SB StartupBasket
All ideas
74 /100 GO Medium complexity

CODPasti — return-loss gate for Indonesian COD sellers

Auto-confirms every COD order on WhatsApp in Bahasa and ships only the buyers who firmly say yes.

Retail SaaS SEA WhatsApp-firstAI-agentMultilingualSMBCompliance-driven
— views
Evaluation Scores
74/100

GO

Overall Score

16
Problem
13
Demand
11
Build
12
Distrib.
12
Revenue
8
Time
4
Defense

CODPasti — return-loss gate for Indonesian COD sellers

1. One-liner

Auto-confirms every COD order on WhatsApp in Bahasa and ships only the buyers who firmly say yes.

2. Trend signal — why now?

Cash-on-delivery is not a payment method in Indonesia — it is e-commerce. BPS data puts COD at 83.11% of Indonesian e-commerce transactions. And COD is bleeding sellers dry: failed COD delivery runs ~15% in the Philippines vs 3–5% for prepaid, and in COD-dominant markets return-to-origin (RTO) rates land at 20–40% depending on category. Every rejected package means the seller eats two-way shipping, tied-up stock, and — as of 2026 — a per-rejection fee.

Three things changed in the last 12 months that make now the moment:

  1. The rejection got expensive. Indonesian sellers report a new Rp5,000 fee per rejected COD package (effective June 2026) on top of return shipping. TikTok Shop rolled out a COD “inspect-before-pay” policy (via J&T) from mid-2026 — buyers can open the box and walk away, and the seller absorbs it. Doorstep rejection is now a line item, not an annoyance.
  2. Confirmation got cheap. WhatsApp Business API moved to per-message pricing — a utility message costs ~Rp55, and customer-initiated replies inside the 24-hour window are free. Confirming an order now costs pennies against a Rp5,000+ loss.
  3. The AI to do it well got cheap. Multilingual LLMs handle Bahasa Indonesia + regional phrasing at near-zero marginal cost, so a bot can hold a natural “are you sure you’ll be home Thursday?” conversation instead of a rigid template blast.

India already proved the model works: GoKwik has raised $70.5M, scores every COD order on 200+ signals, confirms risky ones over WhatsApp + OTP, and claims up to 40% RTO reduction. Indonesia’s 30,000-seller platforms (OrderOnline.id) offer order management and a WhatsApp gateway — but nobody there is running GoKwik’s risk-score-then-confirm playbook tuned to the Indonesian social-commerce order flow.

Provenance:

3. The opportunity

The playbook is validated in India and structurally identical in Indonesia — but nobody has ported it. The gap has three edges:

  • GoKwik-shaped hole in SEA. GoKwik/WareIQ/RapidShyp are India-native: English/Hindi, Shopify-D2C-centric, priced and packaged for Indian brands. They don’t serve the Indonesian seller whose orders arrive through a WhatsApp DM, an OrderOnline landing-page form, or TikTok Shop — a different order flow, a different language, a different fraud pattern (“order fiktif,” fake addresses, box-inspection rejections).
  • SEA incumbents are the wrong shape. OrderOnline.id (30K sellers) and Wati are broad order-managers with a generic “COD reconfirmation” checkbox — a template blast, not a risk engine. They confirm everyone the same way, which trains buyers to ignore the message. No one is scoring which orders are likely to flake and spending the conversation budget only where it pays.
  • The manual workaround is the real competitor. Today the fix is a staffer WhatsApp-ing each COD buyer by hand — “kak, dikonfirmasi ya, besok di rumah?” — one message at a time. It doesn’t scale past a few hundred orders/day, so sellers ship the un-confirmed ones and pray. That labor is exactly what an AI agent collapses.

A focused, Indonesia-first tool that scores every COD order and runs a real Bahasa confirmation conversation only on the risky ones — then feeds the outcome back to shipping — can cut RTO materially for a wallet that global tools never priced for.

