Wishlist→purchase discovery engine · last 12 months · generated 2026-08-26 05:32 · Myntra (subscription sample · n=50)
Business metric: % of users who purchase ≥1 wishlisted item within 30 days of adding it. Solution constraint: non-monetary (no discounts).
Each area is sized by how many relevant items map to it and its share of that brand's relevant corpus. Needs money? flags whether fixing it would require a monetary lever (a discount / price cut) — the brief bars monetary solutions, so “no” areas are the addressable ones.
| Opportunity area | Dimension | n | % of relevant | Needs money? |
|---|---|---|---|---|
| delivery-concern | purchase-blocker | 5 | 12.8% | non-monetary |
| return-hassle-fear | purchase-blocker | 4 | 10.3% | non-monetary |
| price | purchase-blocker | 4 | 10.3% | money |
| value-for-money | residual-uncertainty | 4 | 10.3% | non-monetary |
| reliable, easy returns and exchanges | unmet-need | 3 | 7.7% | non-monetary |
| dependable delivery and accountability | unmet-need | 3 | 7.7% | non-monetary |
| competitive pricing and rewards | unmet-need | 3 | 7.7% | money |
| fit-size | residual-uncertainty | 2 | 5.1% | non-monetary |
| accurate sizing and fit assurance | unmet-need | 2 | 5.1% | non-monetary |
| premium browsing and clear product imagery | unmet-need | 2 | 5.1% | non-monetary |
| fit-size-uncertainty | purchase-blocker | 1 | 2.6% | non-monetary |
| quality-material | residual-uncertainty | 1 | 2.6% | non-monetary |
| color-accuracy | residual-uncertainty | 1 | 2.6% | non-monetary |
Largest addressable (non-monetary) area: delivery-concern — 5 items (12.8% of relevant).
“This app has no safety for customers' orders. I neither received my parcel nor was I contacted by the delivery person. But the status of my parcel shows 'Delivered'. I did not get any OTP and no one called me to get one, so how can this app show that the parcel was delivered? I have raised this query, and now they are ”
playstore · 2026-08-22 · 3★ · gp_06744991-be3c-4f23-a8cd-fe22b409bb26
“Delayed Delivery. Pathetic app, I ordered a bag worth 6.5k. Payment is done and all. I placed my order on 4th august. Its 21st August today, my bag has not arrived yet. The initial delivery was supposed to happen on 11th, then 13th then they raised a ticket on 14th asking until 18th, on 18th they asked for 48 hours and”
appstore · 2026-08-21 · 1★ · as_14454001996
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
Why do users add fashion products to their wishlist?All 39 relevant items, each assigned one wishlist reason. Percentages are of 39, so the rows sum to the whole.
All 39 items appear in the rows below — each in a specific group, nothing dropped (rows sum to 39).
| Theme | n | % | Detail |
|---|---|---|---|
| other | 34 | 87.2% | |
| price-watch-wait-sale | 4 | 10.3% | |
| gift-idea | 1 | 2.6% |
“They sell defective products with extra price. And returning an defective item is almost impossible. I had bought belt, which was defective and they were not ready to take it back. After that uninstall the app”
playstore · 2026-08-23 · 1★ · gp_376a7c76-a57b-4186-8d92-014a52b307fe
“I have never had such a bad experience buying a product. The product itself is good, but first of all, it was not delivered to my home or anywhere near my house, so I had to travel 4 km to collect it. Also, the T-shirt I received was much larger than a normal L size. I requested an exchange, and when I travelled 4 km a”
playstore · 2026-08-22 · 2★ · gp_07015f27-c191-486b-8759-30b46d07d8c3
“Delayed Delivery. Pathetic app, I ordered a bag worth 6.5k. Payment is done and all. I placed my order on 4th august. Its 21st August today, my bag has not arrived yet. The initial delivery was supposed to happen on 11th, then 13th then they raised a ticket on 14th asking until 18th, on 18th they asked for 48 hours and”
appstore · 2026-08-21 · 1★ · as_14454001996
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What prevents wishlisted products from being purchased?All 14 relevant items that name a purchase blocker (of 39 relevant), grouped by the single strongest blocker.
