How Men’s Wearhouse Could Use Data Science Cont.
After writing the brief piece about How Men’s Wearhouse Could Use Data Science. My team and I continued discussing how we would approach the problem of recommending products to customers for Men’s Wearhouse. We believe that if Men’s Wearhouse could move their in-store customer experience and fashion knowledge(which is great) online, then they would grow from their current 1.65B USD market cap to several times that. Stitch Fix somehow has a 666 P/E ratio and a 2B USD market cap(which is more than Amazon’s P/E). In the same way, Men’s Wearhouse has that opportunity to create a similar or better service that helps recommend products which would likely increase online purchases.
In creating a system to recommend products our first step would be to break up the customers into categories. This could be done using an algorithm like K- nearest neighbor or SVM. This will also depend on if the customers are already labeled in categories or not (which we assume it is not). Thus, this step will require an unsupervised approach first to see if you can develop natural clusters that can be explained.
In the case that this is successful, then the next few steps could be split up and done simultaneously.
One of the important steps would be to calculate the possible value of the customer. This could be calculated by considering past purchasing habits like does the customer come in and buy one product very frequently or do they purchase $2000 worth of products once a year and do these customers come for sales, or do they buy full priced items. They could also purchase 3rd party data to calculate other spending habits and discretionary income of a customer outside of their stores. This would provide the ability to more accurately market the right price point to each customer as well better steer the customer towards brands the customers already know.
Another important step would be to consider which products each person might buy. Will the customer want to purchase a pair of green striped socks, a bow tie, a tie, a new suit, etc. Since Men’s Wearhouse already has a lot of past purchase history they already know what apparel might go with what their customers already own. This would be a separate set of heuristics or business logic that would occur after each group was separated into categories as it would help better fit predictions and get rid of some of the noise that you get when looking at data sets that are too large (one of the things you will notice is we will use this concept of breaking up the data or steps a few more times, because we believe this helps increase accuracy and reduce the noise of complicated problem).
This could either be one algorithm or possibly multiple algorithms to breakdown the problem further. For instance, you might just try to predict which product a person might buy. However, it might be easier to try to create an algorithm to predict purchases that are likely to be replacement purchases, new purchases, and spontaneous purchases. These each might have distinct patterns and features that might get skimmed over when trying to make an end all be all product recommendation algorithm.
This would have to be tested in order to confirm which approach would be better. Once that is decided, then the algorithms which could be anything from a decision tree to a neural network could be used to predict that the next product the person should purchase is a new purple striped dress shirt to go with the grey 3 piece suit and solid purple tie ( I am actually not sure if that works) that is why the system would be helpful for their customers! It takes away the fear of being wrong and provides the convenience of not thinking.
Once the product is decided on, it would be about packaging the ad correctly. We could foresee the ad being placed in email and also in Men’s Wearhouse “Perfect Fit” app. If they could get a good user base on the app, that would be optimal, then every few months have a notification pop up with a list of 2–5 products that the customer might be interested in. Instead of just telling the customer about their shirt size. Create actionable steps for more purchases. At the end of the day, that is their real goal. Might as well be helpful in doing it. This type of system has the potential of removing the stress of knowing what to wear. Maybe it could even recommend what to wear for special events and or every day business attire. The more Men’s Wearhouse can provide a service, not just a product would cause an increase in loyalty and likely in purchase. So if Men’s Wearhouse could scale their in store feel online, we foresee a similar output. There are a lot of ways we could see this kind of data product being utilized to increase customer satisfaction and sales.
Finally, once the ad has been shown to your customer it will be important to track the success of the ad. This is one of the most valuable steps because knowing the outcome will allow the company to create more accurate algorithms and products in the future. If the customer purchases right after seeing the ad online, then it will be very easy to reasonably say when an ad has been successful. The customer might also purchase the product several days later in person or online, so that will also have to be considered. That might require some form of business rule. Maybe if a product is purchased in less than 15–30 days after seeing an ad they could consider it a success. There will be false positives in that range, but again, this would require actual testing.
We are very curious to see how Men’s Wearhouse and other retailers that have been around for a while will start integrating their digital strategies more and more with all the data they already have. The way we see it, there are a lot of old concepts like product recommendations and ad targeting that still have a lot of work that could be done to help increase company’s profits and overall efficiencies. In addition, moving core competencies like in store fashion advice online would be majorly beneficial.
Interested In Reading More About Being A Better Data Scientist?
How To Grow As A Data Scientist
Boosting Bagging And Building Better Algorithms
How To Survive Corporate Politics As A Data Scientist