Split Testing – Introduction
What is Split Testing and why is it important for my Amazon business?
Split Testing is a great tool to use if you want to further increase your sales and improve your ranking on Amazon.
Split Testing (or A/B Testing) is a method used by both Marketers and Developers to understand with real data what works best with their customers/users to drive the result that they want (for example, an increase in sales).
The concept is to test an existing element (called “Control” or “A”) compared to a change we want to make to this element (called “Variant” or “B”).
For example, you might want to know which Main Picture you should use for your listing in order to drive more shoppers to your product and to increase your sales.
You think Picture A is the best, but your friend Bob thinks Picture B is much better.
So which one do you choose for this critical decision?
Without A/B testing you’ll be choosing with your gut.
With A/B testing you’ll be choosing based on data.
How does Split Testing work on Trendle?
Read our Step-By-Step article.
With just a few clicks you can choose which product you want to perform a test on. You choose what variable to test, over what time period and how often to rotate between the Control (“A”) and the Variant (“B”).
The application will then apply the changes as defined by you until the test finishes. At any point during the test you can open the ongoing results to view sales, revenue, sessions and conversions.
Once the test finishes you can review the data and apply the winning variant to your listing immediately and permanently … or at least until your next split test!
You can also end tests early if you want.
You can currently run up to 50 split tests in parallel at any one time across all Amazon Marketplaces (USA, Canada, Mexico, UK, Germany, France, Spain, Italy, India, Japan, Australia)
What can I test?
- Main Image: A shopper’s attention span is limited. They’ll probably rarely read your beautifully crafted titles. They’ll instead probably scan images on the result page, glance at the price and click, or not. Images are all about emotions. If it looks like what the shopper is looking for then he’ll click. If it looks professional and trustworthy, he’ll click.
- Other images (coming soon): Once a shopper is on your product listing, other images, as well as their order, will help convince a shopper to buy your product.
- Price: Is cheaper always better? There’s only one way to find out! Word is that Amazon shoppers are getting wiser and would prefer spending a bit more money to get better quality than to buy always the product with the cheapest price. Test that and find out for yourself.
- Sale Price: Many people believe showing a discounted price will lead to increased in conversion. It doesn’t matter what they believe. Test it on your products, and find out the truth about what works with your customers.
Note: Amazon is now beginning to hide the discount % (especially in the USA), therefore this test might become less relevant over time.
- Title: Title = Keywords. What are the keywords that will convince your customers to click? Let’s not forget which keywords are important for Amazon to index your product on in search results. Balancing the two can be tricky. So how to know which one works best? Split Test can help with that.
- Bullet Points: It’s widely believed most shoppers won’t scroll down to the listing description too often. So your Bullet Points are critical in converting a browsing shopper into a customer. You may want to test the order of your bullet points. Experiment with emoticons. Compare information filled sentences to snappy short bullets.
- Description (Note: EBC description changes are currently not possible through the Amazon API): For those customers who will make it down to your product’s description, you can try out different sentence structures to see which lead to better conversion.
- Hypothesis: Write down what the test hypothesis is. For example, “I think my friend Bob is wrong, Photo A is much better as a main image than Photo B. Photo A will drive more shoppers and result in more sales”.
- Metrics: Define your Hypothesis further by deciding which Metric(s) you will measure. Quite often it will be “sales”. But it can also be “sessions” which is how many shoppers have visited your product.
- 1 Element: Only test one element at a time per product per country. Try to avoid changing other variables whilst the test is running otherwise it becomes difficult to determine what factor influenced change the most. For example, if you’re testing Image A vs Image B, avoid doing price changes or marketing activities such as giveaways which would lead to non-typical traffic and sales. If you run Sponsored Products campaigns, we recommend you minimise the amount of changes you do to that campaign. The goal is to try to keep everything as constant as possible, except for the element you are testing.
- Time: Run the test for as long as possible. The more data you collect the clearer the result will be. We recommend at least 14 days as shopper behaviour is different on weekdays than on weekends therefore you want to cover at least 2 weekend. In addition, as you are testing 2 variants, you want to make sure each variant is tested on each day of the week. The longer you run the test, the more reliable the results will be.
- Environment: Consider the impact of seasonality on your products. If you sell toys, you’ll see a huge increase in visitors and sales in Q4. So if you run A/B tests on toys in Q4, the results might not be as clear-cut as at other times of the year.
- Iterate: Split Testing never ends. It has to be done continuously as you pursue perfection in an age where change is the only constant.
- Statistical Significance (coming soon): Once the results are in you need to establish whether they are reliable or not. Currently, you have to decide that for yourself by looking at the data we present to you. However, we will soon be assisting you with this decision by giving you a statistical probability of which variants will drive the best results for your product.