Landing page optimization: Using data, hypotheses & tests for more conversions

Your landing page is there, but the desired success is missing?

Then it is probably time that you take care of the landing page optimization.
But to optimize your landing pages, you first need to measure their performance. This requires defining goals and goal metrics. You can then test changes, for example using A / B tests or multivariate tests.

But to optimize your landing pages, you first need to measure their performance. This requires defining goals and goal metrics. You can then test changes, for example using A / B tests or multivariate tests.

How can I analyze the success of my landing page?

To evaluate the effectiveness of your landing page, you must first have defined a specific goal. Ideally, this happens before you create a landing page, so I assume that this is the case.

In the case of "Lead Generation Landing Pages" (or "Mail Opt-in Landing Pages") you want to measure, for example, the number and the "qualifications" of the people who filled out the form on your landing page. And usually in relation to the total number of visitors to the page.

In the case of "Click Through Landing Pages", on the other hand, you primarily want to measure the click rate on a linked page (for example your online shop).

These values ​​represent the starting point for the landing page optimization - with a view to conversions or Google rankings. From here it is important to find out how you can change your page in order to influence this value positively.

Good to know: We speak of "uplift" when we increase the conversion rate and in the best case generate bottom line growth, for example in the sense of more sales.

Landing page analysis the easy way.

For a first evaluation you can use the Unbounce Landing Page Analyzer . The tool focuses on the optimization potential of your landing page and measures nine different categories:

1. Industry performance (corresponds to a kind of industry benchmark)
2. Page Performance (refers to the page loading speed)
3. Copy analysis (especially headlines)
4. Mobile responsiveness
5. Conversion centered design
6. SEO (with focus on keywords)
7. Trust & Security
8. Message match
9. Social open graph tags

The analyzer can test any website, but it lacks advanced functionality such as natural language processing. In my eyes it is more of a marketing tool than a real optimization tool. However, it recognizes how certain words can affect the conversion rate and serves as a good point of reference for the further analysis of your landing page. Because a look at your Open Graph, the mobile version of your landing page as well as Google rankings and corresponding SEO factors, if relevant, are certainly useful.

The 7-level model for landing page analysis

"With the help of the 7-level model, weaknesses of a website can be identified from the user's perspective" - this is how André Morys describes his model. The highlight is that it is based on customer thinking and decision-making processes instead of linear funnels, as are often artificially constructed by marketers. It is based on the assumption that a visitor has to reach different "levels" (in the correct order) before finally deciding on a purchase and completing it mentally satisfied (ergo no purchase loyalty develops).

In essence, this means that a customer must first recognize the relevance of your brand, your products or services and, in particular, what you offer on your landing page before he takes a closer look. They need to be able to trust and be able to orient themselves easily so that they can take the next steps themselves - such as the selection of suitable products. From then on, it is up to us to provide sufficient stimuli to really put these products in the shopping cart or to carry out an alternative "micro conversion".

We achieve this, for example, through shortages or urgency. The closer it is towards conversion, ergo purchase completion, the more security the user needs, for example in the form of social proof or guarantees.

The levels of comfort and evaluation then put the crown on the user experience: the more fun the purchase and above all the receipt (keyword: unboxing) and the use of what is bought, the better you will be remembered by the customer and the higher the probability that he will buy from you again in the future.

Retention marketing should not be underestimated and is far too often neglected!

The greatest potential lies in the combination of qualitative AND quantitative methods.

While Unbounce 'Landing Page Analyzer is like an expert evaluation, the 7-level model is a mixture of an expert evaluation and a heuristic evaluation (ideally based on personas). Both are qualitative methods, similar to user tests, for example in the context of Usability labs or user surveys.

If there is a lot of money involved that you can generally gain by optimizing your landing page or website, I recommend a combination with quantitative, i.e. data-based methods such as web analysis or 5 second tests . This gives you a complete picture of the strengths and weaknesses of your site and allows you to formulate more specific hypotheses for conversion optimization.

What are typical measures for landing page optimization?

This question is often asked and unfortunately also often answered without the indication that these measures are in no way a general recommendation for every landing page. Rather, it is tips and "typical mistakes" that are addressed. The crucial question in individual cases, however, is how to fix them and what you can deduce from them to optimize your own site.

Please consider the following list as a source of inspiration and not as a checklist of landing page optimization!