All 14 items appear in the rows below — each in a specific group, nothing dropped (rows sum to 14).
| Theme | n | % | Detail |
|---|---|---|---|
| delivery-concern | 5 | 35.7% | |
| return-hassle-fear | 4 | 28.6% | |
| price | 4 | 28.6% | |
| fit-size-uncertainty | 1 | 7.1% |
“This app has no safety for customers' orders. I neither received my parcel nor was I contacted by the delivery person. But the status of my parcel shows 'Delivered'. I did not get any OTP and no one called me to get one, so how can this app show that the parcel was delivered? I have raised this query, and now they are ”
playstore · 2026-08-22 · 3★ · gp_06744991-be3c-4f23-a8cd-fe22b409bb26
“Delayed Delivery. Pathetic app, I ordered a bag worth 6.5k. Payment is done and all. I placed my order on 4th august. Its 21st August today, my bag has not arrived yet. The initial delivery was supposed to happen on 11th, then 13th then they raised a ticket on 14th asking until 18th, on 18th they asked for 48 hours and”
appstore · 2026-08-21 · 1★ · as_14454001996
“Very Poor Delivery timelines. They are always delay in delivering items to locations in South. Ekartel people don’t even call or answer our calls”
appstore · 2026-08-22 · 2★ · as_14461074154
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What uncertainties remain after a user identifies a product they like?All 8 relevant items that voice a residual uncertainty after liking a product, grouped by uncertainty.
All 8 items appear in the rows below — each in a specific group, nothing dropped (rows sum to 8).
| Theme | n | % | Detail |
|---|---|---|---|
| value-for-money | 4 | 50% | |
| fit-size | 2 | 25% | |
| quality-material | 1 | 12.5% | |
| color-accuracy | 1 | 12.5% |
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · 1★ · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
“worst app ever so expensive not good productss”
playstore · 2026-08-23 · 2★ · gp_564985dc-d630-4f3f-ac60-b00b485d5d19
“Good app to find good quality trend wears also can't find that mutch offers that's a bad point”
playstore · 2026-08-23 · 3★ · gp_1f14962f-58aa-4dad-aea9-8dda5e575157
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What causes users to postpone a purchase?All 2 relevant items that state a reason for postponing the purchase, grouped by cause.
All 2 items appear in the rows below — each in a specific group, nothing dropped (rows sum to 2).
| Theme | n | % | Detail |
|---|---|---|---|
| waiting-sale | 2 | 100% |
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · 1★ · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
“Good app to find good quality trend wears also can't find that mutch offers that's a bad point”
playstore · 2026-08-23 · 3★ · gp_1f14962f-58aa-4dad-aea9-8dda5e575157
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
How do users compare multiple shortlisted products?All 2 relevant items that describe how they compare shortlisted products, grouped by comparison behaviour.
All 2 items appear in the rows below — each in a specific group, nothing dropped (rows sum to 2).
| Theme | n | % | Detail |
|---|---|---|---|
| across-apps-compare | 1 | 50% | |
| in-app-compare | 1 | 50% |
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · 1★ · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
“Used to be great. Now its just stupid. This was probably my favourite app for shopping. Now its just bad in terms of user experience. The product image quality are quite low resolution. There are stupid sponsored ads shoved so aggressively that its almost stupid. There stupid buttons here and there (why do i need a cli”
appstore · 2026-08-23 · 3★ · as_14461824398
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What information do users seek outside the platform before purchasing?All 0 relevant items that mention seeking information outside the app before buying, grouped by source.
No rows — this question had no usable signal in this corpus.
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What role do fit, size, styling, price, reviews, occasion and social validation play?All 28 relevant items that name at least one factor (of 39 relevant). A single item can touch several factors, so the percentages are shares of 28 and sum to more than 100%.
