Develop a seller score algorithm for Walmart Marketplace.

Question

Develop a seller score algorithm for Walmart Marketplace.

Asked at Walmart in the past year

887 views

Asked at

887 views

How to answer Technical questions

Interview Guide

Top Technical interview questions

Imagine you're the product manager for Facebook Marketplace. Since many sellers don't mark items as sold, what existing functionality and metrics could you use to determine whether an item has likely sold?7 answers | 20.9k views
What happens when you enter a URL in your browser?6 answers | 10.8k views
How does TinyURL work?5 answers | 317k views
See Technical PM Interview Questions

Invite members
Invite by email

Add another

Mohit Khatwani · Answer 1 · 2024-05-04T10:42:34+0000

Develop a seller score algorithm for Walmart Marketplace.

Some clarifying questions

What is the objective of seller score algo ?
- The score will help us determine how genuine the seller is ?
Ok if we find that out how will we help the seller ?
- We will add a tag called Genuine Seller/ Walmart Recommended
The final objective is to increase engagement or revenue
- To begin with Engagement
Geo - india
No contrainst to the project
Clarfication - Sellers can come down and list thier products and users can buy that from walmart site.
Do we have a programme already in place for rating sellers ?
- No we dont
Why are we doing this seller score algo ? what is need for it ?
- Lot of users have been complainin that some products were not of good quality
- so we should have a mechanism to evaluate the seller performance not just by sales but other factors too

Let me think through this and come back

for the sake of this exercise i will follow below structure

Mission, type of users, Pain points, Solutions, Sucess metric , does this soud ok?

Mission - To make shoping experience and easy and simple

Type of users

Seller
Buyer

Because we are focusing on sellers main focus group would be seller but buyers would esentially play a role in scoring algo so we will consider both user types

Sellers

Type	Reach	Impact
High Revenue	Medium	High
Mid Revenue	Medium	Medium
Low Revenue	Low	Low

Buyers

Type	Reach	Impact
High Freq	Medium	High
Mid Freq	Medium	Medium
Low Freq	Low	Low

I am considering reach which is number of users will it reach and impact it wil create and as we can se High Revenue and High Freq qualfy hence basis that i will pick these 2 users groups , if the scoring model is scuessful than we can extens it to other groups

Factors for Scoring seller model

Same product as promissed delivered - yes /No
No of times seller product delivered is not up to mark 1 time - exemption , more than 1 will result in negative model
1. (defect rate/% - 1% is tolerable limit)
User rating less than 3 or less than 3
Sellers claim to the life of product
1. Claimed 1 years , worked only 6 months

How will we and at what stage will we collect this info ?

Based on past purchase history user will be prompted to rate
- Quality 1 to 5
  - with even 2nd purchase we can prompt the user if the user is tryin to buy same product asking about scoring past experience on quality
- Complaints will be registered agants seller through call center or contact us option from app
  - Life of product claim versus actual
  - No of complaints needs to be more than 2
- User rating
  - Post purchase of product
    - poor rating more than 1 time and in span or gap of 1 month
    - Not to consider ratings which are being given by same user for multiple products of same seller in span if let us say every 15 days (there by safegurarding sellers intrest too)

Pain points(seller)

Type	Depth	Size
Loss of Repuation/Trust	high	high
Impacts Repeat purchase /Revenue	high	high
Results in banning of Account after threshold	high	high

Pain points(buyer)

Type	Depth	Size
Loss of Trust	high	high
Impacts Repeat purchase /Revenue	high	high

Solution

Sol	Reach	Impact	Effort	Phase
1 - Use a AI/maching learning algo which will be feeded with Quality Score, Ratings, No of complaints and Come up with a score Score above 5 being a good score and 1 being poort	H	H	H
2 On getting poor score notfy sellers about the wrong doings	H	H	M	1
3. Sellers which score 4 & above to be shown with Walmart recommended tag	H	H	M	1
4 - Use wallmart recomemded tag as a input in existing recommendation engine to recommend products	H	H	M	1
5 Build a process post idenifying poor score 1 - Build a SOP to walk seller through it and give a chance to better the experience (Examination period) 2. On Failing examination period ban the seller for 6 months	H	H	H	1
6 Use AI to check 1 If there is any plagerism of sellers meta data to run seller down by some users 2 Catch hold of fake reviews like same users posting againt a single seller to impact its repuation	H	H	H	2

Ujjwald · Answer 2 · 2024-07-13T14:50:00+0000

Step 1 - Ask clarifying questions

Just to rephrase, we need to develop an algorithm to score sellers on Walmart market place.

Interviewer - correct

I assume this will be used for visibility on the platform. So, higher the seller score for a product, that seller will show up first.

Interviewer - That will be the use case for this.

Are we looking at any particular category or will this algorithm impact all sellers?

Interviewer - We think it should impact all sellers.

Lastly, I would assume that this score is used to rank sellers with the best sellers (sellers with the highest score) coming right at the top provided the search terms match

Interviewer - Yes. That's correct.

Step 2 - Describe the product

Walmart Marketplace is a virtual market where sellers can place the items that they want to sell, and the buyers buy the items that they are interested in. There are a number of sellers for a given item, so I am assuming that is where this algorithm comes in. One important thing that exists in e-commerce is the buy box, and how a particular seller can capture it. So, these sellers need to stand out in some meaningful way so that they are moved to the top of the search results when a buyer searches for a certain search item.

This algorithm will probably play a part in recommending/bumping up certain sellers to the top so that the buyers can purchase the goods from them. The scoring becomes important because a high score would mean that buyers would buy from them as they will rank at the top of the search results. This will also incentivise the sellers to put their best foot forward.

Step 3 - List the attributes/goals

The important goals here will be :

Help buyers choose trusted sellers with good history of selling on the platform
Improve the quality of sellers overall
Improve average reliability scores of the seller

Step 4 - Develop the algorithm

Some parameters we should consider are :

Number of returns for an item sold by the seller
The customer rating of the seller
Average return % (For e.g. in the fashion category there are returns, but if the average return rate is 10% and the rate for a seller is 20%, something could be wrong there)
Price being sold by the customer vs avg price being sold for the item in the category
Availability of items with the seller (we want to flag a seller if they have many items that become unavailable suddenly)