4. Target market

  • Primary customer: Indonesian social-commerce sellers (UMKM) doing 300–3,000 COD orders/month, 1–5 staff, where COD is 60–80% of orders. Their orders arrive via WhatsApp, OrderOnline-style landing-page forms, and TikTok Shop/Shopee. Fashion, muslimwear, beauty, supplements, home goods — the high-RTO categories.
  • Why they buy (in their words): Sellers on MediaKonsumen (June–July 2026) document package after package rejected or returned with return shipping charged back to them — one seller charged Rp18,900 ongkir on a damaged/short return, another’s appeal rejected despite full video evidence. The pain is concrete and monthly: rugi ongkir on every doorstep rejection, plus the new per-package fee, plus stock stuck in reverse logistics for 18+ days.
  • Rough TAM reasoning: OrderOnline.id alone serves 30,000 Indonesian sellers; the broader COD-seller population running WhatsApp order flows is in the hundreds of thousands. Capture a few thousand mid-volume sellers and the ARR math (below) clears $1M comfortably — well short of needing the whole market.
  • Why now for them: The Rp5,000/rejection fee and TikTok’s inspect-before-pay policy turned a tolerated leak into a per-order tax in 2026. Sellers who shrugged at RTO last year now feel every rejection in cash this year.

5. Product sketch (MVP)

  • One-tap order intake — connect a WhatsApp number + paste/CSV/webhook from OrderOnline or a landing-page form; every new COD order lands in CODPasti.
  • Risk score per order — flags likely-to-flake orders using signals available on day one: repeat-address history, price-vs-category outliers, incomplete address, time-of-order, prior rejection on that number, product category RTO baseline.
  • AI confirmation conversation in Bahasa — for risky orders, a natural WhatsApp exchange (“kak, ordernya [produk] Rp[X], COD ya — bisa dipastikan Kamis di rumah?”), not a template — handles “ganti alamat,” “reschedule,” “batal” gracefully, and switches to regional phrasing.
  • Ship / hold / cancel verdict — each order gets a clear disposition: confirmed → ship, reschedule → hold, no-response/decline → cancel before it costs you ongkir.
  • Reason-tagged cancellations — so the seller (and the risk model) learns why orders die: fake number, address wrong, buyer cold feet, price shock.
  • Return-loss dashboard — Rupiah saved this month = (orders held/cancelled that would have been rejected) × (return shipping + Rp5,000 fee + margin). The number that justifies the subscription.
  • Bahasa + PDP-compliant consent — messaging templates and phone-list handling built to UU PDP from day one.

6. AI angle — what’s load-bearing

Two places, both load-bearing:

  1. The confirmation conversation. A template blast (“Balas YA untuk konfirmasi”) gets ignored — buyers are trained to swipe past it. An LLM that actually converses in Bahasa, absorbs “besok aja,” “alamatnya ganti,” “harganya kok segitu,” and steers to a firm yes/no is the difference between a confirmation rate that moves RTO and one that doesn’t. Remove the AI and you’re back to the useless template every incumbent already ships.
  2. The risk score. Deciding which orders to spend a conversation on — and eventually predicting RTO probability — is a model that compounds on the seller’s own accumulating outcome data. Without it you either confirm everyone (annoying, costly, low-signal) or no one.

If you strip the AI out, this is just another WhatsApp broadcast tool — which is exactly why the existing WhatsApp broadcast tools don’t fix RTO.

7. Localization angle

This is the localization play — it’s the whole moat against GoKwik.

  • Language: Bahasa Indonesia + regional register (Jawa, Sunda phrasing) in the confirmation flow. Global tools speak English/Hindi.
  • Payment/rails reality: built around COD-as-default (83% of orders), not prepaid-with-COD-as-exception — the inverse of how Western/Indian D2C tools are shaped.
  • Local pricing: a Rp199K–799K/mo tier lands where a $49–99/mo global SaaS never could for a UMKM.
  • Local fraud pattern: tuned to “order fiktif,” fake-address rejections, and the specific Shopee/TikTok/J&T COD-inspection flows — not generic chargeback fraud.
  • Distribution: WhatsApp-native, sold through the exact channels (seller Facebook groups, OrderOnline community, TikTok seller creators) where Indonesian UMKM actually gather.