All 28 items appear in the rows below — each in a specific group, nothing dropped (rows sum to 46).
| Theme | n | % | Detail |
|---|---|---|---|
| quality | 13 | 46.4% | |
| delivery-returns | 13 | 46.4% | |
| price | 9 | 32.1% | |
| size | 4 | 14.3% | |
| fit | 3 | 10.7% | |
| occasion | 1 | 3.6% | |
| styling | 1 | 3.6% | |
| reviews | 1 | 3.6% | |
| brand-trust | 1 | 3.6% |
“They sell defective products with extra price. And returning an defective item is almost impossible. I had bought belt, which was defective and they were not ready to take it back. After that uninstall the app”
playstore · 2026-08-23 · 1★ · gp_376a7c76-a57b-4186-8d92-014a52b307fe
“Used to be great. Now its just stupid. This was probably my favourite app for shopping. Now its just bad in terms of user experience. The product image quality are quite low resolution. There are stupid sponsored ads shoved so aggressively that its almost stupid. There stupid buttons here and there (why do i need a cli”
appstore · 2026-08-23 · 3★ · as_14461824398
“worst app ever so expensive not good productss”
playstore · 2026-08-23 · 2★ · gp_564985dc-d630-4f3f-ac60-b00b485d5d19
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
When is the wishlist genuine purchase intent vs a bookmarking mechanism?All 39 relevant items, each assigned a wishlist nature. Percentages are of 39.
All 39 items appear in the rows below — each in a specific group, nothing dropped (rows sum to 39).
| Theme | n | % | Detail |
|---|---|---|---|
| unclear | 21 | 53.8% | |
| genuine-intent | 16 | 41% | |
| price-watch | 2 | 5.1% |
“They sell defective products with extra price. And returning an defective item is almost impossible. I had bought belt, which was defective and they were not ready to take it back. After that uninstall the app”
playstore · 2026-08-23 · 1★ · gp_376a7c76-a57b-4186-8d92-014a52b307fe
“Size issues. Clothesline is good but order processing is badly handled.Size issues are mostly there forcing you to exchange”
appstore · 2026-08-21 · 2★ · as_14454508909
“I am facing a lot with my exchange order, it's getting frustrated now. how much I have to wait. once they attempt n the delivery agent didn't receive my order as it was his first day don't know much. Calling myntra customer care service is also not helping much ..!”
playstore · 2026-08-25 · 2★ · gp_98c2b13d-1402-4d59-9f71-ee69f0c36a08
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
How do these behaviours differ across user segments?All 39 relevant items grouped by shopper segment (segments with <3 items are pooled). % = the segment's share of the relevant corpus; the note gives that segment's top purchase blocker.
All 39 items appear below (rows sum to 39): 79% fall into a named theme, the remaining 21% are grouped in the “other / not stated” row so nothing is dropped.
| Theme | n | % | Detail |
|---|---|---|---|
| general shopper | 21 | 53.8% | top blocker: delivery-concern (1) · 9 genuine-intent |
| frequent returner | 5 | 12.8% | top blocker: return-hassle-fear (4) · 1 genuine-intent |
| deal / offer seeker | 5 | 12.8% | top blocker: price (3) · 2 genuine-intent |
| smaller / unlabelled segments | 8 | 20.5% | accounted for, but with no single dominant theme |
“delivery parter are very bad”
playstore · 2026-08-22 · 1★ · gp_9643705c-2abb-40dd-ad1f-b088ecc1f93b
“the quick exchange experience made it give it 5 STAR. Thank you for providing such great PRODUCT and Service.”
playstore · 2026-08-23 · 5★ · gp_007b70ff-5523-40c3-985b-0d6948be2cab
“"Myntra is my absolute go-to app for fashion and lifestyle shopping! The user interface is very smooth, search filters are highly accurate, and the sizing guides are mostly perfect. Delivery is usually on time, and refunds are processed swiftly. Highly recommended!"”
playstore · 2026-08-21 · 5★ · gp_51aba27e-3f91-4d42-a157-a32b94d53713
*** n = the number of real feedback items behind each row (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What unmet needs emerge consistently across conversations?All 13 relevant items that state an unmet need, grouped by theme. Themes in 2+ independent sources are the durable ones.
