With increasing online eCommerce competition, we recommend you steal this CRO process and level up your Shopify store success.
8 Step Process Overview
Map out present page types
Map out user types
Measure Baseline performance
Set up tracking
Check actual impact vs. expected
Map Out Present Page Types
First, you need to know what pages you have, how you can categorize them, and how many pages are in each category. Imagine that you are selling 200 products, and you need to check if your product pages are performing well. How would you do that? Out of those 200 pages, which would serve as your 'good' and 'bad' benchmarks? That’s why we need page categories - to be able to divide those 200 pages into actionable batches.
Map Out User Types
The next step is looking at user data and applying the same categorization approach to your actual customers. The simplest way to do this is to look at new and returning users separately. Then view the transition and drop-off rates for both groups. However, the important thing in terms of CRO is the products these users are buying.
For new users who have less trust and knowledge about your brand, some specific items could be enticing for both advertising and organic conversions. We call those entry-level products. One of the main goals here is to understand what those products are for people finding your store for the first time.
For returning customers, things get a bit more complicated and less generic. That’s why segmentation is key at this stage. We recommend using default age groups as presented in Google Analytics for easier data processing. Gender would typically serve as the secondary dimension. Here’s an example of how we would describe a segment:
The age group of 45-54 generated about XX% of all user sessions
Blended group metrics: Bounce and Conversion rates
Gender split: XX% users are male, and XX% users are female
After that, we would execute the same process for both sub-segments - female and male users, using blended stats as our baselines to assess performance. Now it's time to map out your basic funnels for both new and returning users, based on the insights you have so far.
Measure Baseline Performance
Time to assess the performance of your store as a whole. We usually approach this in two stages - page performance and customer performance. As you have partially done the second stage earlier, it will be much easier to look at data now, adding more metrics to the mix.
For pages, we'll look at two sets of values - general and behavioral. General metrics include total sessions, landing page sessions share, bounce and exit rates, size and load times, etc. It’s okay to come up with a mix that can be more or less detailed, depending on the needs of your current project. Behavioral metrics are things like previous and next pages, the rate of CTAs within the average fold, and others.
For existing users, we are looking for:
Average order value
The average number of orders per segment
Congratulations! Now you're finished with data collection and it’s time to move towards more fun things like looking for trends and building your first projections.
By building projections, we compare the current performance of any given dataset - user group or page to a baseline of our choice. It can be an industry standard, an average based on third-party data, or a gap between the worst and best-performing segments. In some cases, you might want to separate your projections into groups or segments. So, in this step, we are trying to describe as many improvement possibilities as possible while keeping them:
Based on actual data
For example, a technical projection would be something in the lines of “It is possible to decrease the load time of XX pages by XX second." A behavioral projection would be something like, “It’s possible to decrease the bounce rate for new users landing on XX entry-level product pages by XX%." Our recommendation is to have 3-5 projection groups with 1-3 projections each.
When your projections are finished, it’s time to use them as grounds for building hypotheses. A good hypothesis tells you three things:
What you want to achieve
Why that goal was selected
What you want to do
For example: To decrease homepage load time by 1-2 seconds, reaching industry performance baselines, we will optimize images and replace unnecessary apps.
Usually, it's enough to have 1-2 hypotheses per projection group as you will probably want to control how many you’re testing at any given moment - launching too many simultaneous experiments is not encouraged.
Lastly, ensure that you have proper tracking set to catch necessary data and hit 'play.' Set a specific timeframe for each hypothesis and try not to extend it.
Check Actual Impact vs. Expected
Just in case you don’t get enough actionable data - that might also be a data point on its own. As soon as the experiments finish, assess the actual impact vs. your expectations - that way you’ll always be clear about the goals you set and the means you use to get there.
This seems obvious, but now it’s time to adjust based on what you learned!