8. Business model — path to $1M–$5M ARR

  • Pricing: tiered by monthly COD order volume. Starter Rp199K/mo (~$12) up to 500 orders; Growth Rp499K/mo ($30) to 1,500; Pro Rp799K/mo ($49) to ~3,000+. Blended ACV ≈ $300/yr. (Buyer’s WhatsApp API/message cost passes through or is bundled at higher tiers.)
  • Rough math to $1M ARR: ~2,800 sellers × ~$30/mo × 12 ≈ $1.0M. Against a base of 30K+ reachable OrderOnline sellers alone, that’s <10% penetration of one channel.
  • Rough math to $5M ARR: ~14,000 sellers at blended ~$30/mo, or fewer sellers at higher ACV via a success fee on RTO saved (e.g. a slice of documented Rupiah saved) layered on the subscription. The dashboard already computes the saved number — charging against it is natural.
  • Expansion path: upsell from confirmation into (a) full COD risk-scoring / auto-block, (b) NDR re-attempt automation, (c) prepaid-nudge (convert flaky COD buyers to QRIS/BNPL upfront), (d) multi-channel (Shopee/TikTok/Lazada order sync). Each raises ACV without a new customer.

9. Go-to-market wedge — first 100 customers

  • OrderOnline / order-form seller communities. These sellers already run WhatsApp order flows and feel RTO acutely. Post a free “RTO calculator” (paste your last month’s rejection count → see Rupiah lost), then DM the sellers who run the numbers. Target the ~50 most active Indonesian COD-seller Facebook/Telegram groups.
  • MediaKonsumen / complaint-thread outreach. Sellers publicly documenting rugi ongkir on MediaKonsumen and platform seller forums are pre-qualified, angry, and named. Reach out directly with “we can hold the order before it becomes this complaint.”
  • TikTok/Reels seller creators. A dozen Indonesian “jualan online” creators teach UMKM how to sell; a demo showing “Rp X saved last week” is native content for that audience — affiliate/rev-share deal, not paid ads.
  • Free 200-order pilot. Onboard a seller, run confirmation on their next 200 COD orders free, hand them the Rupiah-saved dashboard. If it saved more than the subscription, they convert. Falsifiable and fast.

10. Build complexity — justification

Medium. Off-the-shelf: WhatsApp Business API, a multilingual LLM for the confirmation conversation, standard web dashboard, CSV/webhook order intake. Custom work is the risk-scoring logic (starts as heuristics, becomes a real model as outcome data accumulates), the conversation state machine (reschedule/cancel/address-change branches), and the integrations into OrderOnline/TikTok/Shopee order sources — the last is where the weeks go. A small team ships a credible v1 in ~10–14 weeks; the heuristic risk score is good enough at launch and the model earns its keep later. No research breakthrough, no dataset that doesn’t exist.

11. Gating checklist

GatePass?Note
Legal in target market✅Order confirmation is legitimate; must honor UU PDP consent/data rules for phone lists.
Ethical — no harm / dark patterns✅Reduces waste for sellers and couriers; confirming genuine buyers is pro-consumer. No pressure/dark patterns.
Market exists (evidence above)✅83% COD share, 20–40% RTO, documented seller losses, funded India analog.
1–5 person team can build this✅Off-the-shelf AI + WhatsApp API; integrations are the main lift.
Launchable with <$50K / ₹40L✅No capex; message/inference costs scale with usage.

All five pass.

12. Feasibility score

AxisWeightScoreNotes
Problem intensity2016/20Monthly, in-cash pain now that rejection carries a per-package fee; sellers actively hand-confirm today. Not quite daily hair-on-fire for the smallest sellers.
Demand evidence1513/15Hard signals stack: 83% COD, 20–40% RTO, dated seller complaints, $70M-funded India analog. A skeptic nods.
Build feasibility1511/15Off-the-shelf AI + WhatsApp; order-source integrations and conversation branching are real but bounded work. ~10–14 wks.
Distribution clarity1512/15Named channels (OrderOnline communities, seller FB/Telegram groups, MediaKonsumen, TikTok creators) with a concrete pilot funnel. Conversion unproven.
Revenue mechanics1512/15Pricing benchmarked to local WhatsApp SaaS; clear ROI vs Rp5,000+ loss. ACV modest, so needs volume — but volume exists.
Time to first revenue108/10Free-pilot → paid in weeks; ROI is a single visible number.
Defensibility104/10Execution + accumulating outcome-data moat; copyable. GoKwik or a local player could enter. Speed + Bahasa depth is the edge.
Total10074/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy (AI conversation + risk model + WhatsApp/order integrations) · operations-heavy (hands-on seller onboarding, local community distribution, Bahasa content)

