E-commerce User Journey & Conversion Diagnostics

Where does on-site conversion break down?

A behavioral case study across browse, intent and purchase signals in a nine-day Taobao observation window.

100.1MAnalyzed events

988KActive users

4.16MItems

9,437Categories

2017-11-25 to 2017-12-03

Asia/Shanghai

Historical Taobao public behavior data.
Portfolio case study.

Historical public dataset · Descriptive analysis · No causal claims

User structure

Users buy, but intent still leaks.

68.1%

Observed Buyer Penetration

User-level participation within the analysis window. This is not a site conversion rate.

25.9%

Intent, No Purchase Users

Favorite or cart behavior appears without an in-window buy event.

Browser Only59,5756.0% of users

Interested, No Purchase256,01225.9% of users

Buyer228,54623.1% of users

Multi-buy-event Behavior Proxy443,85844.9% of users

Favorite and cart are parallel intent signals. The analysis does not force users through a linear funnel.

Journey diagnostics

Cart signals stronger intent — but conversion is delayed.

1.40%

View → Later Same-item Buy

Later Same-item Buy

4.24%

Favorite → Later Same-item Buy

Later Same-item Buy

6.06%

Cart → Later Same-item Buy

Later Same-item Buy

19.41h

Median Cart → Buy Lag

Observed among later same-item buyers.

These are window-observed same-item later-buy proxies. Missing an in-window purchase does not prove permanent abandonment.

Corrected category cohort

Traffic alone does not guarantee conversion.

Category popularity by same-category later-buy rate

Conversion Audit Candidate

2355072

361,785 viewers · 2.64% later-buy rate

Core Performer

4145813

368,450 viewers · 6.55% later-buy rate

Smaller-scale Exposure Test Candidate

3880463

10,343 viewers · 16.27% later-buy rate

Anonymized category IDs only. The matrix uses Viewer → Later Buy Same Category Rate and a minimum of 10,000 viewers.

Decision support

From behavior signal to operational experiment.

Signal

Hypothesis

Test

Success metric

Cart intent + long lag

Reminder timing or checkout friction

Cart reminder holdout test

Incremental same-item purchase vs holdout

Favorite without later buy

Delayed consideration

Favorite-user re-engagement test

Incremental purchase vs holdout

High traffic + low category later-buy

Conversion friction

PDP and recommendation audit

Viewer-to-later-buy lift

Low traffic + high later-buy

Exposure opportunity

Controlled recommendation exposure

Incremental qualified purchase

The goal is not to push every user harder.
It is to identify where intent deserves the next experiment.

Historical 2017 dataTaobao, not a cross-border platformNo price, GMV, revenue, or marginNo order IDNo session IDNo causal inferenceNine-day windowWindow-based abandonment may include later purchases outside the dataset