1. Show (your) personality

The sympathy factor often counts more than a certificate: we decide in a few seconds whether we want to work with a person or not. Regardless of whether you sell products or services under your own name or act as a brand: Showing a face and expressing your personality never hurts.

Because customers are increasingly buying the "why" behind brands and products and not because of individual features. But landing pages in particular often appear sterile and unemotional - because they are often created on a large scale using templates for certain marketing activities. But I tell you, it's worth showing your personality and optimizing the emotional communication of your website - especially if

2. Answer the most important customer questions

Companies that blog or maintain an FAQ section may have already answered the most important questions of their customers there. If not, the landing page is a good place for it - whether in the form of links to the same blog articles or as directly integrated questions and (short) answers.

Why? Because this can resolve concerns among visitors. This is exactly why we often see information in e-commerce about return conditions or delivery charges; because the typical customer asks these questions when making a purchase.

3. Let your customers speak for you.

Everything you say on your landing page (or anywhere else on your website) can basically be interpreted as advertising. So it is worth considering letting others speak for you in order to generate trust . And who is better suited than your satisfied customers?

There are no limits to your creativity when it comes to design, but as always, test the different options to identify those that lead to more conversions. Does the combination of a portrait with a short statement work better or a video statement? Only written statements from three customers or a longer one from one person? As a separate section on the landing page (as with Unbounce) or rather in a direct context to individual features (as with ConvertKit)?

An important note: Never use names, photos or other personal information of customers without your explicit consent! Few real testimonials that are used in accordance with data protection are worth more than fake ones, because sooner or later they stand out and then do more damage than they used!

4. Use the right visual elements

Use visual elements to control your visitors' attention. Customers not only love solutions to their problems, but also prefer when these solutions are skillfully staged and show them their desired scenario, so to speak. You can then imagine even better to use it and are more willing to buy it. Freeletics has used this principle from the very beginning:

Visual content also offers other advantages: Because if you show your visitors how your offer can make them a better version of themselves, they may only become aware that this is possible. They then develop a desire for your offer.

5. Establish direct contact with your customers

Make it as easy as possible for your customers to contact you; be just one click away: offer a live chat function and make sure in the long term that you can get in touch with each other at any time via newsletter or email.

Yes, bidirectional! This makes you independent of communication channels such as social media, the future of which you can hardly influence. Long live the good old mailing list. ;-)

6. Give clear instructions. Multiple times.

You may be wondering about this tip in several ways. On the one hand, because an instruction sounds very demanding at first, and on the other hand, because the repetition puts a lot of pressure on the visitor. In practice, however, it is shown time and time again that a lack of calls to action - so-called "calls-to-action", or CTA for short - leads to visitors consuming our content, but then leaving our site satisfied, so to speak.

We leave the opportunity to make direct contact or generate leads, for example with regard to point 4 of this list. And depending on how extensive your landing page is, the more important it is to place your (one!) "Offer" several times. A look at data such as the scroll depth can certainly be a valuable guide.

7. Adjust the scope

Depending on the offer and target group, it may make sense to adjust the scope of your landing page. Create a long-form landing page with detailed information to anticipate as many questions from your customers as possible. This is particularly advantageous if you offer products that require explanation or if you want to attract organic traffic with an SEO landing page and aim for good Google rankings. CrazyEgg has prepared a few interesting examples in this regard . In contrast, create "short-form landing pages" to advertise easily understandable products or to quickly motivate the user to take an easy action (eg a click on another page). But less is often more.

8. Pay attention to the shape

Usually we read websites according to a Z or F pattern, i.e. from top left to right, bottom left and back to the right. Accordingly, the recommendation is to place important elements such as logos, product images and, above all, buttons at these focus points.

How do I formulate good hypotheses for my A / B tests?

I'll put it this way: if you test, you already do a lot of things right (you as others).

Whoever bases these tests on hypotheses is almost at the forefront.

However, there are "good" and "bad" hypotheses, because tests are often viewed in isolation or only guesswork. These then end in sometimes more and sometimes less successful A / B tests - whereby "successful" in this context often means that there was no significant improvement.

With such unstructured tests, you can temporarily lower the bounce rate or increase the conversion rate, but the performance will not necessarily change permanently. And even worse: after the test, you usually have no idea what led to the observed change in performance, and certainly not why.

So before you do a test, think about what you want to learn from it. Because even a test that does not result in an increase in performance generates valuable insights on this basis , which in turn can be useful for the next test.

