For online storeowners, it’s sometimes difficult to take a chance. As humans, we tend to go with what has worked for us in the past. But taking chances is an important part of creating a successful business.

Merchants should always follow best practices. At the same time, every online store is different. Strategies that have found success for other retailers don’t always work for entrepreneurs with unique business goals.

Taking a chance requires a strong basis for testing. As any scientist will tell you, testing is the key to successful experimentation. When you experiment, it’s crucial to plot divergent (but related) strategies beforehand.

Check out some tips for ecommerce experimentation below.

Strong web developers use A/B testing to ensure a successful campaign, feature or strategy. As you experiment, the practice will become a crucial part of your testing arsenal.

To perform A/B testing, you’ll want to designate two separate strategies that have similar goals. For instance, a simple A/B test may designate two separate groups of customers to whom you’ll send out promotions via email. Choose a variable (the subject line, for instance). Blast out the emails with different subject lines to the different test groups and measure the open rate. That way, you can find out which subject line performed better and tailor your future email promotions based on the data.

It’s important to understand what’s successful and what isn’t for your business. To do this, you’ll need to realize certain factors (called key performance indicators, or KPIs). By comparing KPIs in an A/B test, you’ll have a much better grasp on which campaign performs better.

For instance, if you’re testing two different homepage designs, take into account factors like the bounce rate and the percentage of visitors that buy. That way, you can understand how effective the homepage is in driving shoppers through the purchasing process.

In order to judge your KPIs, you’ll need the right software in place to collect the data. Most shopping carts will give you access to built-in analytics. But in some cases, you may need more advanced functionality.

For instance, if you’re trying to get a really strong idea of which Google Ads drive the most traffic, it may make more sense to rely on Google Analytics. If you’re measuring email campaigns, fall back on MailChimp’s built-in ecommerce tracking.

Every good scientist has to get his or her feet wet. When you start to experiment, kick off the process with a low-risk item like a coupon or discount promotion. That way, you can get the hang of experimentation without much at stake.