Explore

Kerala Ayurveda — Product Analyst II Assignment

Dataset: GA4 clickstream Scope: D2C India website (web funnel analytics) Expected effort: ~6-8 hours Deadline: 18th Jan Submission method: Link-only (no attachments)

Background

You’ll work with GA4 clickstream data captured on our India D2C website. The dataset includes event-level touchpoints from session start through purchase. Your goal is to (1) build a clean, reusable funnel model and (2) produce an MBR-style narrative: what happened, why it happened, and what to do next.

⁠

1) Dataset structure (what you’re getting)

Files

A ZIP containing a folder of parquet files containing event-level data of one month captured by GA4 on

keralaayurveda.com⁠

website. Download

here⁠

⁠

Base event columns (high level)

User/session

user_id (GA4 pseudo user identifier)

session_id (session identifier)

Event

event (event name)

event_ts (timestamp)

date_ist, time_ist

Page / context

page_location (URL)

page_type (categorized page type, if available)

device

Parameters

event_params (GA4-style nested list of key/value pairs)

event_params (nested GA4 params)

event_params is a list/array of key/value objects (not a flat dict). You will typically need to flatten/unnest it into columns to use it.

Common keys you may see (not exhaustive):

Attribution: source, medium, campaign, term, gclid (often on session_start)

Commerce: transaction_id, value, currency, payment_type, coupon, shipping, tax, discount (often on purchase)

Funnel events (relevant set)

Session start

session_start

Product view (choose definition)

view_item

view_product_page_loaded

Add to cart

add_to_cart

add_to_cart_custom_event

Checkout start

begin_checkout

gokwik_checkout_initiated

Checkout progression

add_shipping_info

add_payment_info

Purchase

purchase

Not every journey is linear; events can be missing or out of order. Handling this sensibly (and calling out QA findings) is part of the assignment.

⁠

2) Your task

Part A — Build a reusable funnel model (session-level)

Create a session-level table/dataset session_funnel across the full month.

Required fields

Identifiers

session_id, user_id

Dimensions

device

source, medium, campaign (from session_start params; null/unknown allowed)

landing_page (first page_location in the session)

Funnel steps For each step, include:

flag (0/1) per session

first timestamp (optional but encouraged)

Steps (use OR logic where there are variants):

product view (view_item OR view_product_page_loaded)

add to cart (add_to_cart OR add_to_cart_custom_event)

begin checkout (begin_checkout OR gokwik_checkout_initiated)

add shipping info

add payment info

purchase

Purchase outputs

orders (distinct transaction_id preferred; state your approach)

revenue (sum of purchase value, with deduping logic as needed)

AOV (revenue / orders)

Data QA (must include)

A short QA section with checks like:

duplicate purchases / duplicate transaction_id

sessions with purchase but no checkout events

null spikes in source/medium

any other anomalies you notice

⁠

Part B — MBR-style analysis

Write a concise memo + supporting analysis.

Required analyses

Monthly KPI snapshot

sessions, users

product view rate, add-to-cart rate, checkout-start rate

purchase CVR (session → purchase)

revenue + AOV

Funnel performance

overall step conversion rates

break down by at least two cuts, e.g.:

device

source/medium

landing page group / page_type

Driver diagnosis (surgical) Compare:

Week 1 vs Week 4 or

First half vs second half or

Best week vs worst week

Explain what moved and why:

traffic mix shifts (source/medium/device/market)

step conversion changes (view→ATC, ATC→checkout, checkout→purchase)

3 recommendations Prioritized actions (product/growth/tracking), with:

expected impact (direction + where it helps)

how you’d validate (experiment, tracking check, next analysis)

Output format

Memo: 2 pages max

Supporting: 3–5 charts/tables (can be in notebook)

⁠

3) Submission instructions (Link-only)

Please submit a single link to a folder that contains your materials.

Allowed: Google Drive / OneDrive / Dropbox / Notion / GitHub repo (public or private with access granted) Not allowed: email attachments

Folder contents (required)

README (how to run + assumptions + QA checklist)

Notebook (.ipynb) or SQL scripts + run instructions

Memo (pdf or md)

Outputs (session_funnel as csv)

Naming convention (important)

Name the folder: KeralaAyurveda_ProductAnalyst_<YourName>

Link permissions

Make sure the link is accessible (viewer access at minimum)

If it’s a private repo/folder, grant access to: nishant@keralaayurveda.biz

What to include in your email response (copy/paste template)

Subject: Product Analyst Assignment Submission —

Body:

Submission link:

Approx time spent:

Any assumptions / known limitations (3–5 bullets)

⁠

4) Evaluation criteria

Accuracy & definitions (30%)

Driver identification (25%)

Business usefulness (20%)

Communication (15%)

Pragmatism in chaos (10%)

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

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

CtrlP

) instead.