How to increase website conversion rate (CR): instructions for split tests

How to increase website conversion rate

The main task of the site is to convert traffic coming from various advertising channels into customers. However, it often happens that visitors leave the site without having completed any targeted action on it. For a business, this is lost leads, and, accordingly, profit. We explain why this happens and how to increase website conversion rate (CR).

Site conversion is an indicator that allows you to evaluate the percentage of visitors who completed the target action on the site to their total number.

How to increase website conversion

A conversion is a user’s transition from one status to another. Before making an order, the user overcomes a certain path. And when moving from one stage to another, the conversion is measured. For example, 100 people went to the site and 10 of them completed the target action (for example, left a request). This means that the conversion from a site visitor to a submitter was 10%. Website conversion usually depends on two factors – the content of the sites and the quality of the incoming traffic. If you change the traffic source or its settings, the conversion of users on the site will also change.

To increase the conversion rate (CR), a set of measures is being taken to optimize this indicator – CRO (Conversion Rate Optimization). This work usually begins with basic tasks:

  • identification and correction of technical errors;
  • Ensuring a page speed of  at least 50 points according to PageSpeed ​​Insights ;
  • checking the mobile version of the site, etc.

These are universal tasks suitable for any site. After all the technical issues have been fixed, you can go further – draw up a set of measures necessary to increase the conversion of your particular site.

Read Also: How to return a user who left the site – “Come back, I will forgive everything!”

How to start writing these recommendations? Of course, from funnel analytics and identifying weaknesses on it – the stages at which most of the visitors leave the site. Further, based on this information, it is necessary to build a hypothesis and conduct an experiment – split testing. According to its results, it is already possible to decide on further changes on the site.

So, to increase the conversion of the site, you need to go through 4 main stages: funnel analytics, hypothesis building, split testing and analysis of the results.

Let’s take a closer look at the work at each stage.

1. Analytics

Analyzing the sales funnel

Usually, standard analytics systems are used to identify problems with site conversion: Google Analytics. Using these tools, you need to study the entire funnel that the average visitor goes through – from searching for a product in Google to making a transaction. Your task at the first stage is to find the places where the most customers leave.

Most often, the loss of visitors occurs at the checkout stage. The vast majority of customers place the product in the cart, but do not take further action. You can identify the loss at this stage using the “Webvisor” or “Click Maps”, which collect information about the behavior of site visitors. The report shows the chain of transitions on the site. It makes it easy to see the top performing pages and the best customer journey sequences.

As a rule, it’s more important not to even hypothesize… Everyone might think that the red button works better, move it up, change the text, or remove the left block from the site. For example, we replaced the “More” buttons in the listings with “Reviews” to specify what the user wants to see.

It is much more important and more correct to look for insights in analytics and build on them when building a hypothesis. For example, to analyze user behavior and see that on mobile devices on specific pages the conversion is lower than on the desktop. Maybe the buttons are not lined up correctly, they are too small, or there are repeated elements, etc. – all this allows you to identify the analyst.

Read Also: How chatbots help improve business performance ?

Conduct surveys and interviews

If the analysis of visitor behavior on the site did not give the desired results – namely, ideas for putting forward hypotheses – go to plan B. What do you know about your customers? Why do customers buy this or that product, what do they really want to get, what result to achieve (we recommend reading our article on the popular Jobs to be done theory on this topic )? What difficulties do potential customers have that stop them from making a transaction? You can get answers to these questions in the course of the good old survey or spot interviews of your customers (or potential customers).

There are several ways:

  1. Placing special forms on the site  – just a few simple questions, and you will receive all the information you need.
  2. Calling clients. Buyers leave their contacts that can be used for communication.
  3. Survey mailings. Perhaps the most inefficient way to get the information you need, but there is a trick – see below.
To motivate customers to participate in surveys, come up with a special loyalty program. Give bonuses or give a discount to those who are willing to spend their time and answer your questions.

2. Building a hypothesis

Analyzing the sales funnel and / or conducting surveys allows you to build a hypothesis – put forward an assumption about what needs to be changed on the site, or on a particular page, in order to increase the conversion.

Here it is necessary to say about the importance of choosing the right hypothesis. It is not possible to test all assumptions at once. Firstly, this is due to the waste of large resources, and secondly, you simply cannot understand what ultimately brought the desired result. Therefore, the task of a specialist is to determine hypotheses that will give the maximum effect in the shortest possible time.

In the process of constructing and confirming hypotheses, it is most convenient to keep a table – to record all assumptions, experiments and their results in it. You should start with a hypothesis, the implementation of changes for which requires less time and financial resources.

