Reason codes for returns and churn, read from Hinglish chats and reviews
TL;DR Whysort reads what a D2C brand's customers say in support chats, chat-app voice notes, marketplace reviews and courier remarks, in Hinglish and regional languages, and tags each message with a plain reason tied to the order and SKU. Brands pay a monthly fee to learn which products and promises drive returns and lost repeat buyers.
India's online shoppers don't look like the people most CX tools were built for. More than 270 million Indians bought online in 2024, and almost 60% of new customers since 2020 came from Tier-3 and smaller towns[1]. Hindi is the first language of 43.63% of the country, and the Constitution lists 22 scheduled languages[2].
So a size complaint shows up as "fitting chhota hai" in a voice note, not as a neat survey answer.
Most mid-size brands read this by hand, or not at all. Feedback sits split across a helpdesk, chat apps, the big marketplaces' review pages and the courier panel, and nobody joins it to the return log, so the same bad size chart or misleading photo keeps costing money for months.
The seed promised AI that decodes customers' emotional drivers, and I reframed it. A mood score doesn't tell a merchandiser what to fix, and the global tools that work, from global feedback-analytics vendors, sell themes tied to churn and revenue and not emotions[3]. So Whysort sells reasons.
It connects to the brand's helpdesk, chat inbox, review pages and order system. A language model transcribes voice notes and sorts every message into a short, brand-approved list of reason codes - size runs small, colour differs from photo, late delivery, damaged pack, gift not as expected. Each code links to the SKU, pin code and return outcome.
A weekly report ranks the reasons by cost and shows ten real quotes for each, so the category team can check the machine's work.
The brand stays the data fiduciary under the DPDP Act. Whysort processes on its behalf and deletes raw text on a fixed schedule.
The users are the CX lead and category managers. The payer is the founder or growth head of a D2C brand that sells through its own site and marketplaces. Indian brands already pay for chat-app tooling, and one popular tool lists plans at ₹2,799–3,799 a month[4]. I'd test ₹15,000–60,000 a month, priced on message volume, but that's my assumption until pilots confirm it.
Language models can now transcribe and label mixed Hindi-English text well enough for a human to check by sampling quotes, which wasn't cheap three years ago. The buyer base has shifted too. One consulting report says Gen Z makes up about 40% of online shoppers, and growth slowed to 10–12% in 2024[1]. Slower growth pushes brands to fix returns and repeat rates instead of buying more traffic.
There's no public figure for India's feedback-analytics or voice-of-customer software market, and I found no fetchable public figure for D2C return rates either. The firmest number is the shopper base - over 270 million online buyers and about US$60 billion (roughly ₹5.8 lakh Cr) of e-retail GMV in 2024[1].
Next door, online-first brands grew revenue more than threefold between 2020 and 2022[5]. Treat these as context, not your market.
One Indian loyalty vendor bundles a feedback-management module into its CRM suite[6], but it sits inside a loyalty and marketing platform. Global feedback-analytics platforms do deep analysis for large software and mobility firms[3][7], and they're priced and built for English-first software and retail companies.
The gap is a low-cost, India-first tool that reads Hinglish voice notes and ties each complaint to a return.
| Difficulty | Moderate (34/100) |
| Starting capital | ₹45–80 L |
| Time to launch | 6–9 months |
The models exist. The hard parts are integrations with lots of helpdesks and marketplaces, and keeping reason codes accurate across languages without a human labelling everything. Selling is a grind too, because each brand wants its own codes and CX leads are wary of yet another dashboard.
No special licence here. This is analytics software for brands, so you only need the baseline and the platform rules.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't trust the model on its own for anything a brand will act on in money terms. A person should read the sample quotes before a size chart or supplier changes.
Want to work out the AI layer for Whysort? Talk to Yash →
Here's what can go south -
I like this for a founder who has run CX or merchandising at a D2C brand and can write decent data pipelines. If you want a pure AI product with no services work then skip it. Before raising, prove that five paying brands changed a size chart, photo or pack because of a Whysort report, and that returns on those SKUs fell.
Plan for roughly ₹45–80 L, mostly for integrations with helpdesks, marketplaces and chat apps, plus the first sales hires.
It can be, but small brands pay small cheques. Pricing is my assumption until pilots confirm brands will pay ₹15,000–60,000 a month.
Yes. Mixed-language feedback exists in many markets. Retrain the reason codes for local languages and follow your own data protection law.
The free Whysort validation kit has the line-by-line capital breakdown, the effort scoring, the data-protection checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for brand teams, and 5 assumptions to test. Download the kit →
Thinking about building Whysort or something close to it? The make-or-break work is getting reason codes accurate enough in Hinglish that a category team actually acts on them. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.