{"id":103632,"date":"2023-03-24T13:38:26","date_gmt":"2023-03-24T20:38:26","guid":{"rendered":"https:\/\/thegood.com\/?post_type=insights&p=103632"},"modified":"2025-05-14T15:12:54","modified_gmt":"2025-05-14T22:12:54","slug":"cro-for-small-businesses","status":"publish","type":"insights","link":"https:\/\/thegood.com\/insights\/cro-for-small-businesses\/","title":{"rendered":"CRO for Small Businesses: Optimizing Low Traffic Sites for Conversions"},"content":{"rendered":"\n

At The Good, we love to use the phrase, \u201cLet’s test it.\u201d There\u2019s no sense in making assumptions or guessing what an audience wants. It\u2019s smarter – and better for your budget – to run experiments in order to determine what actually boosts your conversion rate.<\/p>\n\n\n\n

A\/B testing<\/a> is the standard experimentation methodology for digital marketing. It provides statistical significance and a high level of confidence. The challenge, however, is that it requires a decent amount of web traffic. If you don\u2019t have enough traffic, outliers, and anomalies can influence your results, thereby obscuring the truth.<\/p>\n\n\n\n

Fortunately, A\/B testing isn\u2019t the only conversion optimization process you can use to make smart decisions. There are other tools that will arm you with sound insight so you can improve your conversion goals<\/p>\n\n\n\n

In this article, we discuss the different types of testing that business owners with smaller sites and low traffic can utilize to optimize for conversions. But first, it\u2019s important to understand why you can still test even if your site doesn\u2019t get many visitors.<\/p>\n\n\n\n

The Confidence\/Tolerance Relationship<\/h2>\n\n\n\n

Confidence is an important part of statistical analysis. It refers to the level of certainty we have that the results of a test accurately represent the real world, expressed as a percentage. Note: Confidence rate should not be confused with the success rate of the experiment.<\/p>\n\n\n\n

If a test yields a confidence rate of 95%, we can be pretty sure that the results of the test align with the results in the real world. 50% confidence means the results could be accurate, but there\u2019s an equal chance they could be totally off base.<\/p>\n\n\n\n

Testing based on small sample sizes because you have low traffic tends to yield low confidence values. This means it\u2019s hard to tell if the results of your test represent what will happen in the real world. This is true for macro conversions like sales or micro conversions like category browsing, filtering, or adds-to-cart.<\/p>\n\n\n\n

But whenever we discuss confidence, we also have to think about tolerance.<\/p>\n\n\n\n

Tolerance is your ability to deal with the consequences of being wrong. If you implement a change to your business based on the results of the test, but the results turned out to be inaccurate, could you deal with the consequences?<\/p>\n\n\n\n

For instance, suppose you test product page variations. At the end of the test, page variation #1 produced more conversions. But after pushing all of your traffic to that variant, you learn that the test was inaccurate and conversions ultimately fell, which represents lost revenue.<\/p>\n\n\n\n

In a case like this, the reduced conversions represent the cost of learning. Can you tolerate that cost? Or would it cause irreparable harm to your business?<\/p>\n\n\n\n

Some organizations have very little tolerance. An aircraft manufacturer that is conducting stress tests needs 99.99% confidence that the results they’re getting from the test they’re running are correct. If a bolt breaks on their plane in the sky, it is a high-risk situation. So they have a very low tolerance for error.<\/p>\n\n\n\n

In ecommerce, we have a higher tolerance for error. Some companies may find that 75% confidence is sufficient, while others might only need some confidence to move forward.<\/p>\n\n\n\n

When you optimize conversions for a small site with low traffic, you\u2019re simply reducing the confidence of your testing. But if your tolerance is high, confidence is less important.<\/p>\n\n\n\n

CRO for small businesses is unique because you have a high tolerance for the consequences of a bad experiment. If you\u2019re only getting three conversions per day, a 30% or 50% drop doesn\u2019t really mean much. Your job, therefore, is to determine what your tolerance is, and based on that, decide what kind of validation methods meet your needs.<\/p>\n\n\n\n

Rapid Testing: Validation Beyond A\/B Testing<\/h2>\n\n\n\n

The confidence vs. tolerance balance begs the question: How do we test different tactics on a site without much traffic if we can\u2019t use A\/B testing?<\/p>\n\n\n\n

In the absence of data to validate statistical significance, we have to lean into rapid testing. When it comes to CRO for small businesses, this is the bread and butter.<\/p>\n\n\n\n

Rapid testing is the practice of getting quick answers from people who aren\u2019t site users, but still match your customer profile. The goal is to get feedback quickly on isolated elements. This type of testing is fast, efficient, and doesn\u2019t require a large user base.<\/p>\n\n\n\n

That said, there are two downsides to rapid testing. First, testing quickly means you give up some statistical certainty. Second, you aren\u2019t testing your actual customers, so results are less reliable (but not unreliable).<\/p>\n\n\n\n

Nevertheless, when we talk about CRO for small businesses, rapid testing is your best option. Let\u2019s look at the different types of rapid testing that help you optimize a website with low traffic.<\/p>\n\n\n\n

1. Preference testing<\/h3>\n\n\n\n

Preference testing is when you present two or more variants to users who match your target audience and ask for feedback on which variant is more likely to make a user perform a specific action or feel a certain way.<\/p>\n\n\n\n

\"two
This preference testing example from UserQ shows how research participants are asked to choose between simple variations.<\/em>
Source<\/a>
<\/figcaption><\/figure>\n\n\n\n

For example, you might show three ad designs to a number of people who match your target audience and ask, \u201cWhich of these is more likely to make you click the blue button and why?\u201d<\/p>\n\n\n\n

