Conversion Rate Optimization (CRO) for Retailers. Is it Just Merchandising?
We have been getting some great questions from online retailers about CRO (Conversion Rate Optimization) and how it is similar or different from merchandising.
So we sat down with our CRO Pro Gary Marx to learn more...
How much of CRO work involves merchandising?
For this discussion we are defining merchandising as how to organize retail products (in this case digital) to convert best around specific SKUs, often utilizing Content Management Systems (CMS) and Data Feeds to ensure optimal customer experience.
At RBL we think about Conversion Rate Optimization (CRO) as a prioritized process and system that is definitely not specific to merchandising. We take in quantitative & qualitative data, feedback from all over to inform testing hypotheses. Does some of CRO overlap into merchandising? Absolutely.
So Gary, what do you think?
You're right - CRO is a prioritized process AND there's major overlap with Merchandising.
A mature CRO process at companies like Amazon and others can take a full time team with data scientists, developers, and a Product Manager to lead the task of testing and validating decisions on Merchandising. Naturally, Amazon is Amazon and you are you - so start from a single person or small team and expand as your confidence and skills grow. It's really fun!
Imagine if your merchandising team tested everything the way we test funnels, growth loops, and landing pages via CRO. No brand people (no offense it’s just a different mindset) imagining they know better and holding marketing, sales, and development to opaque decisions -- but data driven brand people collaborating with a team of testers to prove increases in sales due to specific placements and proven decisions.
That's the potential and a lot of what Whole Foods has done on their shelves with their data. This is why they move products so much more efficiently than other groceries and why a wonderful startup in Oakland called Bedrock Analytics holds such a promise in the CPG space - a world which does not really utilize the data available.
It's why Amazon eats every online retailer's lunch, breakfast, dinner, afternoon snack, and dessert.
So how would someone execute CRO for Merchandising?
Great question - One word: Segmentation
There are different levels of maturity in any CRO program.
In short, you start out with one version of one page, then as traffic increases, you segment users at a first level (geolocation, device, etc.) then you continue to increase your segments and divide further as you understand what users are worth what real and potential $$ and dial in on increasing the real bucket for each segment.
Here’s where the CRO process combines with Merchandising - you increase how customized the product display is per segment as you discover what converts better through testing. This will result in many segments of users and a constantly tested product display that gives weekly (if not daily, pending traffic and sales) feedback to the PM (Product Manager) or CRO lead to make data-driven decisions.
All of it revolves around CRO - get the qualitative data from users, PM's, salespeople, then test it all through A/B testing via segments.
Sure, that’s great in theory, but how does it work in real life?
I've done tests that required apparel photoshoots that tested different model personalities (mood, facial expression), backgrounds, models (skin color, diversity, hairstyles), etc. sell better (they do and it depends on your audience).
That’s awesome and similar to how we take a rigorous approach to testing nearly every impactful variable in paid marketing: headline, image, call to action, even test matching images of people to their digital targeted counterparts.
Totally! In this apparel example, we ran a small photoshoot on a subset of shirts and pants, using the most popular ones and then ran an A/B test to measure what performed differently. A set of new photos resonated better with the audience and lifted sales!
The client saved money and increased sales because we did a sample test run of multiple photoshoots, got rapid feedback from the testing, then did one major photo shoot to change all products to the new environment with the models having the new expressions & natural background.
We saved time by NOT doing 5 major photoshoots and increased sales by showing customers the RIGHT one -- with the benefit of adjusting the brand guidelines with customer actions/feedback. To this day, that is how this client’s website shows their products and it’s carved a unique corner for their brand to own. Apparel is a fun industry.
What about this one, how do retailers show a shirt to Stephanie in 75 degrees Santa Barbara but a pullover to Peggy in 55 and raining Portland?
Segmentation based on Geotargeting, and potentially a subset of female visitors - you absolutely should show different locations different things based on different seasons, weather patterns, etc. The Weather Channel does this with Optimizely fantastically - and super complex. Merchandising needs to figure out the things that drive their buyers to different products (qualitative feedback) then validate that information with testing.
If a retailer wanted to sell to both Peggy & Stephanie, I would run a test that identified the geolocation of all users in Portland or Oregon - we are talking Portland, Oregon, not Portland, Maine here - depending on customer feedback about the city or the state sharing Peggy’s interest. The rural area outside Portland may not share Peggy’s stylish Portland tastes. Portland has over 600,000 residents, so not much concern about traffic being low in the city.
Then, I’d have another test that segments by geolocation of Santa Barbara - a beach town of roughly 90,000 people. This is where traffic considerations come into place - normally, women are about half the population, so forty-five thousand. Then, how many of those are in the adult age group for the shirt and how many of those people love this client? We may need to increase the traffic of this segment to hit statistical significance in a meaningful time period by running paid marketing to feed us more users in that region.
Are there ways we already see this?
Two examples include 'purchased with' or 'most people who buy your cart item also buy…’ They’re somewhat simple (I believe Shopify has plugins for that and ruby has gems to help with it) and you can either do it manually or have the code continuously analyzing purchases for commonalities.
When you have low traffic levels or are unsophisticated in this approach:
+ a 'best sellers' list or carousel is usually shown - it’s the same list for everyone
+ simplistic device type segmentation (iPhone vs. android)
+ no difference in return visitor states
And all of those are GREAT starting points - but test them to ensure they resonate with your visitors. They could turn them off or distract them from existing designs that work really well, That’s the wonderful thing about testing - we can be certain what works.
I hope that helps!
For top retailers thinking about CRO and Merchandising there are opportunities to collaborate to get maximum results. Building a holistic system that considers both qualitative and quantitative data ranked intelligently against the business’s key objectives will get you there and ensure that conversion rates for various customer segments are strongly considered to drive the best possible ROI.
Gary Marx is a CRO lead for Round Barn Labs and has run thousands of A/B tests for brands like CNN, Disney, and Singularity University.