Your website is the showcase for your organization.

You set this up to help a visitor make choices and to make a purchase, for example.

Unfortunately, the proper design of a website is often a major challenge.

Elements are regularly added to the website with the idea of ​​better helping the user, while this has the opposite effect.

By applying conversion optimization it is possible to better understand website behavior.

As an organization you can then respond well to this!

Conversion optimization is the scientific validation of adjustments to a website. This way you test whether an adjustment is actually an improvement or whether it is based on chance. 

In this way you ensure that your organization only makes pre-validated adjustments that actually have a positive impact. 

This is an essential part to grow as an organization and to better serve your users. 

Loss-making adjustments are avoided in this way. Risk management from the top shelf.

However, there is also a caveat to conversion optimization. 

To be able to make a scientific judgment, there are requirements for the amount of visitors and transactions. 

That can be difficult, especially if you have too few visitors. Because what does this mean if you only have a handful of website visitors per day?

Fortunately, there are still opportunities to improve your website data-driven if you have little traffic. 

Analyzing the users on your website is essential to see where users drop off and how you can serve them better. 

In addition, you can see what has a positive effect and what does not in a data-driven way.

In this blog we discuss how to improve a website for your users when you don’t have enough traffic to actually test for conversion.

Let’s go!

What steps can you take to improve low-traffic websites?

Low website traffic is a problem for organizations of all sizes. 

As mentioned earlier, a certain group size is needed to validate whether an effect is due to chance or not. 

Yet data-driven working is always better than blindly implementing changes. Having insight into website behavior helps your organization to set priorities. 

Data-driven working is divided into two phases. Researching your users and actually validating your solutions to user problems.

1. Research your users

There are all kinds of surveys that can help your organization understand how users use the site, where they drop out and what objections they have to not make a purchase. 

Gaining insight into these bottlenecks is the starting point of your website optimization. If you do not know what is happening and for what reason this is happening, getting more transactions is quite difficult.

Web analysis/quantitative research provides insight into what is happening on your website. Which pages do your users come to, on which device do they visit your website and where do they drop out? 

The possibilities are endless and this is an essential part of any analysis. 

Surveys (questionnaires) offer you the opportunity to ask questions directly to your users. This can be done in many ways. Think of showing a questionnaire to users who want to leave the website or asking questions to users who have made a purchase.

In this way you can ask specific questions about a user’s reasons for making a purchase or not.

Customer service research is also an easy way to identify customer problems. 

Customer service employees are in daily contact with your customers and know exactly what is going on and what customers encounter. 

Session recordings and heatmaps provide insight into how users actually navigate through your website. By viewing session recordings, you can immediately see what users encounter. 

Heatmaps also provide insight into which part of a page is viewed and to what extent important elements on your page receive sufficient attention.

User testing or usability research gives the possibility to have users perform an assignment on your website. You can ask the user directly about the motivation behind all the actions he or she performs on your website. 

This provides a lot of insight into the user-friendliness of the website and whether it meets the expectations of the user.

After all, something can be designed in a certain way, but get in the way of visitors finding what they are looking for. 

Competitive analysis provides insight into how other organizations have set up their website to help users make a purchase. 

This way you can immediately find out differences and you also get direct insight into how you can distinguish yourself from the competition. 

You have to take into account that competitors may implement things without testing. So copying competitors does not necessarily increase your chance of success. 

Copy ideas and A/B test from competitors

2. Validate your solutions

Now that you as an organization have a good insight into the bottlenecks of your user, it is time to actually implement improvements. 

You do this by converting your ideas to improve it into hypotheses. We do this on the basis of the so-called ABC principle: assumption + evidence + verification. 

Assumption: Based on [data], we expect [adaptation] for [user group] to motivate them to [behavior] for this [reason].

Evidence: This assumption holds if we measure an [effect] on [growth metric]. 

Control : The assumption appears to be [(in)correct]. We learned that [lessons]. The next step is [follow-up action].

By performing a runtime calculation, you know exactly how long your test should run and what the impact must be on your metric to achieve significance. 

Significance is the scientific basis that determines the probability that your adjustment is due to chance or not. 

Your duration calculation depends on two factors:

  1. How much traffic is there on the page you want to customize?
  2. And how many people convert on the target you want to influence?

There are a lot of tools that can help you make a maturity calculation. An example is this tool .

A/B testing calculator
A/B testing calculator
A/B testing calculator

Although this tool can be very useful to help you gain more insight into your optimizations, we strongly advise you to apply one of the strategies below.

This is because the calculator may display unrealistic numbers when entered. This can give a wrong picture on which you may want to base your approach. 

Example: you will achieve a 30% conversion uplift to measure a significant difference in four weeks. Such a large improvement does not often occur in practice.

But how do you deal with low-traffic pages that do not benefit from the runtime calculation?

Unfortunately, there is no standard answer to this. However, there are ways to gain insight into this. 

Combine pages that are similar

If you have pages with a similar layout such as product detail pages or category pages, combining these pages in an experiment can provide enough traffic to get a significant result. 

Using a micro conversion as the target

Transactions and leads are the clearest metrics to hang results on. 

These are very tangible and can be directly calculated into expected extra turnover. If these metrics do not take place sufficiently for experimentation, it can be very relevant to measure micro conversions. 

For example, you can think of the number of users who click through or the number of users who add something to the shopping cart or come to a product page. 

When using micro conversions, it is essential that there is a logical link with your end goal. And that you also take the impact on your end goal into account when analyzing the results. 

Lower your significance level

As indicated earlier, the significance determines to what extent an improvement is based on chance. 

You often see that a limit of 95% certainty is used here. That would mean that of every 20 winners, 1 experiment is not an actual winner. 

By lowering this limit to, for example, 80%, you can test more things, but the risk that something is based on chance is also greater. 

And therefore also the risk that you implement a change with a negative effect. 

If you choose this, always include other insights outside of your most important metric to get a good overall picture of your experiment.

With low numbers, this method is the least recommended, so let’s quickly move on to the next method.

Implement and monitor long-term adjustments

If there is really very little traffic, it may be an option to simply implement changes and monitor what happens to your metrics through quantitative analysis (web analysis) from the moment it is implemented. 

Also add an annotation to Google Analytics right away.

Do you see a structural positive effect on your conversion metric?

Then your adjustment does not seem to do any harm. If it shows a negative effect from that moment on, you can therefore consider reversing your adjustment.

Be aware that you are now looking at correlation rather than causation. 

In other words: you can never prove afterwards that an improvement or deterioration in your conversion rate is the result of your change.

User testing / usability research

For larger re-designs or adjustments, a user survey is recommended. 

This gives you direct insight into the reactions to adjustments you want to make. 

The disadvantage here is that often only a handful of users participate in the research, which may not be representative of all users of your website. 

In addition, users will be more aware of your site because they have to give feedback. This can cause different behavior than when they actually visit your website themselves.

Stay practical

It is now clear that data-driven working is of great importance. Even if your website is not immediately suitable for A/B testing, there are still many options. 

All in all, it’s essential to know what helps your users and what doesn’t. If you don’t work data-driven, you won’t get any wiser. 

So it may just be that you are making the wrong adjustments. 

Of course you don’t want that for your organization. So try to use the most practical way for your organization and learn what works and what doesn’t. Also in your approach. 

Good luck actually improving your website and helping your users!

error: