Modelling Magazine Choice to Optimise Layout Elements
First Published August 2008
A visual experiment on 200 possible layouts discovering a ‘best’ 10 times better than worst
This choice modelling experiment showed how people choose a magazine and has set the magazine cover style since 2010. Actual results are reproduced below with kind permission of NewsCorp Australia.
Gardening Weekly is just one of over thirty magazine titles produced by News Corporation Australia but was failing in the market. In an effort to boost magazine sales, a redesign of the cover was proposed. The designer presented a variety of concepts that included elements from the existing layout and website as well as some new ideas. Naturally there were many opinions about “what works” SurveyEngine was hired to help put some data into the discussion.
Initially, NewsCorp wanted to simply market test the handful of concepts produced by the artist but SurveyEngine proposed to test all of the combinations of ideas to explain what elements drive a consumer’s choice.
How we did it?
The artist produced four general ideas on revitalizing the existing design for the choice modelling experiment.
The designer suggested the masthead could be solid white as it had been or ‘bleed-through.
There were three possibilities – uppercase, lowercase, and a cursive script style
Four options were presented – a side feature layout, two horizontal feature styles as well as ‘no features’
Eight backgrounds were suggested – a number of flowers, a garden, a bowl of berries and a gardener.
These raw ideas, were developed as an experiment within SurveyEngine which reported that there were exactly twenty-four test layouts that would be needed for the experiment. This list was given to the artist who produced finished artwork that was attached to the experiment.
The experiment was sent out to 1,000 people appearing as a regular online survey. Once enough data had been collected, the results were automatically modeled.
What is the best layout?
We can bypass all the mathematics and find the best layout by picking the most positive effects with a choice modelling experiment.
Predicted Choice 93%
The optimal setting for each of the attributes is a:
- See-through masthead
- Simple uppercase font
- Background of a field of lavender
- Simple article layout
- With a predicted choice 93% compared to the worst.
The cover on the left was created from the above optimal setting. Before the model, it had never existed and it was never directly tested, it was predicted.
The cover on the left was created synthetically from the above optimal setting. Before the model, it had never existed and it was never directly tested, it was predicted.
…and the worst
Predicted Choice 7%
As an exercise, the worst layout was also created. This time all the most negative effects are picked, which are:
- Solid masthead
- Script font
- Picture of a gardener
- Side layout
- With a predicted choice of 7% compared to the best.
This means the optimal would be chosen between ten and twenty times more often than the worst.