One of the most common comments we hear from restaurant CMOs/CCOs is “we test, but our tests aren’t as predictive as they should be”.  Here’s a case in point.  Per the Wall Street Journal, Burger King is running out of ribs for their LTO.

Just three weeks ago, we wrote that BK Ribs was an idea worth testing, despite a lot of skepticism from industry analysts and food bloggers.  Well, what’s worth testing is worth testing well.  A scientifically designed, rigorously analyzed test is predictive.  Specifically, it predicts the two critical things you need to know for this type of program:

1. Should I do it?
   • Will it work?  Can I reliably predict which products will be successful (based on incrementality, not just mixing), net of halo and cannibalization effects, so I can go national with the right product set?
2. If it will work, how do I plan for success?
   • How many units will I sell, by location?  How do a plan for LTO length, media support, and product supply to make sure I can take full advantage of the success?

Burger King actually did test the rib offering in Indianapolis, Las Vegas, Greensboro, and Orlando.  But the testing obviously didn’t do its job.  BK underestimated the national demand.  It sounds from the Journal that this was driven by underestimating the sales of larger piece orders.

There are many reasons why restaurant tests aren’t predictive:  poor test group selection, not enough stores in the test, reliance on comps without good controls, control group matching techniques that don’t net out enough noise, lack of tools to predict how test results flow to rollout results.  All of these challenges can be addressed with the right analytics.

And then you won’t run out of ribs.

Sign up for APT’s monthly Food for Thought newsletter