Talk to the team →
Case study · pharmacy chain

Every store has a leaking bucket.

We show you which drops to catch first — a ranked list of the customers, the items, and the rupees walking out, sorted by who's worth saving today.

Talk to the team → Hear a real customer conversation
One shift · one counter

5 revenue leaks in one
14.6-minute conversation

Leak detection timeline for a walk-in visit at a leading pharmacy chain. May 18, 2026 — 4:25 PM. All amounts in ₹ annualised at current visit frequency. Wavy underlines mark where the AI flagged uncertainty — hover for details.

5
Coachable
leak events
4
Stockouts with
no alternative
1
Customer bounce
unhandled
₹3,700
Estimated
value at risk
Detected leaks — chronological order
0:00
Stockout Pentagon 40 — out of stock, no alternative offered −₹800
Customer सिक्स का दो पैंटा काइन 40। कितना वन ठीक है अभी?
Customer मुझे अभी दे दो।
Staff दे दो।
Customer हो जाएगा, ऐसा नहीं होगा।
Staff नहीं हो जाएगा, बट ऑर्डर भी डाल दो — अभी आपके पास एक भी नहीं है क्या?
Customer इमर्जेन्सी में।
Plain English
Customer asks for two strips of Pentakind 40 right away. Staff acknowledges but says it'll have to go on order — there isn't a single unit in stock. Customer mentions it's an emergency. No alternative brand, no similar molecule, no callback offered.
Product: Pentagon 40 (anti-epileptic)
Confidence: 95%
Wall clock: 4:25:58 PM
▶ Raw capture (what the device heard)Raw
▶ Cleaned (what the AI transcribed) — 32s from 0:00Denoised
2:04
Stockout Generic substitute — customer objects, staff does not close −₹800
Customer आपका जेनेरिक होगा ना, मैडम?
Staff जेनेरिक नहीं है। मतलब टॉप कंपनी का है। ये जो कंपनी ने जो बेंटाकन वगैरह बना दिया है ना? मतलब बॉडी में उनके भी मैन्युफैक्चरर रहते है — टॉप फोर मैन्युफैक्चरर है।
Staff बस ये है की बस वो एम आर नहीं रखा है। एडवर्टाइजमेंट नहीं करती। पिछले प्राइस का।
Customer कल का भी पेमेंट — वो जो ऑर्डर प्लेस है, उसका भी पेमेंट अवलेबल नहीं।
Plain English
Customer asks the female pharmacist whether they stock a generic version. Staff explains the only alternative is their own-label product — framed as "from the same top manufacturer, just without the MR rep and advertising cost." Customer doesn't engage with the substitute and pivots to a separate concern about pending payment on a previous order. Staff has no comeback to either question. Opportunity drifts away without closure.
Product: Generic / own-label alternative
Confidence: 90%
Wall clock: 4:27:45 PM
▶ Raw capture (what the device heard)Raw
▶ Cleaned (what the AI transcribed) — 30s at 2:04Denoised
2:10
Bounce Three-store shopper — minimal retention effort −₹500
Staff ओके?
Customer तो वही मैं बोल रहा हूँ की आप एक ही ऑर्डर मांगिए, so —
Staff दैट — हो सकता था नहीं, बना के रख देता बिल। विदाउट बिल के पेमेंट ले रहा हूँ कोई तो।
Customer क्या आप कल आप स्टॉक का जगह दे देना? मुझे कोई इश्यू नहीं।
Staff है, पता भी नहीं चलेगा।
Customer तो फिर आप एक काम करो — ऑर्डर ले।
Staff लो, ऑर्डर ले लेता हूँ।
Customer हाँ, पेमेंट मैं कर देता।
Staff हूँ, अरे —
Customer यार ये नहीं हो पाएगा। फिर मैं ऑफिस में होऊंगा ना, मैं आज जल्दी आऊं — मैं 10:30 आता हूँ।
Staff 10:30 ना।
Customer मतलब 10:30, 11 — हाँ वही मैं बोल रहा हूँ। मैं आपका फर्स्ट टाइम, इसीलिए — क्योंकि मुझे खुला दिखा है. मेरे जो आपका टाइपिंग का भरोसा नहीं।
Staff है। इनका मेडिसिन का ऑर्डर है, वो मेडिसिन अपने पास अवलेबल है — नहीं, इनको पेमेंट।
Plain English
The customer pushes to consolidate everything into one order and offers to pay upfront. He casually reveals this is his first time here — he only walked in because the shutter was open, and he doesn't trust this store's billing yet. He proposes coming back at 10:30 the next morning. Staff agrees to take the order but never positions the store's reliability, never offers a callback when stock arrives, and lets the customer dictate the terms. The trust-building moment is missed.
Products: Lovax 300mg, Levipil 1g, Pentagab
Confidence: 90%
Wall clock: 4:30:09 PM
▶ Raw capture (what the device heard)Raw