Key assumptions to validate (3–5)

  1. Assumption: An AI Bahasa confirmation conversation lifts confirmation-response rate meaningfully above a template blast (enough to move RTO). How to test: A/B a real seller’s next 400 COD orders — template vs AI conversation — measure response rate and doorstep-rejection rate.
  2. Assumption: Mid-volume sellers will pay Rp199K–799K/mo when shown Rupiah-saved. How to test: Run 15 free 200-order pilots; count how many convert to paid after seeing the dashboard.
  3. Assumption: A heuristic risk score (no ML) already beats “confirm everyone” on cost-per-RTO-avoided. How to test: Compare RTO outcomes on scored-and-confirmed vs blanket-confirmed cohorts in the pilots.
  4. Assumption: Order-source integration (OrderOnline/TikTok/Shopee) is reachable without official partnership. How to test: Build webhook/CSV intake for the top 2 sources; confirm a seller can wire it in <30 min.

Risk flags

  1. Platform dependency: Reliant on WhatsApp Business API pricing/policy and on order-source access. Meta reprices messages or a marketplace closes its order flow → economics or intake shift. Mitigate by supporting multiple order sources and passing message cost through.
  2. Incumbent entry: GoKwik (funded, proven) or OrderOnline (30K sellers, incumbent distribution) could bolt on real risk-scoring. The window is the head start + Bahasa/COD-flow depth, not a durable moat.
  3. Market timing / policy: If platforms shift the Rp5,000/return-shipping burden off sellers (consumer-protection pressure exists), the acute ROI softens. Watch the regulatory direction.
  4. Buyer fatigue: If confirmation messages proliferate across all tools, buyers may tune them out — the AI-conversation quality has to stay ahead of the template noise.

14. Structured verdict

Score:                  74/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder (AI + integrations) with an Indonesian ops/distribution partner
Time to revenue:        8–12 weeks (free pilot → paid)
Capital to launch:      $8–15K (₹7–13L) — mostly WhatsApp/inference usage + local ops
Top 3 assumptions to validate first:
  1. AI Bahasa confirmation beats template blast on response + RTO — A/B 400 real COD orders
  2. Mid-volume sellers convert at Rp199K–799K/mo after seeing Rupiah-saved — 15 free pilots
  3. Heuristic risk score beats confirm-everyone on cost-per-RTO-avoided — cohort comparison in pilots
Kill criteria:
  - Abandon if AI confirmation fails to beat a template blast on doorstep-rejection rate in the A/B
  - Abandon if <3 of 15 free-pilot sellers convert to paid after seeing the saved-Rupiah number
  - Abandon if a marketplace/regulator shifts the return-shipping+fee burden off sellers, gutting the ROI

15. Next step — 1-week validation sprint

  • Day 1–2: Recruit 3 mid-volume Indonesian COD sellers from OrderOnline/seller FB groups. Pull their last month’s order + rejection data to baseline RTO and current confirmation practice.
  • Day 3–4: Wire a WhatsApp number to a hand-built AI Bahasa confirmation flow. Run it live on the next batch of one seller’s incoming COD orders, split against their normal template/manual process.
  • Day 5: Decide go/no-go on a single number: does the AI-confirmed cohort show a lower doorstep-rejection rate than the control cohort, by a margin that clears the message cost?

Falsifiable outcome: rejection-rate delta between AI-confirmed and control cohorts, in Rupiah saved per 100 orders. No delta → no product.

Interested in a detailed proposal?

Get a deep-dive with market research, competitive analysis, and implementation roadmap.

Contact us

info@startupbasket.ai