History, live-app walkthrough, diagnosis, competitors, market size, the "Jaisa dikha, waisa mila" strategy across Choose, Buy, Receive and Resolve, prototype, roadmap and business case. Every fact, hypothesis and assumption is labelled.
Fact From a filing or credible report, source citedHypothesis A belief to test with users or dataAssumption A modelling input; replace with internal dataTarget A proposed goal, to be agreed
Stop competing on price. Compete on "it will be what you saw."
Where Snapdeal is
No. 2 in India in 2014–15; lost that place by 2017. Today: ₹1,093 cr NMV and 1.22 cr buyers in FY26, 65% of units to Tier 2+, still loss-making. The app is clean and easy; Instagram has 3.1M followers but little engagement.
What's wrong
Meesho, Shopsy and Amazon match the low prices with zero seller commission. Buyers can't judge unbranded goods before buying. "100% Quality Guaranteed" is a claim, not proof. Let-down buyers are not won back.
What to do
Make one promise, "Jaisa dikha, waisa mila", and keep it when buyers choose, buy, receive and resolve. Start with six near-free fixes; pilot checked listings in two categories.
"Not as described" returns −2 pp vs control · 90-day repeat +3 pp · unhappy buyers reorder · conversion drop ≤ 1% · refund ≤ 5 days · seller retention ≥ 90%.
Decision requested
Approve ₹35 lakh for a 90-day research quarter (plus six near-free fixes now). Release the rest (≈ ₹2.2 cr over 4 quarters) only when each gate is met.
The fall came from a capital-heavy GMV war and spread focus, not from closing down. Snapdeal still operates.
Feb 2010Founded by Kunal Bahl and Rohit Bansal as a daily-deals site
Oct 2011Becomes an online marketplace
Oct 2014SoftBank invests ~US$627M for ~30%
2014–15No. 2: ~32% share vs Flipkart 44%, Amazon 15%
2015Buys FreeCharge, Unicommerce; stake in GoJavas
Feb 2016Peak valuation ~US$6.5B
Jul 2017Flipkart talks end; FreeCharge sold for ₹385 cr; "Snapdeal 2.0"
Oct 2026AceVector lists; ₹32 issue, ₹26.08 day-1 close (NSE)
Was it really India's second-largest?
Yes, around 2014–15, per a Morgan Stanley estimate reported by CNBC (17 Nov 2015). The exact metric isn't stated; GMV share is likely. By July 2017 Amazon had taken the No. 2 spot.
Did Snapdeal close?
No. It runs as AceVector's marketplace beside Unicommerce (SaaS, listed Aug 2024) and Stellaro Brands. It lost its position in horizontal e-commerce, not its business.
Sources: CNBC (17 Nov 2015); Global Venturing (29 Oct 2014); Business Today (31 Jul 2017); Wikipedia, each date traced to its cited news source; Chittorgarh (Oct 2026).
No profit figure doubled in FY24–26. About 70% of the FY26 improvement came from a one-off ₹57.9 cr FY25 provision not repeating; the rest from revenue growth. Cash burn fell from ₹209 cr (FY23) to ₹27 cr (FY25).
Sources: AceVector RHP restated consolidated figures via Chittorgarh, Oquilia (Sep 2026); MediaNama UDRHP (9 Dec 2025). Different bases (e.g. standalone) are not mixed.
7"Log in to view your cart" even when it is empty: a wall before the shopper is ready
8Mobile number + 4-digit OTP with auto-fill: simple, familiar login
9"Recently viewed" (₹118, ₹227) and "Extra 10% off your next order" auto-applied
What this means
The app already looks good and is easy, so the product problem is trust, not design. Quick wins with no media spend: guest cart, ask for notifications after the first order, and put proof where the claim is.
In-app ratings reflect buyers who were happy enough to stay. The ones who were let down leave and post elsewhere. Snapdeal replies by asking them to email an order ID. A recovery flow inside the app would catch them first.
Posts on X could not be opened in this research; the same themes repeat across the sources here.
Sources: Google Play (12 Sep 2026); App Store; Trustpilot (Oct 2026); consumercomplaints.in (2025). Reviews are self-selected, not a random sample.
Value shoppers can't confidently judge quality, fit, material or seller reliability before buying.
→
Consequences
Hesitation and lower conversion · "not as described" returns · support contacts · weaker repeat purchase. Each NAD return costs a forward and a reverse trip.
→
What we must learn
Which doubt matters most, how often, for which segments, and whether reducing it changes behaviour.
Unbranded is not the same as poor quality
Many value goods come from the same clusters (Surat textiles, Moradabad metalware, Delhi NCR home goods) that supply branded labels. The gap is often brand and ad cost. Quality varies by seller, so the buyer's problem is uncertainty, not low price.
Signals a marketplace can build
Specs checked against photos (fabric, size chart, dimensions)
Verified-buyer photos before seller photos
Seller accuracy score from NAD returns
Predictable returns, so trying is low-risk
Evidence boundary: no primary research or Snapdeal return data is claimed. The problem is ranked as the key research gap.
snapdeal02 / Diagnose · Where to winexplains 7-page p.3
Position: the trusted value app, not the cheapest one
Illustrative positioning by the author Hypothesis, to test in the Q1 journey audit.
Q1 equivalent-journey audit
Buy a ₹399 kurta and a ₹299 storage set on all four apps to one Tier 2 pin code, then return one. Record specs, buyer photos, delivery promise vs actual, COD fee, refund days and languages.
Strengths to build on
Clean, aesthetic, easy app (observed)
15+ years of value sellers; Tier 2+ reach
Unicommerce in the same group, used by many sellers
Receive and Resolve: keep the date, fix fast, prove it
Receive
One promised date, re-check before dispatch, Hindi WhatsApp updates, ₹30 back if late. Then one tap: "did it match the photo?"
