You are on page 1of 2

TAILORED CLOTHING RECOMMENDATION ALGORITHM

Before any recommendations, app should:


Loop through potential recommended items and make sure:
1. They are not disliked
2. They are not on the wish list
3. They are for users set gender
Then:
1. Check for closet items, country sizes and profile sizes.
2. If app contains only country sizes, recommend based on country sizes
3. If app contains anything other than country sizes, recommend based on that and not country
sizes.
Algorithm for Country Sizes:
Loop through users country sizes and recommend items that are same as country sizes (ex: M=M and
32=32)
Algorithm for Accurate Closet Items
Loop through users closet items
1. Perform for each closet item individually, not combined or averaged
2. Gather waist size of closet item
a. If no waist size on closet item, proceed to next
3. Loop through spec sheets and find all items with similar waists (up to 1 smaller and 1.5 larger)
to get list of potential recommendations.
a. If no waist size on spec sheet item, proceed to next
4. From list of potential recommended items, compute the following (weighted):
a. ((closet item rise - spec sheet rise)*1)+((closet item thigh - spec sheet thigh)*1.3)+((closet
item hip spec sheet hip)*1.4)+((closet item leg opening spec sheet leg
opening)*1.2)/4 = Value
b. If a value from above is missing, disregard on both closet item and spec sheet, divide by
3 instead of 4, or 2 instead of 4, and so on
c. Disregard items with Value greater than 2
5. Rank items in order of closest waist, ascending Value and most similar inseam
6. If no value exists for given measure, ignore measure on both user garment and recommended
garment
7. When new garment added to users closet, loop through all potential recommendations and apply
steps above and put new items in user discover feed
8. When new garment added to admin panel, loop through all users garments and apply steps from
above, and send push notification if there is a match
9. Do not duplicate item recommendations

Algorithm for Accurate Profile Measures


(Recommendations for Profile Measures will include more items and be more broad than closet item list)
1. Loop through users profile sizes and gather waist size
2. Loop through spec sheets and find all items with similar waists (up to 1 smaller and 1.5 larger)
to get list of potential recommendations.
3. From list of potential recommended items, compute the following:
a. ((profile rise - spec sheet rise)*1)+((profile thigh - spec sheet thigh)*1.3)+((profiile hip
spec sheet hip)*1.4)/3 = Value
b. Disregard items with Value greater than 3
4. Rank items in order of closest waist, ascending Value and most similar inseam
5. If no value exists for given measure, ignore measure from profile and recommended garment
6. When user profile sizes are updated/changed, loop through all potential recommendations and
apply steps above
7. When new garment added to admin panel, loop through users profile sizes and apply steps from
above, if new item is a recommendation, send push notification.
8. Do not duplicate item recommendations
ITEMS RECOMMENDING THROUGH PROFILE SIZING SHOULD NOT CONTAIN THE UI:

*Blend the two lists together in ascending order*

You might also like