Explore

Whatsapp chatbot PRD

Objective -

Reduce lost orders due to restocking delay for users driving overall order growth

Make restocking easier through a interactive Whatsapp chatbot

Success metrics -

Restock rate% (TBD)

IHAO% of suppliers (TBD)

Channel adoption - #Recos interacted / # Total Recos

Problem Statement -

Understanding effectiveness of Whatsapp chatbot to support multiple supplier use cases -

Inventory recommendations (First use case)

Interacting with price / product recos

Supplier support

Scope - The goal of V1 is to allow Inventory management via Whatsapp chatbot through proactive notifications as well as supplier initiated conversations.

Bot flow -

Supplier initiated conversation -

⁠

Suppliers initiating the conversation have the ability to view

Currently OOS recos

Low stock recos

The next set of actions are identical for the recommendation types and encapsulates -

Updating stock for the reco

Updating stock for other siblings / variations of the recommended PID

Reject reco

Capturing the rejection reason from supplier with vSKU details

Viewing more recommendations (Limit TBD)

Proactive notification (twice a day) -

The top OOS recos to be shared twice a day per supplier (TBD)

The notif will have the similar catalog card structure and quick action buttons as seen in the case of supplier initiated flow

Action button responses will trigger bot says directly as per bot definitions

Backend API builds -

Proactive notifications (Interakt) -

Dynamic user addition on Whatsapp business account / Interakt

Use case - Adding / Updating users through backend DB as and when new suppliers / parameters are generated

API doc -

https://www.interakt.ai/resource-center/api-doc#How-to-add-update-User-Details-on-interakt-via-User-track-APIs⁠

⁠

In case a user had originally been added via the interakt dashboard (and not by calling the User Track API),then, no userId would exist for that user. In that case, you could either -

Call the User Track API without specifying a “userId” (and only specify the “phoneNumber” & “countryCode”), or,

Include a userId in the API call, which will then get added for that user, and you could use that userId to reference that user in future API calls

Dynamic variable setup

Use case - Adding dynamic variables / images in the message template which is unique per supplier and keeps changing at a timely frequency

Event track APIs can be setup here with the required variables <

https://www.interakt.ai/resource-center/api-doc⁠

We can send variables which are configured in template from request of event track API itself Eg.u can use request body like -

{

"userId": "12035448-36a0-3aa24",

"phoneNumber": "9999999999",

"countryCode": "+91",

"event": "product_out_of_stock",

"traits": {

"Catalog ID": "1",

"SKUid": 2,

"Variation": 500,

"Net order loss": "44"

"createdAt": "2020-11-05T13:26:52.926Z"

}

Whenever the product is out of stock for the supplier, we can simply call the event track API and pass the data, and it will trigger the notification as we have set the ongoing campaign for it with a specified event

Whatsapp consent tagging

Use the current Whatsapp consent tags to be sent to Interakt to identify the suppliers who’ve accepted receiving updates from WP

Chatbot API builds (Haptik’s conversation studio) -

Serving inventory recommendations to suppliers when they initiate conversation / try to view more through post viewing the proactive notification

Suppliers would have a toggle to view more recommendations

Introduce counters / pagination in the recommendation feed to serve out recos incrementally as per input

We can retrieve supplier’s phone number from Haptik by defining user_data in code node for configuring API calls

Updating inventory value and reflecting that to inventory service

Suppliers will have the control to directly update stock through Whatsapp

Recommendation vSKU details can be fetched across nodes by configuring conversation details in the code node

Inventory updates through chatbot need to be reflected in logs for attribution

Update source ‘Whatsapp_chatbot’ in inventory_change / inventory_product_inventory_log_v2

Capturing rejection reason and relaying it back to DB

This would help filter out the rejected recommendations from future reco feeds

Displaying live stock values for other variations of a selected recommendation

Suppliers would have the functionality to update stocks for other variations of a given recommendation

Proactive recommendation notification setup

Leverage current CRON responsible for inventory Whatsapp notifications to build the recommendation feed logic here

Backend details

Both type of recos - OOS and low stock will have separate Spark / Presto model tables with the following schema -

Supplier ID (INT) - Both recos

Product ID (INT) - Both recos

Style ID / VPID (VARCHAR) - Both recos

Variation name (VARCHAR) - Both recos

Catalog ID (INT) - Both recos

Catalog name (VARCHAR) - Both recos

Sorting score (INT) - Both recos

Current stock count (INT) - Low stock recos

Days on hand (INT) - Low stock recos

Estimated order loss (INT) - OOS recos

Frequency - Thrice a day (TBD)

Mode - Overwrite

Model to be setup on Presto and output to be dumped in S3 to take care of any scaling issues

Key callouts -

Conversation window on Haptik’s chatbot is applicable for 8 mins post which the session gets reset

Conversation window in case of proactive notification through Interakt - TBU

We’ll first test this with Haptik’s staging account on a test WP number

Tentative items to be picked up in the next version -

Introducing reminder nudges for inventory communications

Integrate price / product reco capabilities

Design a FAQ / Support section as a separate flow

Design prototype -

meesho_flows_for_suppliers_10march21__1_.pptx⁠

⁠

Resources -

⁠

Haptik⁠

⁠

Interakt⁠

⁠

Want to print your doc?
This is not the way.

Try clicking the ··· in the right corner or using a keyboard shortcut (

CtrlP

) instead.