Using AI/ML, how can Amazon detect and prevent fraud done by retailers listed on Amazon e-commerce website

Amazon is the one of the successful e-commerce platforms in the world

Clarifying questions

-          I am the PM at Amazon à right

-          Amazon is a global company, I will build the solution keeping Indian geography in mind that could be scaled further à ok

-          Some of the frauds that I can assume are

o   Listing price higher than Mrp

o   Showing discounts / offers in limited deals

o   Selling counter feit items

o   Fake reviews

Do you have something specific in mind that as an e-commerce platform we should tackle head on à Let’s say we need to tackle the fake review

Amazon vision

-          To be earth’s most customer-centric company

User segments (affected by this)

-          Retailers

o   Get undue benefit

-          Customers

o   Cheated

-          Amazon

o   Lesser trust and customer-experience

Pain-points

Retailers grabbing fake review

1.      Identification of fake reviews

2.      Customer resolution

3.      Seller intervention

Point 2 and 3 are already in place and what we need to refine further is the identification of these fake reviews and highlight as quickly as possible

So, I would like to prioritize the identification of fake reviews part

Solutions

Amazon already has pretty strong algorithms in place to detect fake reviews. Suggesting few solutions adding on to what Amazon might be already doing 

1.      Plagiarism review checker [ Impact – M | Effort – L]

a.      Checks internet for similar feedback

b.      Estimate plagiarism and highlight the retailer

2.      History scrubber [ Impact – M | Effort – M ]

a.      Explore the history of the customer and retailer

b.      Check for the feedback given by the user and retailer’s feedback for similarity

3.      Authenticity rating [ Impact – High | Effort – H ]

a.      Scrutiny list à recent reviewer

b.      Reviewer level for genuine feedback based on the feedback and tenure on Amazon

c.      Match the return rate, complaints vs feedback

d.      Give reviews authentic rating and based on A/B testing decide to show it on platform

NSM

-          % correct prediction

Other key metrics

-          # products have fake reviews

-          # retailers identified forging fake reviews

-          # fake review profiles

Guard rail metrics

-          # of false positives product