.

3. Copying other people’s test results. A lot of marketers do this often. They see high-performing campaigns or designs then the next thing they do is get into the brand’s funnel and begin to copy her approaches.

As an external party, there are limitations to the information you have access to regardless of how comprehensive you feel the result of your research on your competitor is.

You only have access to little information about what makes the campaign you studied, successful. You don’t know if there are different premises that must have influenced the campaign that you don’t have access to.

4. Testing with a low sample size usually doesn’t provide reliable results. If you don’t have at least about a thousand monthly traffic visits to your website then don’t test.

5. Not running the tests for enough periods of time can also bring erroneous results. Ideally, you should test on a 7 days cycle without any breaking pattern. And the test must be done for a minimum period of 14 days to 28 days.

For instance, if you begin a test schedule for 14 days on a Monday, then the test shouldn’t end until the Sunday of the following week and neither should the test be paused and later resumed during any of the days. The test should and always keep running till the end of the scheduled period.

However, if you have thousands of data coming in then you can reduce the testing duration.

6. Sample pollution is also another common mistake made by conversion optimizers.

For instance, having someone enter your website from different devices instead of a single device type consistently and performing the same actions on the website can adulterate the test results.

7. People often delete their cookies

8. Giving up after failing at the first test

9. Ignoring small wins

How to Measure the Effectiveness of a Testing Program

In order to avoid wasting time by testing the same testing mistakes, that is repeating a test with existing elements of error. It is important to continually self-assess the efficacy of your testing programs such as the following:

Testing velocity: Though this depends on your amount of traffic.

Percentage of wins: One of the reasons that people don’t record enough win tests is because they are testing the wrong problems, to begin with

Impact per test successful: The degree of the influence or benefit recorded from a test is as well a good factor.

How Long Should a Test Run

Is it when you reach a 100% conversion per variation? Apparently, no.

Should it be when you reach a 95% statistical significance? No.

Is it whenever the testing tool tells you? This also is a big no.

Statistical significance is an algebraic system that tells you nothing at all until two previous conditions have been met.

Thus, what are the criteria to determine when a test or experiment is complete?

Criteria to Declare a Test Done

1. Sample size: Does it have enough sample size?

Has it gathered enough data?

Do you have enough data?

Do you have enough people going through the experiments?

2. Multiple business cycles: Usually conversion rate fluctuates from day to day because people’s behavior changes from day to day as well.

For instance, surprisingly, how people behave on a Monday differs from how they behave on a Friday or over the weekend and this reflects in their consumption abilities.

3. Statistical significance reached: The statistical significance will help you determine if your test results are due to some chances or to some set of factor interests.

Conclusion

There are plenty of people (some who call themselves optimizers) who claim to know immediately what’s wrong with a website, and how to improve it.

Sure, some problems might be plainly obvious. But you might be wrong — your personal preferences and bias get in the way — and the better the website, the less obvious the problems are. So, you are left with opinions and the problem with opinion is that they often don’t count

As a growth marketer, you should know that CRO is 80% research and 20% experimentation. Thus, you need to master how to conduct the right research to get sustainable results over time.