Predictive Analytics and Fit Finder

I stumbled upon another interesting application of predictive analytics, The North Face’s Fit Finder.  On their product detail pages, click on the Fit Finder link and it will open a modal with a series of forms.


Height + Weight = Waistline

They ask for your height, weight and your preferred fit (tight or loose).  After you save these values, they give you a recommended size.


Now I work for an ecommerce website and one of our main issues is getting our customers to order the right size of apparel.  We do this by giving them the measurements of chest, waist, and hips that best fit a specific size.  Height and weight don’t usually factor into measurements – or so I thought!

I did a little research and found that there is a 2012 statistical study predicting waist circumference from one’s body mass index (BMI).  The body mass index is defined as the weight in kilograms divided by the height in meters.  In the study, they developed linear regression models to determine waist circumference as a factor of BMI.

The North Face is known for their jackets and outerwear so if you know the customer’s waistline, you can guess which jacket would fit them.


The next form asks if you have a flat, average or curvy stomach.  If you are not sure, it asks for how wide your hips are.  They adjust their measurement based on this added information.  They probably fit your size based on ranges and if you are curvy-er – they fit you in the upper range of the predicted size or one size bigger even.


For products made for women, they even ask for bra size.  This gives them chest measurement to find the right jacket fit.



Then they ask for your age.  They explain that your age has an impact on how your weight is distributed.  They are still adjusting your measurements based on added information.  I think age could be a co-variable to height and weight in the linear regression model as well.


Historical Data

Based on all the data given, they compare them with historical customer data.  They look at their returns data. How many similar customers who ordered the size they recommended returned the product?  They display the percentage of those who did not return the size to you.

If a majority did not return the product in the size they recommended then you feel more confident ordering that size.  And if their prediction model works, you will be added to the statistic of customers who did not return the recommendation.



I actually like this Fit Finder more than ours.  Height and weight is easier to remember than all your pertinent measurements.  And most of us remember how old we are so that would not be an obstacle.  The visual way the customer can indicate their waistline or hip curves is very user friendly.

Using historical customer data will help fine tune the predictive model.  This is especially true if your customers belong to a niche market who tend to have similar body shapes like dancers.

North Face did a great job on this application. Making predictive analytics improve user experience.  A great solution to every online apparel store’s problem of minimizing returns due to the customer’s incorrect choice of size.

This solution is that beautiful intersection of big data and UX that I keep talking about!