Product Selection from China for E-commerce Sellers
The 100→30→10→3 funnel — a 10-stage system that lifts seller year-one ROI to 1.6× on average

Hello, this is GreenFrog Seoul.

"I found a great-looking product on 1688, ordered it, and it doesn't sell in Korea."
"I went in expecting a 30% margin — after ad spend it's a loss."
"I built it up to #2 in three months, then a Chinese seller stepped in directly and cut prices in half."

Among Korean sellers, "China sourcing isn't about finding a great factory — it's about picking a winning product" is settling in as conventional wisdom. In our 7+ years on the ground, 70%+ of seller failures don't come from the factory, quality, or logistics — they come from "a loss already decided at the product-selection stage." Bottom line: "finding a great factory" matters less to ROI than "picking a product you can win with."

E-commerce product selection is a "funnel, not gut feel" business. 100 candidates → 30 samples → 10 market tests → 3 mass-production runs. Sellers who systematize this 100:30:10:3 funnel see year-one ROI 1.6× higher on average than sellers who go from gut feel straight to mass production.

Today we condense GreenFrog Seoul's 10-stage product-selection method — refined over 7+ years on the ground with Korean e-commerce sellers — onto one page. Demand validation, category color, margin structure, 1688 hunting, competitor analysis, sampling, differentiation, test launch, SKU expansion, and pivot rules — including where the seller's job ends and where mediation steps in.

1. Why product selection — the 5 places selection-blind sellers break

Product selection isn't "intuition" — it is "the area where Korean sellers, if they don't know it, repeatedly take losses across orders, ads, and inventory." Five gaps that break sellers without a selection system:

2. Stage 1: Demand validation — "data, not gut feel"

Every selection journey starts the same way: "find products that sell in Korea, not products that look good in China." Demand validation must finish before opening 1688.

4 demand-validation data points

4 demand-validation principles

• Enter only when both numbers hit: 10K+ monthly searches AND 30M+ KRW monthly sales for #1–2 — never one alone
• Only categories with +15%+ trailing-12-month growth — flat / declining go to the back of the queue
• Look at Naver, Coupang, and social in parallel — single-channel reads are biased
• Validate demand before searching 1688 — starting on 1688 means letting "what's visible" decide

3. Stage 2: Category selection — "red, blue, and green oceans"

Once demand is validated, the next variable is category color. Same demand, different category — ROI swings 3×.

The 3 category colors

4 category-selection rules

• New sellers: do not enter red oceans — ad spend / CPC will eat the margin
• Blue ocean is optimal — demand validated, pricing still alive
• Green ocean only when the differentiation card is clear — market formation takes time
• Realistic ceiling: 30% of #1–2's sales on entry

4. Stage 3: Margin-structure analysis — "why a 30% gross margin is a 5% net"

"30% margin" means gross margin (revenue – cost). The actual net margin a Korean seller takes home is what's left after ads, fees, returns, logistics, and tax.

7-line net-margin calculation

4 margin rules

• Enter only categories with 35%+ gross — to defend a 5–10% net
• Cap ad spend at 18% of revenue by default — review the moment it crosses
• Categories with 8%+ return rates aren't fit for new sellers — apparel / food need extra caution
• Run net-margin sims before opening 1688 — "if margin doesn't pencil, change the category itself"

5. Stage 4: 1688 / Alibaba hunting — "build 100 candidates fast"

Only after demand, category, and margin pass do you open 1688 / Alibaba. The goal is "100 candidates fast."

4 hunting channels

4 efficiency moves on 1688

• Pre-build a Chinese keyword dictionary — KO→ZH translation increases search hits by 30–50%
• Default filters: 1,000+ sales, 20%+ repurchase, 2+ years registered
• Recognize that 5–10 factories make the same SKU — gives real price comparison
• Compress 100 → 30 with a 4-axis scorecard: price, MOQ, certifications, lead time

6. Stage 5: Korean competitor analysis — "look at #1, #2, and the worst"

With 100 candidates from 1688, return to Korean platform competitor analysis. Same SKU, but who sells it and how in Korea decides ROI.

5 competitor-analysis lines

4 ways to use the analysis

• #1–2 are benchmarks; the worst seller is your differentiation hint — read both
• Drop categories in price-war mode — 3-month price drop -20%+ is a red flag
• Categories with 70%+ ad-share concentration aren't fit for new sellers
• Mining 1–2 star reviews is the starting point of differentiation

7. Stage 6: Sample validation — "30 candidates → 10"

After compressing 100 → 30, the next step is real samples. The sample stage's goal: "30 → 10."

5 sample-validation checks

4 sampling rules

• Cost per sample = 0.5% of mass-production margin — 30-sample cost ≈ 15% of margin
• Korean-shipping box simulation by default — 1%+ damage rate disqualifies for mass production
• SKUs without certification: drop immediately — last-minute cert at mass-production stage adds 6+ weeks
• 2-week real use + 70%+ approval from 5 close users = qualitative pass

8. Stage 7: Differentiation cards — "winning on the same 1688 SKU"

10 Korean sellers see the same 1688 SKU; 9 take it as-is, 1 plays a differentiation card and becomes #1–2. Differentiation is decided right before OEM ordering.

The 5 differentiation cards

4 ways to choose a card

• Solve the bottom seller's 1–2 star review reasons at the OEM stage — fastest differentiation
• Korean-fit sizing / manuals are the best ROI card — 5% cost, 30% lift
• Bundles lift ad-spend efficiency by +30% — by raising AOV
• Adopt private-label branding from the 5th SKU onward — SKUs 1–4 are still in validation

9. Stage 8: Test launch — "throw 10 SKUs into the market"

The 10 SKUs that pass sampling and differentiation don't go straight to mass production — they go to test launch. The goal: "10 → 3."