So a good hypothesis is

1. falsifiable (Am I right or wrong in my assumption?),
2. justifiable (How do I get this assumption?),
3. generates learnings (what do I learn from it?) and
4. can have consequences (what can I deduce from this?)

"The bigger the button, the more users click on it." For example, it's not a good hypothesis. Although it is falsifiable (ergo: size does not matter), it stops at the reason. What do we learn from this? What exactly are we changing on the landing page?

Testing doesn't just mean trying it out!

The following wording would be better:

"If we enlarge the button, the likelihood of a click increases because users are now aware of it".

This is falsifiable (the probability does not increase), justifiable (users do not see the button - this is where data comes into play!), Generates learnings (for example: the size alone is not decisive) and can have consequences (this is how buttons have to be designed in principle) so that the user perceives them).

So in future use the scheme "If... then... because..." to formulate your hypotheses.

How can I test the hypotheses on my landing pages?

Have you discovered potential and formulated your ideas as testable hypotheses? Then let's get down to business: testing the hypotheses.

A / B tests (or split tests) are usually best for this. The original version of a landing page (the so-called "control variant") is compared with one variant (A / B) or more than one variant (A / B / n), each of which changes a variable (eg " Layout "or" button color "or" social proof "). It is crucial that these variants are selective and you can always understand why the target metric of your landing page has changed.

Sometimes, however, it makes sense to test several variables within one variant - for example an image and the button text. We then speak of multivariate tests . With such tests, we want to determine which combination of variables performs best.
And then there are multi-armed bandit tests (MAB test for short). From now on it gets really complicated - or actually not, because here an algorithm usually takes over and tests different variables and variants sequentially in order to find the best combination as quickly as possible.

In the A / B test we have to learn (= test phase) before we can benefit (= implementation of the winner variant). However, visitors are forced to land on the variant that will later turn out to be less than optimal, which means that we have lost profits in the retrospective. We are talking about the so-called "explore exploit dilemma". Bandit algorithms, on the other hand, try to minimize opportunity costs and remorse (the difference between the actual uplift and the uplift theoretically possible through continuous optimization).

Alex Birkett explains this very well in his article at CXL, including a discussion of the advantages and disadvantages compared to the other two methods.

At the latest in this discussion it becomes clear that we need tools for landing page tests. And you are spoiled for choice. An important aspect is certainly always data protection compliance and the basic functionality with regard to your planned test scenarios is also important. But now the differences between the individual tools are much deeper in the details. There is no way around it that you have Adobe Target , Google Optimize (free), VWO , Optimizely , AB Tasty , Kameleoon , iridion , Bunchboxetc. have a look yourself and / or use a tech consultant to help you in the first step just to make a tool comparison like this one.

Strategy is essential - also with the landing page optimization

Finally, if we summarize the rough steps for testing and optimizing landing pages, we can probably agree on the following six or seven:

• Analyze: The first step - as soon as your landing page is live - is the quantitative and qualitative analysis to identify optimization potential.
• Concept: Define target metrics and the number of variables to be tested.
• Forming hypotheses: In order to realize this potential, change projects are required, which we formulate as testable hypotheses (and ideally also prioritize, but that would go beyond the scope here; see the PIE method).
• Testing: Implement one hypothesis after the other as tests - in direct A / B comparison or in the form of multivariate or even MAB tests. Remember to always compare the variants with the control variant - in real time.
• Evaluate and document: The test data are now evaluated and the winner variant with the highest conversion rate (with a corresponding level of significance) implemented. What should happen now, however, is the documentation of the test: the hypothesis, the solution (incl. Scribbles, layouts etc.), the results and their interpretation. This knowledge forms the basis for further experiments.
• Communicate: According to the documentation, (internal) communication, not only of test results but above all the resulting learnings, is still an underestimated growth factor. The more colleagues have the knowledge from your experiments, the sooner it will also be used in other areas and will be gradually expanded.
• Repeat: Start at 3 or 4. Depending on whether you have further hypotheses to verify or need to start earlier in the process.Overall, testing landing pages to sustainably increase the conversion rate is a complex matter that involves a lot of work for marketing managers, designers and developers. The information about potential customers obtained from this alone is worth the effort, not to mention the uplift generated and the associated increase in sales.

This guest post was written by Vishal Agarwal of Digital Garg.