All 13 items appear in the rows below — each in a specific group, nothing dropped (rows sum to 13).
| Theme | n | % | Detail |
|---|---|---|---|
| reliable, easy returns and exchanges | 3 | 23.1% | seen in 2 of 4 sources |
| dependable delivery and accountability | 3 | 23.1% | seen in 2 of 4 sources |
| accurate sizing and fit assurance | 2 | 15.4% | seen in 2 of 4 sources |
| competitive pricing and rewards | 3 | 23.1% | seen in 1 of 4 sources |
| premium browsing and clear product imagery | 2 | 15.4% | seen in 2 of 4 sources |
“They sell defective products with extra price. And returning an defective item is almost impossible. I had bought belt, which was defective and they were not ready to take it back. After that uninstall the app”
playstore · 2026-08-23 · 1★ · gp_376a7c76-a57b-4186-8d92-014a52b307fe
“I am facing a lot with my exchange order, it's getting frustrated now. how much I have to wait. once they attempt n the delivery agent didn't receive my order as it was his first day don't know much. Calling myntra customer care service is also not helping much ..!”
playstore · 2026-08-25 · 2★ · gp_98c2b13d-1402-4d59-9f71-ee69f0c36a08
“Refund not received. I am disappointed with Myntra refund don’t buy with this app”
appstore · 2026-08-22 · 1★ · as_14459333595
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
Why do users add fashion products to their wishlist?SAMPLE (n=50): wishlisting intent is mostly genuine, with a clear price-watch minority
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
“Good app to find good quality trend wears also can't find that mutch offers that's a bad point”
playstore · 2026-08-23 · gp_1f14962f-58aa-4dad-aea9-8dda5e575157
“This app has no safety for customers' orders. I neither received my parcel nor was I contacted by the delivery person. But the status of my parcel shows 'Delivered'. I did not get any OTP and no one called me to get one, so how can this app show that the parcel was delivered? I have raised this query, and now they are ”
playstore · 2026-08-22 · gp_06744991-be3c-4f23-a8cd-fe22b409bb26
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What prevents wishlisted products from being purchased?What blocks purchase: delivery reliability, return-hassle fear, price, and sizing
“Delayed Delivery. Pathetic app, I ordered a bag worth 6.5k. Payment is done and all. I placed my order on 4th august. Its 21st August today, my bag has not arrived yet. The initial delivery was supposed to happen on 11th, then 13th then they raised a ticket on 14th asking until 18th, on 18th they asked for 48 hours and”
appstore · 2026-08-21 · as_14454001996
“This app has no safety for customers' orders. I neither received my parcel nor was I contacted by the delivery person. But the status of my parcel shows 'Delivered'. I did not get any OTP and no one called me to get one, so how can this app show that the parcel was delivered? I have raised this query, and now they are ”
playstore · 2026-08-22 · gp_06744991-be3c-4f23-a8cd-fe22b409bb26
“Very Poor Delivery timelines. They are always delay in delivering items to locations in South. Ekartel people don’t even call or answer our calls”
appstore · 2026-08-22 · as_14461074154
“They sell defective products with extra price. And returning an defective item is almost impossible. I had bought belt, which was defective and they were not ready to take it back. After that uninstall the app”
playstore · 2026-08-23 · gp_376a7c76-a57b-4186-8d92-014a52b307fe
“I have never had such a bad experience buying a product. The product itself is good, but first of all, it was not delivered to my home or anywhere near my house, so I had to travel 4 km to collect it. Also, the T-shirt I received was much larger than a normal L size. I requested an exchange, and when I travelled 4 km a”
playstore · 2026-08-22 · gp_07015f27-c191-486b-8759-30b46d07d8c3
“Refund not received. I am disappointed with Myntra refund don’t buy with this app”
appstore · 2026-08-22 · as_14459333595
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
“Size issues. Clothesline is good but order processing is badly handled.Size issues are mostly there forcing you to exchange”
appstore · 2026-08-21 · as_14454508909
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What uncertainties remain after a user identifies a product they like?Residual uncertainty is dominated by value-for-money and fit/size
“Size issues. Clothesline is good but order processing is badly handled.Size issues are mostly there forcing you to exchange”
appstore · 2026-08-21 · as_14454508909
“I have never had such a bad experience buying a product. The product itself is good, but first of all, it was not delivered to my home or anywhere near my house, so I had to travel 4 km to collect it. Also, the T-shirt I received was much larger than a normal L size. I requested an exchange, and when I travelled 4 km a”
playstore · 2026-08-22 · gp_07015f27-c191-486b-8759-30b46d07d8c3
“Used to be great. Now its just stupid. This was probably my favourite app for shopping. Now its just bad in terms of user experience. The product image quality are quite low resolution. There are stupid sponsored ads shoved so aggressively that its almost stupid. There stupid buttons here and there (why do i need a cli”