For example, changing the name of a product will be much faster than refining the design. Even if the specified hypothesis does not demonstrate the proper result, its implementation will take a minimum of time and it will be easier to roll back the changes.

Some ideas for building hypotheses were shared with us by German Antonov, marketer at TiboMedia.

I will tell you about several TiboMedia approaches that we have taken as a basis for providing website conversion optimization services. We consider these methods to be quite effective and applicable in most cases.

So what can you do to increase conversions:

1. Change the call to action;

The user on the site performs the target action for some reason. You can strengthen his intentions, for example, by providing some benefit in return. For example, a free consultation, guide, checklist, book, cost calculation and much more. In the language of marketers, this is called a lead magnet.

2. Change or add communication methods;

The vast majority of sites provide the ability to call, order a call back, send an application form, etc. as the main ways of communicating with customers. Each action has resistance. It is harder for visitors to call than to order a call back. Ordering a callback is harder than requesting a chat in a messenger. Requesting communication in a messenger is harder than writing in an online chat on a website. Etc. All targeted actions on the site have their own resistance. To increase the conversion of the site, you can revise the formats of communication with users. The easier it is to complete the target action, the higher the conversion of the site.

However, there is one consequence of this approach that marketers need to consider. The easier it is for the user to complete the target action, the lower the value of the conversion. Therefore, when experimenting with different methods of communication, it is necessary to monitor the quality of incoming calls so as not to reduce it.

3. Add information to the site;

There may not be enough facts on the site to explain the value of your offer. The user may leave the site without seeing any reason to cooperate or purchase a product.

4. Change the client’s path;

If you have a complex product and a multi-page site, marketers can design the customer journey between the pages of the site. The simplest example of such a path is choosing and placing an order in an online store. First you need to go to the desired catalog page, filter the products according to your criteria, select the purchase options (customization / number of products / products in the kit, etc.), select the delivery method, specify the delivery address, delivery time, packaging and pay.

At each stage of the customer journey, some users will “fall off”. It is important to analyze the entire customer journey on the site and put forward hypotheses that will help reduce losses when moving from stage to stage and “straighten” the funnel.

5. Conduct a technical audit of the site;

Sometimes the user does not perform the target action on the site for technical reasons. For example, the order page does not load, or layout problems occur when the page loads. There can be many reasons. To put forward hypotheses, you need to conduct a technical audit of the site.

6. Change landing page;

Sometimes changing the landing page helps. For example, they drove traffic to a product catalog page, and then transferred it to a page with special offers, to the main page, or to a specific section of the catalog. In this case, the main thing is not to be afraid to experiment. Sometimes the obvious decision to lead users with the interest “office desks” to the directory section “office desks” is not the most effective.

In each case, it is necessary to evaluate the customer journey, as well as correctly prioritize the hypotheses for testing. Unfortunately, there are no universal solutions in this matter.

3. Confirmation (or refutation) of the hypothesis

When you have formulated and fixed a number of hypotheses, you can begin to test them. Here it is important to introduce changes to the site consistently, step by step tracking the results of each change. Thus, no other design improvements that are not included in the changes according to the hypothesis should be implemented during the testing phase. Your task is to keep the experiment pure.

The hypothesis is confirmed through A/B testing (or split testing). During it, visitors are randomly divided into several groups and then they see different versions of the site. This process is dynamic and cyclical – after each successful hypothesis, we fix a new CR indicator and proceed to the next experiment.

What are the tools for split testing?

Not so long ago, we tested various button labels, button colors, block layout, and block removal from pages. Worked with Google Optimize and Tag Manager. I won't say that these are ideal tools for experimenting, as they slightly increase the load on the site. However, this is the easiest way for the marketing department to test hypotheses and make at least some decisions.
When it comes to increasing the conversion of any one particular page, everything is much simpler. Another thing is to test the hypothesis on a sample of similar pages in order to implement these changes for 1000 pages. Here the results may vary. If the old and new versions of the pages show the same dynamics, then it is better not to make changes.

If you plan to run ads, then A/B tests are very important. For several thousand rupees, you can test several page options and choose the best one.

4. Analysis of results

How to understand how much your CR has changed as a result of the test? The conversion rate is calculated very simply:

(test CR / pre-split CR – 1) * 100%

Let’s calculate the test result using a simple example:

1% – CR of the main version of the site
1.3% – CR of the test version
(1.3 / 1-1) * 100 = 30%

Thus, the conversion rate increased by 30%.

Track any type of conversion — from a visit to a case or from a case to a deal, etc. – it is more convenient in the personal account of your end-to-end analytics service .

Instead of a conclusion – cases!

Here are some cases that experts shared with us. They clearly show how in practice work is built to optimize the conversion of the site.

