Data Sources (Momo)
Published 2026-05-14Updated 2026-06-302 min read
1. Data Scale
| Item | Scale |
|---|---|
| In-house members | N=5,000 (PII masked) |
| SKUs | ~50K (actual PoC), ~100K synthetic |
| OrderTransaction | ~250K |
| TVPurchase | ~30K |
| LiveStreamPurchase | ~50K |
| LiveStream broadcasts | ~500 (1 year) |
| DeliverySLA logs | ~250K |
→ ~800K Neptune edges
2. cohort_tag
| Value | Meaning |
|---|---|
real | PII-masked in-house |
synth | 49.5K synthetic |
external | Social · Weather · Economy · Competitor |
3. Four External Data Sources
3.1 Social
- Dcard · Instagram · X · Xiaohongshu (live-broadcast reviews · SKU trends)
3.2 Weather (Critical for Delivery)
- Central Weather Administration (Taiwan) — heavy rain and typhoons impact delivery SLA
3.3 Economy
- DGBAS (Directorate-General of Budget, Accounting and Statistics) consumption indicators
3.4 Competitors
- PChome · Yahoo奇摩 · Shopee TW public campaigns
4. Live-Broadcast Synthesis Strategy
# Live-broadcast simulation (1-hour average, 5 hosts, 30 SKU pins)
def gen_live_session():
duration_min = 60
viewer_curve = poisson_growth(start=1000, peak=5000, decay=2000)
pin_events = sorted(random.sample(range(duration_min*60), 30))
purchases = [(t + lognormal(0, 1)*60, random_sku) for t in pin_events]
return {viewer_curve, pin_events, purchases}
5. Delivery-SLA Seasonal and Weather Impact
| Event | Impact |
|---|---|
| Typhoon (heavy rain) | 24-hour SLA breach rate +60% |
| Double 11 (光棍節) | Daily orders +500%, SLA breach +25% |
| Lunar New Year (春節) | Delivery -3 days (full shutdown) |