appstore · 2026-08-23 · as_14461824398
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What causes users to postpone a purchase?Postponement is mostly waiting for a sale/offer (small n)
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
“Good app to find good quality trend wears also can't find that mutch offers that's a bad point”
playstore · 2026-08-23 · gp_1f14962f-58aa-4dad-aea9-8dda5e575157
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
How do users compare multiple shortlisted products?Comparison is cross-app (price) and, when blocked, in-app by brand
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
“Used to be great. Now its just stupid. This was probably my favourite app for shopping. Now its just bad in terms of user experience. The product image quality are quite low resolution. There are stupid sponsored ads shoved so aggressively that its almost stupid. There stupid buttons here and there (why do i need a cli”
appstore · 2026-08-23 · as_14461824398
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What information do users seek outside the platform before purchasing?Off-platform info-seeking is thin in this store-review sample
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What role do fit, size, styling, price, reviews, occasion and social validation play?Delivery/returns, price, quality, and size are the factors that dominate
“Delayed Delivery. Pathetic app, I ordered a bag worth 6.5k. Payment is done and all. I placed my order on 4th august. Its 21st August today, my bag has not arrived yet. The initial delivery was supposed to happen on 11th, then 13th then they raised a ticket on 14th asking until 18th, on 18th they asked for 48 hours and”
appstore · 2026-08-21 · as_14454001996
“Size issues. Clothesline is good but order processing is badly handled.Size issues are mostly there forcing you to exchange”
appstore · 2026-08-21 · as_14454508909
“I have never had such a bad experience buying a product. The product itself is good, but first of all, it was not delivered to my home or anywhere near my house, so I had to travel 4 km to collect it. Also, the T-shirt I received was much larger than a normal L size. I requested an exchange, and when I travelled 4 km a”
playstore · 2026-08-22 · gp_07015f27-c191-486b-8759-30b46d07d8c3
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
When is the wishlist genuine purchase intent vs a bookmarking mechanism?Wishlist is genuine intent for buyers; a price-watch subset treats it as "wait"
“This app has no safety for customers' orders. I neither received my parcel nor was I contacted by the delivery person. But the status of my parcel shows 'Delivered'. I did not get any OTP and no one called me to get one, so how can this app show that the parcel was delivered? I have raised this query, and now they are ”
playstore · 2026-08-22 · gp_06744991-be3c-4f23-a8cd-fe22b409bb26
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
How do these behaviours differ across user segments?Segments: frequent returners, deal-seekers, occasion/premium, tier-2
“Size issues. Clothesline is good but order processing is badly handled.Size issues are mostly there forcing you to exchange”
appstore · 2026-08-21 · as_14454508909
“very bad,iska saman bahut mahenga hai or mesho shopsy se dubble rate hai ,isse badhiya mesho shopsy hai”
playstore · 2026-08-21 · gp_d4a04135-bfef-4ab6-baa6-0375227a5687
“Very Poor Delivery timelines. They are always delay in delivering items to locations in South. Ekartel people don’t even call or answer our calls”
appstore · 2026-08-22 · as_14461074154
*** n = the number of real feedback items this claim rests on (the evidence count). Higher n = stronger; n=1–2 is a weak anecdote. ***
What unmet needs emerge consistently across conversations?Consistent unmet needs: return confidence, delivery trust, sizing certainty
“They sell defective products with extra price. And returning an defective item is almost impossible. I had bought belt, which was defective and they were not ready to take it back. After that uninstall the app”
playstore · 2026-08-23 · gp_376a7c76-a57b-4186-8d92-014a52b307fe
“Delayed Delivery. Pathetic app, I ordered a bag worth 6.5k. Payment is done and all. I placed my order on 4th august. Its 21st August today, my bag has not arrived yet. The initial delivery was supposed to happen on 11th, then 13th then they raised a ticket on 14th asking until 18th, on 18th they asked for 48 hours and”
appstore · 2026-08-21 · as_14454001996
“Size issues. Clothesline is good but order processing is badly handled.Size issues are mostly there forcing you to exchange”
appstore · 2026-08-21 · as_14454508909
| Product | Coverage | Relevance yield | Avg confidence | Consistency | Human agreement | Triangulation | Citation validity |
|---|---|---|---|---|---|---|---|
| Myntra (subscription sample · n=50) | 100% | 78% | 66% | — | — | 70% | 97% |
Coverage = items that classified into a valid object ·
Relevance yield = share that is substantive shopping feedback ·
Consistency = label stability when a sample is re-classified ·
Human agreement = model vs a hand-labelled gold set (npm run label) ·
Triangulation = top themes corroborated by ≥2 independent sources ·
Citation validity = synthesized claims whose quote ids resolve to real rows.
