HOW IT WORKS

ONLINE KIDS' JEANS SALES:

"A-B SPLIT" TEST...

Weekly sales at the onset of an economic downturn had dropped alarmingly
New graphics were introduced showing the product in a different way
Online shoppers could scroll down to see all the product styles (not shown here)
Two options for the new graphics were tested in an “A-B split” (shown right)
Graphic option “B” is best, but there’s still a sales decline of 236 pairs of jeans per week

... AS HALF OF A 2x2 TEST

Turns out that the best graphic option depends on whether a scroll or “clunky” layout is used
“Clunky” required the shopper to click through several small pages with a few styles on each
No-one had expected “clunky” to work
In fact: “clunky” sells more kids’ jeans, when combined with graphic option “A”

The next case below (glimpsed on our home page) a healthcare insurance client reduced acute hospitalizations for thousands of Medicare patients We asked 35 in-house nurses and physicians: "What might prevent admits?" They said: "A through S" (below).

We tested all 19 in the live business with nurses randomly assigned to the test matrix below (where "1" means test and "0" means don't) so you can see 19 A/B splits. Surprisingly, sample size was the same as for a single A/B split.

STATISTICAL DESIGN and CONTROL pre-test which proves the test is on a "level playing field"

Randomization gives cause and effect directly for A-S. 4 things reduce hospitalizations (e.g. A by -200.7)

STATISTICAL CONTROL GRAPH PROVING THE STATISTICALLY DESIGNED TEST CAUSED A 17% REDUCTION IN HOSPITALIZATION