As a PM at Amazon, if returns and cancellations have recently peaked, how would you go about identifying the root cause?

Returns and Cancellations have recently peaked in Amazon.

Since when is the issue observed? Last one week

Is it gradual/ sudden? Gradually and then peaked

Is it specific to any geography? India

Is it specific to any item category? Not sure

Key metrics impacted -

Cancellation rate -

Do we know these are buyer initiated or seller initiated cancellations? Buyer

For this discussion , would like to limit the core business to e-commerce as the returns/ cancellation are msot common in this business as compared to AWS and advertising

Possible flows -

Order placed -> Order confirmed -> Order processed -> Order in-transit -> Order delivered [No returns/ cancellation]

Order placed -> Order confirmed -> Order processed -> Order in-transit -> Order refused

This typically happens when buyer wants to cancel the order, but the order is already in-transit

Order placed -> Order confirmed -> Order processed -> Order cancelled

Need to evaluate the cancellation reasons

Order placed -> Order confirmed -> Order processed -> Order in-transit -> Order delivered -> Order returned

Need to evaluate the return reasons

Do we have any data on the returns and cancellations? Specific to the split of reasons so as to identify the top issues

Possible hypothesis -

• Buyer related -
  • Got a better price elsewhere
  • Purchased a wrong item
  • Fit issue
  • Color/ Variant issue
  • Undue advantage of the returns policy
• Seller related -
  • Inaccurate product images/ description
  • Wrong item sent
  • Processing was delayed
  • Stock out scenario
• Ops related -
  • In transit delay
  • Package theft/ lost/ damaged
  • Address navigation issue

• If it's a product description issue -> 
  • More stringent QA
  • Flag high return skus
  • Train the seller on the importance of accurately updating the product description in the website
• If it's a delay related issue ->
  • Notify customer and remove sla promise
• If it's a fit related issue ->
  • Add size chart, AI recommender for size in the PDP page
• If there is a high return rate from a specific buyer ->
  • Flag the skus
  • Discuss with customer and check the genuineness of the issue
• If it is related to 3P seller frauds ->
  • Blacklist the seller and remove all their listings

(I only answered the topline bold question and answered my own clarifying questions so a bit more generic)

Context setting
Amazon is a company looking to connect customers across the globe with the products they are looking for from a diverse catalogue of sellers and vast assortment. Customer satisfaction is a c central goal to growing Amazon's footprint and spikes in returns and cancellations may indicate reduced satisfaction so it is imperative to get to the root cause.

Clarifications:
1. I will assume this is order returns and order cancellations (as opposed to subscription cancellations)
2. I will assume the peak is in the NA market (as in NA buyers) and fairly uniformly
3. was this subscribe and save orders or all orders? all
4. What is the recency window? 3 months
5. Have they steadily. increased and now plateaued at a peak? yes
6. Peak for how long so far? 2 weeks
7. So at this timeframe it is not likely resultant from a recent holiday such as Christmas or Black Friday? assume no

Restate problem

In the last three months order returns and cancellations from NA buyers have steadily increased to a peak which has been sustained for a couple weeks. This is not expected to be attributable to a holiday event but the root cause is as yet unknown.

Structure:

Internally controlled factors
1.customer
a decreased order satisfaction due to various factors such as colour shape size function 
b customer ordered more than what was needed with intent to return some items
c customer unsatisfied with order processing delay
2.business
a platform not connecting searched to the right assortment leaving customers wanting more choice
b value proposition not attractive for new sellers leading to stagnant assortment in some categories
c return and cancellation policy nudging undesirable customer behavior
d shipping policies leading to delayed seller shipping
3.seller
a poor item description
b product does not match the description
c sellers shipping damaged or wrong items

Externally controlled factors
1.customer
a customer found items cheaper or more attractive elsewhere
b customer unsatisfied with shipping delay or undelivered items (from contracted or local shipping)
2.business
a contracted logistics teams not meeting SLAs for delivery
b pricing and assortment is not competitive on the market
c competitor platforms are gaining market share
3.seller
a not enough new sellers onboarding
b poor seller customer service

I would identify the root cause from the above list of potential causes (assuming the above list is exhaustive- I would validate this). Not that any one of these root causes would have a number of likely more granular root causes so it may be necessary to do an issue tree or repeat root cause investigation.

I would assume that the internally controlled factors are within the control of the business and have a policy, strategy or technical solution available with internal resource allocation which may be important in prioritizing.

Then my solution would be
1. I would identify segments, sellers, categories, local markets (ie. rural or urban) and other factors that may be causing the peak by investigating distributions of these variables in a descriptive analysis
2. I would narrow my list of root causes ruling out irrelevant ones and assess effort and impact where more than one root cause needs prioritization.
3. Investigate the root causes at play with respect to the key variables from the descriptive analysis. For example if a customer found more attractive product elsewhere we may see that certain categories are not well represented or if there was a wide problem of seller misinformation there would be a a segment of sellers yielding higher returns.
4. Once I have the top root causes and the descriptive data to support I would craft an approach by diving deeper into the issue and listing out the use cases.

For example imagine that sellers from France have poor information provided to their NA market. We see in the data that there are increased wrong item returns from French sellers. In this case I would probably need to do another root cause analysis to determine what the driver of widespread poor product descriptions is like translation software or other. But then I would have

use case one: build GenAI to identify content gaps
use case two: build seller facing tool
use case three:
use case four:
...
use case n:

Perform a prioritization base on effort and impact.

repeat for each root cause.

As a bonus I would also probably tie back to an enterprise strategy ie. what if Amazon aims to have 30% of all product pages assessed by GenAI tooling or aims to have X number of sellers from any representative country etc. I'd also think about things like whether investing effort in a problem tied to a locale is aligned with the financial strategy of the enterprise. Anything that brings a global view.