Myntra (subscription sample · n=50) playstore: 27appstore: 13youtube: 10
Low or zero counts reflect real source availability and are reported as-is.
Every stage below ran for this report. The numbers are this run's, not an illustration.
| Stage | What happens | Myntra (subscription sample · n=50) |
|---|---|---|
| 1 · Resolve | Company name → verified store ids (never hard-coded) | 1 brand |
| 2 · Collect | Play Store · App Store · YouTube, into one schema | 2265 |
| 3 · Clean | Normalise, then drop empty / too-short / spam / bot / duplicate | 1704 |
| 4 · Extract | One structured AI call per item → 16 fields, schema-enforced | 50 |
| 5 · Cluster | Free-text fields → labelled themes carrying their member ids | 27 themes |
| 6 · Aggregate | Plain counting — the 10 questions + money-tagged opportunities, no model opinion | 39 relevant |
| 7 · Synthesise | Narrative written from the aggregates only | 10 answers |
| Rule | Dropped | Rule | Dropped |
|---|---|---|---|
| tooShort | 463 | exactDuplicate | 54 |
| nearDuplicate | 33 | empty | 6 |
| spam | 4 | noLetters | 1 |
2265 collected → 1704 kept (75%), 561 dropped.
Normalised: repeatedPunctuation: 113emoji: 915usernames: 4stretchedWords: 6urls: 3
Every dropped row is kept in dropped.jsonl with the rule that removed it.
Anything we count uses a fixed vocabulary, so counts are reliable. Anything that needs nuance is free text and is clustered — never matched by string, which would produce a list of count-of-1 "themes". Each cluster keeps the ids of the items it was built from, so the taxonomy is derived from this data rather than chosen in advance.
| Field | Method | Raw lines | Themes | Largest theme |
|---|---|---|---|---|
| blocker_note | free text → clustered → label-merged | 15 | 5 | returns and exchanges are hard, slow or refused (4) |
| info_needed | free text → clustered → label-merged | 3 | 2 | accurate size and fit before ordering (2) |
| segment | free text → clustered → label-merged | 38 | 9 | general shopper (21) |
| pain | free text → clustered → label-merged | 20 | 6 | return and exchange ordeal (4) |
| unmet_need | free text → clustered → label-merged | 13 | 5 | reliable, easy returns and exchanges (3) |
| wishlist_reason · purchase_blocker · residual_uncertainty · factors · wishlist_nature · sentiment | fixed enum — chosen by the model from a closed list | — | — | counted directly |
The narrative model never sees the review corpus. It receives only the aggregates above — counts, percentages and quote ids — and must return each claim with the evidence count it rests on and the quote ids it drew from. Those ids are then checked against real rows; the citation validity figure in the quality panel is that check.
| Guarantee | How it is enforced |
|---|---|
| No claim without a number | n is a required field in the output schema |
| No claim without evidence | quote_ids required; rendered inline above |
| No invented quotes | Ids validated against the corpus → 97% valid |
| No reading of raw reviews | The synthesis payload excludes the corpus by construction |
Seven independent checks. Figures are in the Analysis quality panel above.
| Check | Question it answers | This run |
|---|---|---|
| Coverage | Did every item classify into a valid object? | 100% |
| Relevance yield | How much of the corpus is substantive shopping feedback? | 78% |
| Model confidence | How sure was the model of its own labels? | 66% |
| Re-run consistency | Does the same item get the same label twice? | — |
| Human agreement | Does the model agree with a hand-labelled gold set? | not yet run |
| Source triangulation | Do independent communities surface the same theme? | 70% |
| Citation validity | Did the narrative invent any evidence? | 97% |