In order to improve the user experience on  the Fragrant World website , we conducted 15 A/B tests over 8 months. Half of them were successful, resulting in a 30% increase in website conversions. We tested almost every hypothesis separately for desktops and mobile devices, each A/B test took three weeks.

Testing steps:

1. Built a simple CJM on the site.

They decided to start from CJM (Customer Journey Map) – they prescribed the stages that the client goes through on the site, as well as the pages and tools that he uses.

2. Collected 50 ideas for improvements.

For each stage, CJM collected ideas for improvement. We compared solutions with competitors and top players in the e-commerce market. There were about 50 ideas in total.

3. Set priorities. Three criteria were used for this:

  • High priority for tests in the cart. It’s easier and more profitable to increase conversions for customers who have gone all the way and stumbled at the very last stage.
  • For each idea, the cost of development (a quick assessment in hours from the IT team) and the potential effect of improvements (an expert assessment of each team member on a scale from 1 to 10 points) were evaluated.
  • The rating was adjusted for the volume of the test. Low-traffic tests were deprioritized: the increase in conversion for visitors to low-traffic pages would go unnoticed as part of the conversion of the entire site.

4. Regularly tested hypotheses.

The preparatory stages took two months. Then the routine monthly work on testing hypotheses began. The designer designed the changes, the IT team programmed the two branches of the test, the internet marketer ran the test in Google Optimize. If the test was successful, the new version was immediately implemented.

As a result of all the experiments, the website conversion increased by 30%.

I will share one very simple case that helped increase website conversion by 25% without losing the value of conversions. We were approached by a tailoring studio from Moscow. After conducting a comprehensive audit of the site, we came to the conclusion that it is necessary to start experimenting with the ways of communication on the site.

The niche of individual tailoring involves a detailed discussion of the order: the selection of a service, the proposal of options from experts, the demonstration of samples, etc. First of all, we decided to use an online chat. The case is quite old, so we used the most popular Jivo service at that time.

Here are the features of the service that might come in handy:

  • chat for the site (tested as part of this project);
  • virtual PBX;
  • callbacks;
  • receiving letters from the mail;
  • messages from Facebook;
  • messages from VK;
  • accept payments via chat.

Experiment progress:

The site received traffic from paid (advertising in Google, Instagram) and free sources (Google, Instagram social networks, Facebook, 2GIS and other service aggregators). No changes were made to the site content during the experiment period.

Read Also: What is Google Performance Max : how to get the most out of it ?

After a few weeks, it became clear that the chat increased traffic conversion rates. To achieve the maximum result, we began to experiment with chat engagement (a special chat feature that automatically sends a message on behalf of the manager). For example, after a user spent more than 15 seconds on a page for tailoring dresses, a message was automatically sent to him on behalf of the manager with the question: “Good afternoon. What kind of dress are you interested in sewing: for every day or for going out? Initially, we tested other proposals. However, this is what showed the best results.


The site received the following conversions:

  •  direct call (calls were tracked by call tracking );
  •  sending a phone number for a callback ;
  •  the client wrote himself (the client wrote to the chat, starting the correspondence first);
  •  the client responded to active engagement (the client responded to the automatically sent offer to discuss the details of the order in the chat);
  •  sending an offline form (if the manager was not online and could not answer messages, the client was shown a callback form that he could fill out);

We entered all the data in a table. The experiment started on January 14th. 

The first line below the table is the percentage of unique users who made at least one conversion on the site. The data includes both regular conversions (call and send phone feedback) and conversions via JivoSite.

The second row below the table is the percentage of people who converted in the form of a call or submitting a feedback form (that is, data without conversions through JivoSite).

The red line shows the change in the percentage of users who called or filled out a feedback form. Blue – overall conversion rate, taking into account JivoSite. The graph shows that after connecting the online chat, the percentage of users who made a direct call or left a number for a call back fell (by about 10-15%). At the same time, the percentage of users who made any conversion increased by 20-30%, which is good news.

Of course, the question of the quality of leads that contacted via chat, and not by phone, is of fundamental importance. In this project, according to the results of several months of monitoring, it did not fall at all.

Of course, a website conversion rate of 3-4 percent is not the ultimate dream. But in this project, taking into account the conversion at other stages, Internet marketing remained cost-effective with the resulting site conversion.

Important: we do not recommend blindly trusting this case. Test chat and engagement from JivoSite on your site.

We hope you have been inspired by the results of these cases and are ready to start working on conversion optimization on your website. 🙂 Make hypotheses, conduct experiments and analyze the results using end-to -end analytics . High conversions for you!

About Post Author

Leave a Reply

Your email address will not be published. Required fields are marked *