Resolve
Refund on pickup, a checked replacement, or keep it with money back. A person calls in Hindi within 24 hours.
Who to reach first
Signal
Buyers who tapped "Not quite"
Same day
Recent NAD returns or complaints
Last 90 days
1–2★ reviewers on app stores and Trustpilot
Public reply in 24 h
Close the loop: tell the buyer when the seller fixed the listing, and offer ₹50 off a checked item with a free return. Credit is paid only on a completed order.
Fix friction: guest cart, notification ask after first order
Product change; no media spend
Team time
Buyer-photo credits
₹15 credit per approved photo; ~20,000 photos
₹3 lakh
Tier 2 micro-creators
50 creators × ~₹10,000, or product barter
₹5 lakh
WhatsApp updates in Hindi + buyer community
Order updates, festive drops, support
₹2–3 lakh
Capped referral: give ₹50, get ₹50
Paid only after a completed, non-returned order; ~10,000 pairs
₹10 lakh
Seller sharing kit
Sellers share their Snapdeal store link with their own customers
₹0
Use what exists: 3.1M followers, "10% off next order"
Retarget followers and recent buyers with Verified Value items
Existing
Total
About 1% of one quarter's marketing (FY26 marketplace marketing: ₹84.4 cr a year)
≈ ₹20–25 lakh
Track: new buyers, cost per first order, 30-day repeat, share of orders from referrals and creators. Stop any move whose cost per repeat buyer is above paid ads.
1Home with proof Choose2Honest search Choose3Proof, maker, fit Choose4Buyer photos Choose5Guest cart Buy6Checkout promise Buy7Tracking promise Receive8Did it match? Receive9Fix it fast Resolve10Prove it was fixed Resolve11Truth Score fix Seller12Wrong-flag appeal Seller
(Seller) Fix a flagged listing, then appeal a wrong flag
Both done; understands why
Hindi + 1 regional language · 44 px tap targets · text, not icon-only, badges · screen-reader labels.
Designed failure case
The AI wrongly flags a correct listing as "colour mismatch". The listing stays live, a person decides within 48 hours, and the seller can appeal with photos.
Q1 · QUICK WINS + RESEARCH₹35 lakhSix quick fixes in month 1; interviews, survey, returns baseline; low-cost growth testsOwners: PM, UX research, DataGo if NAD is a top-3 return reason; ≥ 70% sellers opt in
Q2 · CHOOSE + BUY₹60 lakhChecked listings A/B for 50% of pilot buyers; Truth Score and coaching; guest cartOwners: PM, Category, Seller Ops, CXGo if NAD −2 pp; conversion drop ≤ 1%; net ≥ 0
Q3 · RECEIVE + RESOLVE₹45 lakhPromised date with ₹30 if late; "did it match?"; refund on pickup; Hindi callbackOwners: PM, Logistics, CXGo if repeat +3 pp; refund ≤ 5 days
Q4 · EXPAND₹80 lakhUp to 4 more categories; more towns where Points workedOwners: Category heads, FinanceGo if contribution/order ≥ baseline
Dependencies
Order and returns data · listing-check model · 6–8 moderators · experimentation tool · carrier data (Q3)
Stop rule
If Q2 net is negative after 8 weeks, stop or redesign. Total ≈ ₹2.2 cr, released only gate by gate.
Costs are assumptions for team time, moderation, research and tooling.
snapdeal04 / Build · AI & systemsexplains 7-page p.5
AI flags. People decide.
Use
How
Evaluation before launch
Human role
Listing checks
Compare title, specs and images; flag mismatches ("cotton" but image shows sheen)
Precision ≥ 90% on 1,000 labelled listings
Confirms flags before any penalty
Review summary
Pros and cons, each linked to source reviews
Faithfulness check on 200 summaries
Hidden under 20 reviews; feedback loop
Support replies
Suggest replies from policy + order status (RAG)
≥ 90% correct on a 150-question test set
Agent sends; refunds by rules or people
Seller risk
Spot rising NAD returns by seller
Lead time vs manual review
Coaching first; penalty after review and appeal
Catalogue
Listings, specs, images
→
Checks service
AI + rules + review queue
→
Reviews & photos
Upload, moderation
→
Seller console
Score, fixes, appeals
→
Orders & returns
Reason codes, refunds
→
Data & experiments
A/B, dashboards
Guardrails: no final refund or penalty decisions by AI · return reasons cleaned to codes first · buyer photos with consent, faces blurred · model calls only on new or changed listings (≤ ₹0.5 per check, assumption).
snapdeal05 / Measure · Business caseexplains 7-page p.6
It pays only if bad returns fall about 2 points
Per 1,000 pilot orders, ₹ Assumption
Cons.
Base
Opt.
Fewer NAD returns (pp)
1
2
3.5
Saved trips (₹110 per avoided return)
1,100
2,200
3,850
Extra repeat orders × ₹15 contribution
150
450
750
Checks, moderation, support, coaching
−3,000
−2,500
−2,000
Net per 1,000 orders
−1,750
+150
+2,600
Per quarter at 3 lakh pilot orders
−₹5.25 L
+₹0.45 L
+₹7.8 L
Break-even
At ₹2.5 cost per order and +3 pp repeat (₹0.45 per order), NAD returns must fall by (2.5 − 0.45) ÷ 110 ≈ 1.9 pp. That is why the Q2 gate is −2 pp.
The ≈ ₹2.2 cr build cost is not paid back inside the pilot in any scenario; payback depends on expanding to more categories. Replace all unit costs with internal data in Q1.