One of the best use cases for preference testing is display ads. If a company has eight versions of an ad and wants to narrow it down to two or three to test in the market against each other, preference testing helps to quickly cut out the worst performers and provides a great gut check with an audience that matches some of the criteria of the test market audience.<\/p>\n\n\n\n

Another good application for preference testing is any marketing asset with small amounts of copy. Whenever you ask people to evaluate multiple options, it\u2019s important to keep the options simple so they can take it all in. This is especially true when you\u2019re testing language.<\/p>\n\n\n\n

For example, you might present test participants with three headlines and ask, \u201cWhich one makes you more interested in the product?\u201d Here, the amount of information that they have to retain in order to make a decision is small. If you want to test a lot of copy, break it into smaller chunks and administer multiple tests.<\/p>\n\n\n\n

2. Tree testing<\/h3>\n\n\n\n

Tree testing is a usability testing method that helps evaluate the effectiveness of a website’s information architecture. It involves presenting users with a simplified, text-only version of the site or app’s hierarchical structure, often in the form of an accordion-shaped tree diagram, and asking them to complete specific tasks by navigating through the tree. It\u2019s used to optimize site navigation, menus, and categories.<\/p>\n\n\n\n

\"hierarchy
This tree testing study from Nielsen Norman Group indicates the correct answer for a user task.<\/em>
Source<\/a>

<\/figcaption><\/figure>\n\n\n\n

The goal of tree testing is to assess whether users can easily find the information they are looking for and identify any issues or potential improvements in the organization of content. It allows you to measure things like how long it takes for someone to find something, and how likely they are to think they can discover something there.<\/p>\n\n\n\n

3. Card sorting<\/h3>\n\n\n\n

Card sorting is a generative type of research useful for earlier-stage work like product categorization and menu organization, which makes it especially applicable to CRO for small businesses.<\/p>\n\n\n\n

In card sorting, participants are given a set of cards, each representing a piece of content or functionality, and asked to organize them into categories or groups based on their understanding or perception of how they relate to each other.<\/p>\n\n\n\n

There are two main types of card sorting. In open card sorting, participants create their own categories or groups and name them based on their interpretations of the content or features.<\/p>\n\n\n\n

\"Example
Source<\/a><\/em> <\/figcaption><\/figure>\n\n\n\n

In closed card sorting, participants are provided with predefined categories and asked to place the cards into the appropriate groups.<\/p>\n\n\n\n

\"Example
Sourc<\/em><\/a>e <\/a><\/figcaption><\/figure>\n\n\n\n

The results of card sorting exercises help designers and developers understand how users perceive the relationships between content or features, informing the design of navigation menus, categories, and overall information architecture to improve usability and user experience.<\/p>\n\n\n\n

4. AI-based eye-tracking studies<\/h3>\n\n\n\n

Traditional eye-tracking studies involve tracking the gaze and movement of a user’s eyes as they interact with a website to understand how users consume information and identify areas that draw attention or cause confusion.<\/p>\n\n\n\n

AI-powered eye-tracking studies replace the need for real people by using machine learning models trained on large datasets of human eye movements. These models simulate human eye-tracking behavior and then predict the likely gaze patterns and attention distribution of users when presented with a new website.<\/p>\n\n\n\n

\"Example
Microsoft’s iTracker trains on countless images real people to understand how we explore sites with our eyes.\u00a0<\/em>
Source<\/a>
<\/figcaption><\/figure>\n\n\n\n

By leveraging AI, you can gather valuable insights into user behavior and attention more quickly and cost-effectively. No traffic is necessary, which makes it especially powerful for CRO for small business. While AI eye tracking is not a perfect representation of how real people use your site, it\u2019s a great way to evaluate which option from a group is more likely to get attention.<\/p>\n\n\n\n

5. Recall testing<\/h3>\n\n\n\n

Recall testing<\/a> is where you show users a design for a short period of time (usually five or 10 seconds) and ask what they remember seeing on the page. You would conclude the one they remember to be more memorable and eye-catching.<\/p>\n\n\n\n

For example, to test an add-on in the cart, you can show one version of the cart page design to 50 users, and another design version to a separate group of 50, and the better ability to recall they saw an enticing add-on would be a good indicator of a better placement or a more eye-catching placement.<\/p>\n\n\n\n

6. Click testing<\/h3>\n\n\n\n

Click testing<\/a> is similar to heat testing, but instead of showing where users move their mouse, it shows you where they clicked when you ask them to perform a specific task. This can be especially helpful when testing complicated interfaces like product configurators.<\/p>\n\n\n\n

When we worked with Fully<\/a>, we tested wireframes of their product configurators. We used click testing to instruct users to \u201cShow us where you would click on the page if you wanted to update the color of the metal base.\u201d<\/p>\n\n\n\n

We assumed the participants would click the word \u201cframe\u201d on the right-hand panel that had an accordion-style menu. Instead, they clicked on the image, assuming the configurator was interactive since the picture was taking up 80% of the screen. This taught us we either needed to make the image clickable or make using the menu more obvious.<\/p>\n\n\n\n

7. Preference test with open-ended responses<\/h3>\n\n\n\n

This type of test is similar to a regular preference test, except that participants are asked open-ended questions after completing their prompts. They might be asked, \u201cWhat did you like about this design and why?\u201d or \u201cWhich one are you more likely to pick and why?\u201d<\/p>\n\n\n\n

In some cases, getting an open-ended response enables you to hear the why behind a user\u2019s action, thus providing you with insight into how you\u2019re making users feel along the journey.<\/p>\n\n\n\n