4-step test launch

4 test-launch rules

• Run the 50–100-unit test even at zero margin — to avoid the bigger 1,000-unit mass-production loss
• Spend 300–500K KRW on ads to verify CPC / conversion — data first, not revenue
• Pass line: 2M+ KRW / 30 days, ROI 1.4×+ — fail = stop immediately
• 3 of 10 will pass on average — accept the other 7 as sunk cost

10. Stage 9: SKU expansion roadmap — "3 → 10 → 30"

The 3 passing SKUs go to mass production. Next: SKU expansion. The biggest trap for Korean sellers is "expanding SKUs too fast."

3-step SKU expansion

4 SKU-expansion rules

• No SKU expansion before the monthly-revenue threshold — ad spend and inventory will fragment
• Adjacent SKUs first; unrelated categories to the back — reuse ad / customer data
• 1M+ KRW/mo per SKU is the pass line — kill anything below it immediately
• 10–15 efficient SKUs deliver 2× the ROI of 30 inefficient ones

11. Stage 10: Pivot and discontinuation rules — "those who can't kill, can't survive"

The last piece of the 10-stage system: pivot / discontinuation rules. Sellers who can't kill non-performing SKUs fast take the biggest losses.

4 pivot / discontinuation triggers

4 pivot / discontinuation rules

• 3 consecutive months below pass line → default discontinuation — "let's wait a bit more" is the trap
• Discontinue on numbers, not feelings — automatic quarterly review
• Clear discontinued inventory within 30 days via discount / B2B — avoid storage cost stack-up
• Recycle data and ad assets from discontinued SKUs into the next SKU

12. GreenFrog Seoul's product-selection mediation service

The 10-stage system above sits in the gap of "too much time, information, and on-the-ground channel required for a seller to run the 100→30→10→3 funnel alone." GreenFrog Seoul mediates the entire funnel from the seller's side via 7+ year on-the-ground consultants.

Mediation package

What this service changes

• Mass-production pass rate 30% → 75% (test-launch effect)
• Dead-inventory share 32% → 8% (quarterly discontinuation effect)
• Net margin 4.2% → 11.8% (margin sim + differentiation)
• Per-SKU monthly revenue 700K → 1.8M KRW (efficient-SKU compression)
• Year-one ROI 1.6× on average (full-funnel effect)

13. Master selection checklist

What not to miss before ordering, after testing, and at quarterly review.

Pre-1688 checklist (demand / category / margin)

• Confirmed keyword volume 10K+/mo and competition under 0.3
• Blue-ocean category — #1–2 sales 30–80M KRW
• Confirmed +15%+ category growth over the trailing 12 months
• Confirmed gross 35%+ and net-sim 10%+
• Capped ad spend at 18% of revenue
• Built a 4-axis scorecard to compress 100 candidates to 30
• Mined the bottom seller's reviews for differentiation hints

Sample / differentiation / test checklist

• Took samples in 3 rounds of 10 / 10 / 10
• Verified Korean-shipping box simulation damage ≤1%
• Decided on 1 of 5 differentiation cards before OEM ordering
• Cleared certifications (KC / KFDA / category-specific) before mass production
• Confirmed 50–100 unit test + 300–500K KRW ad → ROI 1.4×+
• Sent only the 3 of 10 that passed to mass production

Quarterly review checklist

• Quarterly review per SKU at 1M+ KRW/mo
• Default-discontinue any SKU below pass line for 3 consecutive months
• Cut ad spend 50% then review for discontinuation on 2 consecutive loss months
• Replace OEM or discontinue SKUs with reviews < 3.5 or returns 8%+
• Clear discontinued inventory within 30 days via discount / B2B
• Quarterly: extend passing SKUs into adjacent SKUs
• Maintain a 10–15 efficient-SKU structure — avoid the 32-SKU trap

Wrap-up — A great product before a great factory

Compressed to one line each, the 10 stages:

• Stage 1 (Demand validation): Korean data first, not 1688 — keyword / sales / growth, 3 axes
• Stage 2 (Category): blue ocean optimal — new sellers don't enter red oceans
• Stage 3 (Margin): 35%+ gross / 10%+ net is the safe line
• Stage 4 (1688 hunting): 100 candidates → 30 via a 4-axis scorecard
• Stage 5 (Competitor analysis): #1–2 benchmark + bottom-review differentiation hints
• Stage 6 (Samples): 30 in 3 rounds of 10 — never one batch
• Stage 7 (Differentiation): 1 of 5 cards by default — wins on the same 1688 SKU
• Stage 8 (Test launch): 50–100 units / 300–500K KRW ad / ROI 1.4×+ pass line
• Stage 9 (SKU expansion): 3 → 5–15 → 15–30 — efficient SKU pool of 10–15
• Stage 10 (Pivot / kill): 3-month miss → default-kill — dead inventory 32% → 8%

China sourcing is a "pick a winning product, don't just find a great factory" business. Same category, same 1688 SKU, same seller — once the 100→30→10→3 funnel is systematized, mass-production pass rate climbs from 30% to 75%, net margin from 4% to 12%, and year-one ROI to 1.6×. Drawing the funnel right matters more to ROI than finding a great factory. GreenFrog Seoul mediates the seller's full 100→30→10→3 funnel — from demand validation to quarterly review. Whether you're approaching 1688 for the first time or stuck on dead inventory and unable to launch the next SKU, reach out anytime.