Product listings
150
5 public index pages
PRODUCT ANALYTICS · MULTILINGUAL NLP
Mapping museum collectibles
and testing review models
150 marketplace listings reveal price and assortment structure. 500 independent reviews test topics, sentiment, and model reliability.
150 listings500 reviews180 audited labels
A
MARKET DASHBOARD
Product Landscape
Business questionWhat does this market sell, at what price, and in which product forms?
Data used150 public Taobao/Tmall listing snapshots; sales shown for 78
You will learnA market map, five exploratory product groups, and three decisions to test
Median price
CNY 59
137 products ≤ CNY 100
Product formats
6
Figurines and DIY lead
Exploratory segments
5
Minimum cluster size: 18
01 · Price map
Where does the market compete on price?
Most of the assortment is accessible rather than premium: 137 of 150 listings are priced at CNY 100 or below.
02 · Product groups
What kinds of products appear together?
Five exploratory groups summarize products with similar prices, formats, store signals, and sales-field coverage.
Commercial profileListingsMedian priceOfficial shareSales coverage
Official museum DIY kitsOfficial-store excavation and educational kits
26
CNY 74
96%
31%
Mainstream museum gift rangeBroad mix of mid-priced museum souvenirs
62
CNY 51
85%
45%
Low-price marketplace DIY kitsBudget excavation kits, mostly outside official stores
20
CNY 16
5%
60%
Higher-price official figurinesOfficial-store display collectibles at higher prices
18
CNY 69
94%
33%
Character collectibles with sales dataCharacter-led items with complete displayed-sales fields
24
CNY 49
79%
100%
Cluster names are analyst interpretations. Sales coverage measures field availability, not demand.
Entry price drives breadth
137 of 150 listings sit at CNY 100 or below.
Experience separates demand
Collect/display and discover/build are two distinct shopping missions.
Premium remains a test
Only 13 listings sit above the mass market; validate before scaling.
03 · Format mix
What experience does each product format offer?
Figurines are designed for collecting and display. DIY excavation kits add participation, education, and gifting.
Sales field coverage
52%
78 available72 missingMissing stays null and is never recoded to zero.
04 · What to do next
Three decisions a product team could test
Each recommendation starts with an observed market pattern and ends with a measurable next step.
Observed evidence137 of 150 listings are priced at CNY 100 or below
Portfolio decisionDefine the price ladder
Portfolio architectureKeep accessible entry products, then make design, material, and IP the visible reasons to trade up.
Observed evidence71 figurines and 45 DIY kits lead the format mix
Portfolio decisionMerchandise by mission
Customer journeySeparate collect-and-display products from discover-and-build experiences across navigation and gifting.
Observed evidenceOnly 13 listings sit in premium or high-end price tiers
Portfolio decisionTest premium before scaling
Measured experimentRun focused pilots and require conversion, margin, repeat purchase, and inventory evidence.
Decision boundaryClusters describe positioning patterns. They do not prove authenticity, quality, or purchase causality.
05 · Why five groups?
Why use K=5 instead of four or six?
Among K=2 to K=6, five groups give the best balance of separation, repeatability, and usable group size. The separation is still modest, so these remain exploratory profiles.
Selected model · K=5
06 · Evidence check
What can this project claim with confidence?
These cards separate reproducible results, evaluated models, and conclusions the available data cannot support.
REPRODUCIBILITY
Passed
14 automated tests · fixed random seed · source checksums
MODEL QUALITY
Evaluated
Cluster stability · held-out LDA · sentiment and aspect F1
BOUNDARIES
Not supported
Causality · authenticity · revenue or demand forecasting