Conversion optimization is not one of the easiest jobs in the world. Anyway, the name itself is misleading, because if we want to optimize only the conversion, the recipe is very simple. All you need to do is lower the prices of all products in the store and the conversion will jump to the ceiling, only business will go to the bottom. The optimization is about growth, but above all, the the conversion and increase in the average value of the visitor
The average value of a visitor consists of not just the conversion rate and the average value of the basket. The customer’s life value (CLV) is the sum of revenues that the customer will generate for the company throughout the duration of the relationship. Calculating this ratio for the entire period of cooperation between the client and the company, which may last even several dozen years, would be time-consuming and burdened with a high risk of obtaining an incorrect result, therefore CLV usually counts for a shorter time period, eg 12 months. This allows you to overcome the impact of factors such as a change in the offer or competition.
When planning to optimize sales in the store, you need to look for both ways to increase conversions and increase the value of the shopping basket.
A typical optimization approach
You have a whole rack of ideas. You’ve gone through a lot of case studies, read the other two books about conversion optimization and participated in outstanding webinars, where lots of examples are shown. You guess, therefore, which activities work and which do not. Sometimes it happens that a person with high knowledge of user experience is taken for optimization, believing that this is enough. All these approaches are insufficient. Why?
Myths in the world of optimization
All myths in the world of optimization have their source … in you. Because you are convinced that all the knowledge you have gathered is enough to achieve good results. How do we know all this? With our own mistakes. Because each of these myths used to be the truth as well.
Your opinion matters and influences optimization
Opinion matters, but not yours, only your end customer. It is their opinion you need to listen to, because at the end the customer decides whether they will make a purchase in the store, or will resign. Every expert has an opinion or idea how the site should look like. However, it is not the opinion that sells. If you do not clash the idea with the customer’s opinion, it is not worth much. So, when preparing for tests, it is important to start a survey earlier in the shop on abandoning the basket or more generally at the end of the visit and ask about: the purpose of the visit, whether it was possible to implement it, what can be improved on the site.
Conclusion: Facts matter, not opinions. The facts should be presented based on data and analysis.
Experience tells you what works and what does not on the site
You can not assume that if you have corrected the product card 10 times in exactly the same way, then in every subsequent optimized store you have to correct it in exactly the same way. This is a typical mistake of hastily prescribing a medicine without examining what the right disease is. If you have read the Case Study on which the change in the color of the button has increased sales, it does not mean that you should think about such a test on the currently optimized website.
Conclusion: The website should be examined to see how the visitor uses it, and how the sessions of those who bought from those who were just browsing differ. However, the most important information that one can find is the customer persona. That information, if used right will allow you to increase your sales. You should also keep a keen eye on the customer’s on-site behaviour, acquisition channels and technology he uses.
I have ready templates to implement on the site
If you already have several optimized pages on your account, you’ll start to fetch patterns that change more often and which, on the contrary. This is the same trap as in the previous myth, only much more experienced optimizers fall into it. It consists in the fact that you are taking shortcuts. You minimize the emphasis on research and analysis of the website, and even get out of action to start the first changes, because they have helped so many times. However, one should always keep in mind that it is unwise to use pre-optimised templates. If you are using an agency to provide you with some help, you should be aware that no proper agency would use pre-optimised templates for all their clients. Of course, templates might be tempting because they are cheaper than anything else, but the results they bring might either make it or break it for your business. It has been established, that around 90% of businesses that use pre-optimised template die.
Conclusion: avoid routine or pre-optimised templates, examine each page patiently and then diagnose.
When interpreting the results, the duration of the test is the most important
To complete the AB test and correctly interpret the results and draw conclusions and recommendations, several criteria have to be met. First, statistical confidence should be at a sufficiently high level. We encourage you not to go below 95%. Second, the number of conversions should be significant and proportional to the size of the store. Usually, the level of 200 – 300 conversions is enough, but what if the shop makes so many conversions in one day? Then it is definitely not enough. So how many of these conversions should it be? It depends … on the duration of the test. The test should last absolutely minimum 2 business cycles (for stores it means 2 weeks). If during this period you collect about 300 conversions for each test variant + you achieve statistical confidence at the level of 95%, then you can go ahead with the test.
Is that all? Of course not, because the most important is the question of what exactly you want to achieve within the given test. Is only the increase in conversion only? Certainly not. You care about the increase in the average value of the visitor. Therefore, pay attention to tools and calculators on the Internet that count statistical confidence only for the conversion itself. Pay attention that all good tools should analyze the size of your traffic and your conversion rates in order to calculate your statistical significance.
Conclusion: In the interpretation of the results, the most important are three factors: the number of conversions, the duration and the level of statistical confidence. All factors must be met to consider the test result as reliable.
What should effective conversion optimization look like?
Contrary to appearances, the process of effective optimization is very simple. The most important thing is to set goals to achieve. Remember that the shop is not alive by the conversion itself.
The next step is start measuring and analyzing the website. Test according to the developed plan. Finish the test and interpret the results in accordance with the best practices, so take into account the number of conversions, duration and statistical confidence.
Step 1 – Data collection and careful analysis
The process of conversion factor optimization should be started by collecting and then analyzing quantitative and qualitative data. Quantitative data allows identifying elements of the site that need improvement. You should use Google Analytics for this type of analysis (of course, you must first ensure that the tool is correctly configured). Information on which elements of the website are ineffective is just the beginning – after all you need to learn the reasons for their low effectiveness. For this purpose, it is worth conducting qualitative research. These include surveys among website users, usability studies or expert analysis (heuristic).
