Any good marketer knows the key to successful campaigns is A/B testing. Its a standard practice that marketers regularly use. A/B testing puts 2 versions of a campaign in front of an audience and then measures which one performs better. Then the winning campaign gets the financial investment. And rolls out. Marketers use A/B testing for pricing tests, websites, ad campaigns, direct mail promotions… In other words – nearly all marketing investments.
Yet, space planning rarely uses it to ensure that the enormous investment in store merchandising is validated by shoppers.
The biggest obstacle to A/B testing in store merchandising is the cost and time to set up the tests. Stores have to be matched for accurate test and control samples. The execution in the test and control group must be maintained. And the time to run the test, measure results and determine the best selection can range from weeks to months. That is often too late to affect the holidays or other key shopping seasons.
Another obstacle is the scale of the test. Changing merchandising can also mean changing operations for stores. For example, place a refrigerated section into pet food as a test. That requires power and fixtures. Plus it changes how to merchandise and rotate the inventory in the pet food aisle. Which affects store members’ tasks.
As a store testing veteran can tell you, another obstacle is reverting test stores to normal merchandising when the test is over. Too often store tests are abandoned in the market. Leaving store associates to deal with their “one off” store for years. Most retailers have abandoned store merchandising tests that are years old still operating in the market.
Retailers with a fleet of stores in the market have the richness of variation. In other words, if the space planning data were mined effectively, they could identify all of the tests they inadvertently already have in the market. Some stores have milk in the back. Others have it on the right side. What does that mean for milk sales? Some stores have a race track layout. While others have aisles regimented front to back. With the right AI, space planning teams could start to identify the patterns in their data and in their store merchandising to evaluate and treat those “in-market tests” as A/B tests. Consider looking at department and category placement within the store, adjacencies and overall space allocation as variables that can be evaluated and measured in your stores as they exist today.
Getting Serious about A/B Testing
The ideal thing to do is to create a formal A/B test process. First, set levels of A/B testing based on impact and cost. There should be differences in testing small brand or merchandising changes within a planogram versus a complete store prototype overhaul. Then create small, medium and large testing protocols. Each test size should have a tiered approval process and must have a measurable test business case.
Here are examples:
|Size||Approver||Duration||Scope||Min Business Case|
|Small||Director||6 months||<10 stores||$50,000|
|Medium||SVP||8 months||11-25 stores||$250,000|
|Large||Capital Committee||1 year+||>25 stores||$500,000|
An example of a small test may be testing private label placement within a planogram.
A medium test might be adjacency changes to the household cleaning aisle.
An example of a large test may be moving the children’s and baby department to the center of the store and adding a new toy department.
Finally, create a team that is the central clearing house for all testing. That includes setting up the tests, tracking their results and coordinating all of the communications to store teams. Once standard protocols are structured, they should be a template that can be repurposed.
A/B Testing In the Future
Expect smart space systems to help identify matched store sets and to locate close affinity stores when setting up A/B Testing. Start to ask your current vendors about their capabilities. Or include it as part of an RFI when surveying the market. For retailers who want to make intelligent choices about how to merchandise their stores, setting up an A/B Testing program may be the key.