▶ Cleaned (what the AI transcribed) — 50s from 2:10Denoised
6:01
Stockout Kamlend / "नहीं है" — second item confirmed unavailable −₹800
Staff 60 एइट स्टंट ऑर्डर में नहीं रहेगा। ठीक है।
Customer हाँ चलेगा।
Staff लोएक्स — लो, और लोएक्स।
Customer 300 एमजी।
Staff हाँ — लोएक्स 300।
Customer उससे कम पावर की नहीं आती है ना?
Staff शो नहीं कर रही है।
Plain English
Staff confirms the 60mg stent variant won't make it into the order. Customer moves on to Lovax 300mg — and asks whether a lower-dose version exists. Staff replies that the system "isn't showing it." No alternative manufacturer checked, no callback offered, no escalation to the warehouse. The customer accepts the miss and the conversation moves forward.
Product: Kamlend / Penta-kind variant
Confidence: 90%
Wall clock: 4:31:23 PM
▶ Raw capture (what the device heard)Raw
▶ Cleaned (what the AI transcribed) — 26s at 6:01Denoised
9:40
Stockout Prenta Kain — "not one strip" left, no callback −₹800
Staff बोलते है — 2 दिन प्रॉडक्ट दिया, उसमें ये आपने दिया ना?
Customer हाँ, ये मैंने।
Staff अभी इनके पास नहीं है का। एक भी नोट नहीं है।
Customer मेरे पास नहीं है — मेरा ऑनलाइन सेल्लिंग है।
Staff कार्ड का।
Customer सर, आपको ₹100 का चीज़ —
Staff मिला है। फिर मुझे सुबह से रहता है ना — चेंज दे देंगे। सब सीनियर सिटीजन होता है, तो उनका ऑनलाइन रहता नहीं है। और दो अभी कहीं उनको दे देंगे, फिर उनको मना नहीं कर सकते।
Plain English
A 2-day delivery promise from a prior visit comes up. Staff confirms "they don't have it — not a single note left." Conversation drifts into a change/cash-vs-online tangent — staff explaining that many older customers don't pay online, so he keeps change ready. The original stock miss is acknowledged but never converted: no order placed, no ETA promised, no callback offer.
Product: Prenta Kain / Pentagab series
Confidence: 95%
Wall clock: 4:35:12 PM
▶ Raw capture (what the device heard)Raw
▶ Cleaned (what the AI transcribed) — 40s from 9:40Denoised
Total estimated value at risk
Per visit. All 5 events in one 14.6-min interaction.
₹3,700
4 stockouts × ₹800 + 1 bounce × ₹500
Known customer
Returning customer
Loyalty data from same-store records
₹2,336
Est. annual spend
High
Repeat risk category
2 items
Purchased on this visit
Items purchased on this visit (confirmed by billing record)
Myelostat 500mg Capsule 10s High-value ethical
Pregabica 50mg Capsule 10s Generic
Pentagon 40, generic substitute, Kamlend, Prenta Kain — requested but not fulfilled 4 unfulfilled
About the wavy underlines
word
Amber wavy — the AI thinks the transcription may have a word wrong. It shows you what it heard and what it thinks was actually said, and asks you to confirm with one tap. Here: "टाइपिंग" was likely "टाइमिंग" — the customer was saying he can't trust his job's timing to come back, not that he mistrusts the store's typing.
Staff
Slate wavy (on speaker chip) — the device's acoustic signal and the conversation context pointed to different speakers. One tap confirms who actually spoke; the system learns from your correction.
Where you see a wavy line, the AI was unsure. It shows you what it detected and what it thinks might be right — and asks you to confirm with one tap. Every correction trains the model on this store's acoustic fingerprint and conversation patterns.

Where this is heading

01 · See

The miss, while it's still hot.

The customer, the item, the rupee value at risk — surfaced before they're back in their car.

02 · Save

The recovery, queued up.

A message draft ready to send the moment that stock arrives. One tap from the manager.

03 · Steer

The shelf, what to restock.

What customers are asking for that your inventory misses — sorted by rupees, every morning.

Like what you see? Tell us about your hardest customer conversation.
hello@ostronaut.ai

Keep exploring

How it works Six stages between a conversation and a coaching rep. Read the pipeline → Customers Who already runs this on their floor. See customers → Rover The capture device behind every leak on this page. See the hardware →