Conducted in the first stage of the analysis is the basis for creating a list of recommendations for changes in the website. This is a key step in the conversion rate optimization process – it is here that you decide what elements of the website and in what order you will test. Recommendations are best written in a spreadsheet containing such columns as the changed sub-page (eg product card), problem, solution, source of recommendation (eg usability testing).
Step 2 – Developing a recommendation for changes
After writing out all the recommendations, you need to prioritize them. To determine them, it is best to use a two-factor assessment, where you assess the growth potential and the difficulty of implementation (on a scale of 1 to 5). The higher the growth potential, the more points you assign recommendations. Assessing the difficulty of implementation works the other way round – the easier it is to encode the change in the site, the more points the recommendation receives.
For example, if you plan to change the header, which will take the developer only 5 minutes and in your opinion will bring a big increase in the conversion rate, its growth potential is 4, and the difficulty of implementation is 5. The strength of recommendation is the product of these two ratings:
potential for growth X difficulty of implementation = strength of recommendation
If you change the header, the strength of the recommendation is 20 (out of the 25 possible points) – so it should be high on the list of priorities to be tested. When assessing the growth potential of a given recommendation, you should take into account the number of users who interact with the indicated subpage or element.
Step 3 – Testing designed changes
There are two types of tests: A / B tests and multivariate tests. A / B tests consist of testing two or more different versions of the site. The mechanism divides the website visitors into two groups and displays them the appropriate version of the site. Multivariate tests involve testing combinations of different variants of individual sections of the site. For example, if you’re testing a header and button, and each of them has two variations, the total number of combinations equals four. Multidimensional tests require a lot of website traffic to achieve statistical significance.
Regardless of which type of tests you decide (first we recommend A / B tests), you have to remember a few iron rules. First of all, the test should last 10-14 days. Such a period of time is needed to eliminate the influence of external factors on the test result, such as the day of the week, weather or the sending of promotional mailings. It should also include the full customer purchase cycle, which, for example, on Monday and Tuesday searches for available products on the market, on Wednesday and Thursday compares them, on Friday makes a decision, and buys on Saturday afternoon. Capturing a full cycle in the test allows for reliable results.
Step 4 -Selecting the test winner
After 10-14 days it’s time to check which of the tested versions turned out to be more effective. You must pay attention to the statistical significance of the test results. If it is lower than 95%, there is a good chance that the results obtained are accidental and the implementation of permanent changes to the website may not bring an increase in the long-term conversion rate.
It’s important that each version tested achieves at least 100 conversions. Otherwise, the test results may be subject to a large random factor. In the test, it is worth paying attention to which version was more effective among new and returning users, and when deciding about implementing changes permanently, take into account that returning users will adapt to changes over time and will probably start to convert as effectively on this version as new users.
After determining the winning version of the season, implement it permanently on the site and target 100% of traffic on it. It’s worth doing it as soon as possible, because each day of delay is a loss equal to the difference between the conversion rates achieved on the tested versions. Remember not to implement changes from two different tests at the same time. The fact that the changes tested separately brought an increase in the conversion rate does not mean that their combination will be equally effective.
Step 5 – Test analysis
Regardless of the test result, a detailed analysis is required. It’s not enough to compare conversion rates between versions. You need to know how the effectiveness of shopping paths was presented on individual versions, what were the user engagement indicators (eg average session duration or bounce rate) or with what elements users interacted with.
Your goal in the conversion rate optimization process should be to bring a situation where, to a specific segment of users, you are able to match the version of the page that is most effective for them. That is why it is worth checking the effectiveness of the tested versions in relation to user segments, separated for example by traffic sources. Users who visit the website are usually at a different stage of the shopping process than users who are entering from AdWords ads. It is worth remembering and striving to match the marketing message to the needs of individual user segments.
The result of the test analysis should be knowledge that you will use to prepare a recommendation for changes and hypotheses for the next test. At this point the process is looped – you go back to the analysis (which is worth refreshing from time to time) and go through the next steps one by one until you finish and analyze the next test.
What analysis must be carried out before developing hypotheses and starting testing?
- Technical analysis, e.g.: impact of browsers on sales, the impact of devices for sale, the impact of page loading speed for sale.
- Website content analysis, e.g., analysis of problem websites that have a low number of page views or conversions, effectiveness of landing pages, analysis of abandoned sites, the impact of ways to navigate the site for sale.
- Search analysis, e.g.: What terms do clients search for? What do they find? Where are they looking for ?, What things are not found by customers? General search versus precision.
- Thermal Map surveys, eg: analysis of clicks on various types of devices, analysis of mouse movements, page scroll analysis, analysis of completing and abandoning forms.
- Study of Opinions and Intentions, for example: questionnaire about abandoning the page, survey on abandoning the basket.
- Heuristic research, e.g.: examination of the purchasing process, information architecture research, assessment of marketing exposure, verification of UX best practices.
Conversion optimization is a process. If you act ad-hoc based on your ideas or experience, most of your tests will be unsuccessful. Systematicity, patience are much more effective than genius ideas or searching for magical quick solutions, like growth hacking which is often unregulated and should not be your primary goal in conversion optimisation. Do not forget about when to consider that the test can be completed and what is its real impact on the sale in the store.