18% → 69.3%
Night Chats Answered Within an Hour
Before: 28 staff-only days, 30 September to 28 October. Now: the dashboard’s assistant-live comparison. About 3.9× the share answered, not a sales multiplier.
Case study · An independent clothing retailer
Over a month of connected orders, couriers, stock and reconciliation. Since 28 October 2026, the Kotha assistant has also answered customers and taken orders in chat.
October 2026 full-month operations plus November to date, through about 9 November. Chat dashboard: last 30 days, including 12+ live days since 28 October. Client identity withheld.
18% → 69.3%
Before: 28 staff-only days, 30 September to 28 October. Now: the dashboard’s assistant-live comparison. About 3.9× the share answered, not a sales multiplier.
138
Last-30-days chat view: ৳2,16,489 in order value. 106 with no person in the chat, worth ৳1,67,029. Orders taken are not completed or incremental sales.
258,408
Counted receipts for routine work. November to date: 77,471 jobs. Jobs are counts, not hours saved.
100%
1,177 of 1,177 invoiced parcels matched to an order. The dashboard prices monthly reconciliation at 10 hours, ৳950 (estimate), using ৳95 per staff hour.
47 sec
Last-30-days chat view: 86% within 5 minutes. The before/now comparison separately shows 81.5% → 86.2% within 5 minutes.
1,289
Out of 4,806 chats overall, about 27%. The assistant answered 1,802 chats; about 72% of those had no person. These are different denominators.
Orders lived in a daily spreadsheet. A parcel had three separate reference numbers, and stock had three separate records that did not agree. Staff kept answering the same inbox questions while checking orders, making calls and packing parcels. Before the assistant, a customer writing between 02:00 and 10:00 waited 2 hours 40 minutes in the middle case.
One client’s dashboard, not a controlled experiment or a guarantee. Operations have run on dekhval for over a month; the assistant has been live since 28 October. The last-30-days view includes staff-only days; its 4,806 chats are not all assistant-live chats. Before/now reply comparisons use the stated 28-day staff-only baseline, 30 September to 28 October. October operations cover the full month; November is partial, through about 9 November. Return rates describe parcels with outcomes, not a fixed cohort. Order-status bars total 164, a different scope from the 138 headline orders, so they are not a fulfilment rate for those 138. The October reconciliation detail has inconsistent raw counts; the 100% parcel match shown here uses November’s unambiguous 1,177/1,177. Time and taka are dashboard estimates: reconciliation 10 hours/month; export 27 minutes/day; staff hour ৳95, from ৳20,000–25,000/month over 26 days × 9 hours. Job receipts are measured counts. No incremental paid revenue, payroll saving or staff replacement is established.
Historical context from the earlier report: all available customer inbox history, with exact date bounds unspecified. This unchanged hourly series is not the new last-30-days view or October/November totals. Each bar is one local hour, 00:00 to 23:00.
The earlier series recorded 2,864 messages in the midnight hour and 821 at 09:00. It shows when customers wrote, not hours saved or the number of missed orders.
0:00 2,864 Messages
1:00 1,757 Messages
2:00 713 Messages
3:00 345 Messages
4:00 218 Messages
5:00 132 Messages
6:00 175 Messages
7:00 324 Messages
8:00 553 Messages
9:00 821 Messages
10:00 2,743 Messages
11:00 3,661 Messages
12:00 3,827 Messages
13:00 3,666 Messages
14:00 4,211 Messages
15:00 3,571 Messages
16:00 3,231 Messages
17:00 3,319 Messages
18:00 3,369 Messages
19:00 3,906 Messages
20:00 3,427 Messages
21:00 3,823 Messages
22:00 3,912 Messages
23:00 3,627 Messages
Answers stock and product questions, takes chat orders and hands uncertain cases to people. Of 379 handoffs, 366 were picked up, 223 within 15 minutes. In 176 chats an order was read back; 158 customers said yes (90%).
Connects orders, courier updates and dispatch. October logged 27,628 courier updates and 8,132 labels printed. Nightly exports ran 31 times, replacing an estimated 27 minutes by hand each day: about 14 hours, ৳1,325 (estimate).
Connects stock, payments and reconciliation. October’s priced jobs totalled 24 hours, ৳2,275 (dashboard estimate) at ৳95 per staff hour. November to date: 13.6 hours, ৳1,292 (estimate). These are estimated labour values, not measured cash savings; most job types are not priced.
A customer asks about a size or price. Kotha checks product information and stock, replies and can take the order. People handle confirmation calls, packing and exceptions. The last-30-days view covers Messenger: 4,002 chats (83%, assistant alone 25%); WhatsApp: 613 (13%, alone 39%); Instagram: 191 (4%, alone 21%). Orders linked within seven days of a chat totalled 510, worth ৳8,46,959; 178 orders worth ৳2,90,089 followed an assistant answer, including 26 worth ৳42,709 after night answers. Phone numbers were known for only 26% of chats. These groups overlap and are associations, not extra sales caused by the assistant.
Tell us where work gets delayed and which result matters most. We will discuss what to connect first and how to measure the change.