When the pipeline moves — hour-of-day traffic by stage, Sep 2025 to Aug 2026

Every stage timestamp in the loan dump for the last twelve months, real files only (feeds and bulk lead-initiators excluded). Cell = share of that stage's volume falling in the hour; darker = busier. The ADR report carries dates without times, so ADR raise/resolve hours are not here — that timestamp should be added to the Metabase query.

What the clock says

The whole chain runs in the evening because the desk does. RMs initiate at midday, complete applications at 16–18h, the desk logs 17–19h (27% of all logins), U1s decide 18–19h (31% after 18h), ops authorise and pay at 18h (30% of disbursals after 18h). Each stage inherits the previous one's evening and adds a night.
Login hour does not buy the customer anything. A file logged before noon waits a median 56 hours for U1 and is decided the same day 14% of the time; a file logged at 17h waits 69 hours, 2%. There is no same-day U1 lane at all — the queue is measured in days, so shifting the desk earlier only helps if U1 commits to a next-morning clear.
Biz UW is a morning stage that exists to absorb the evening. It peaks 10–13h processing the previous night's U1 output and delivers to U2 by noon; U2 therefore has no morning. Removing the gate gives U2 back its mornings — half a working day on every file.
Payment is a single end-of-day batch. Both authorisations and disbursal peak in the 18h hour; 19% of disbursals fall on Saturday. A 13:00 batch would move roughly a third of disbursals a day earlier and take the Friday/Saturday bunching out of treasury.
People, not policy, decide the hours. On the desk, Rikin Tiwari and Shubham Pawar log 31–33% after 18h and 6–8% after 20h; Sagar Holkar 9% after 18h on the same volume. Among U2s, Sudheer Kumar takes 23% of decisions after 20h and 19% on weekends; Ravi Sharma 1% and 16%. Shifts should be set, not inherited.

1 Traffic by stage and hour

Read each row left to right. The percentage columns and the reading on the right are what to act on.

Stage (actor)	08	09	10	11	12	13	14	15	16	17	18	19	20	21	22	23	00–07	after 18h	after 20h	Sat+Sun	Read and fix
Initiated (RM) n=36,257 · peak 12h	1	3	7	12	14	12	10	10	9	8	6	3	2	1	0	0	0	13%	4%	18%	RMs open files at midday, not on the morning visit. Move initiation to the customer visit (app, geo check-in) — it is the first timestamp on the file and sets the clock.
Application done (RM) n=29,511 · peak 17h	1	3	8	8	10	10	10	10	11	12	9	4	2	1	1	0	0	16%	4%	16%	Completed 16–18h, four hours after initiation. With the app gate this is the same event as initiation.
Login done (desk) n=27,706 · peak 17h	0	1	5	9	10	8	10	11	12	15	13	4	2	1	0	0	0	20%	4%	14%	27% of logins land 17–19h, 6% before 10h. The desk clears the day's pile in the evening; U1 sees it two days later. Two desk shifts (10–14, 14–20) and a 20:00 cut-off with everything logged by then guaranteed to U1 by noon next day.
U1 decision n=24,672 · peak 18h	1	2	4	6	7	7	7	10	11	12	14	6	3	2	2	2	2	31%	10%	16%	31% after 18h, 10% after 20h — U1s work evenings because the queue arrives in the evening. Fix the desk timing and U1 moves into the day.
Biz UW n=23,310 · peak 12h	0	5	10	12	13	8	7	9	8	10	9	4	2	1	0	0	1	16%	3%	15%	Morning stage (10–13h): processes the previous evening's U1s. Removing the gate removes a full working day here.
U2 decision n=14,053 · peak 15h	0	1	6	9	11	9	10	11	11	11	9	4	2	1	1	1	1	19%	6%	16%	Flat 11–18h, 19% after 18h, 16% on weekends. No morning start — the previous day's Biz UW output is only ready by noon. A 10:00 U2 queue review would pull half a day.
Sanction n=12,590 · peak 12h	0	2	6	9	13	9	10	12	11	12	9	3	1	1	1	1	1	16%	5%	15%	Auto on U2; mirrors U2. Nothing to fix here once U2 moves.
Rejection n=9,815 · peak 12h	0	5	9	16	18	9	8	10	6	8	6	1	1	1	0	0	0	9%	2%	13%	Peaks 11–12h: rejections are a morning batch, mostly U1s clearing overnight files. Fine.
Ops done (docs) n=8,622 · peak 17h	0	2	7	8	9	8	9	11	12	13	12	5	2	1	0	0	0	20%	3%	18%	Evening-weighted (17–19h). Docket completion follows sanction by 9 days — the issue is days, not hours.
Payment auth 1 n=8,443 · peak 18h	0	0	2	6	8	4	8	13	14	13	17	8	3	2	0	0	0	31%	5%	19%	Payment authorisation is an end-of-day batch: 30% after 18h, 17% in the 18h hour alone. Two batches (13:00 and 18:00) would move a third of disbursals a day earlier.
Payment auth 2 n=8,443 · peak 18h	0	1	1	5	8	5	6	13	14	13	18	9	4	2	0	0	0	33%	6%	19%	Same batch, minutes after the first. Second signature adds no time — keep.
Cheque printing n=8,614 · peak 14h	0	2	8	10	10	4	12	12	11	12	12	4	1	1	0	0	0	18%	3%	19%	Bimodal 10–12h and 14–18h; prints the previous evening's authorisations. A 13:00 payment batch feeds a same-day print.
Disbursed (OTC) n=8,445 · peak 18h	0	0	3	7	7	4	9	14	13	13	17	8	3	1	0	0	0	29%	5%	19%	29% of disbursals after 18h, 19% on Saturdays, 17% on Fridays. End-of-day and end-of-week — treasury and customer both get the money a day late.

Login hour	Logins	Median hours to U1 decision	U1 decided same day
<12	3,789	56 h (2.3 d)	14%
12–13	4,295	56 h (2.3 d)	9%
14–15	5,092	66 h (2.8 d)	6%
16–17	6,634	69 h (2.9 d)	2%
18–19	3,914	70 h (2.9 d)	1%
20+	854	67 h (2.8 d)	0%

It does not, today. The median wait is 2.3–2.9 days regardless of when the file is logged, and same-day U1 is under 15% even for morning logins. The rule to introduce: everything logged by 20:00 is at U1 by 12:00 the next day. The desk's 17–19h peak then becomes U1's 10–12h peak, and U1 stops working evenings.

3 Logins by zone and hour

Logins by zone	09	10	11	12	13	14	15	16	17	18	19	20	21	22	23	after 18h
NORTH n=6,884	1	5	9	11	8	11	11	13	15	12	2	2	2	0	0	18%
SOUTH n=9,394	1	6	9	10	9	9	11	12	14	11	4	2	1	1	1	19%
WEST 1 n=9,330	1	5	9	9	7	10	11	12	16	14	3	1	0	0	0	18%
WEST 2 n=2,531	1	6	8	6	5	7	9	10	16	16	6	4	3	1	0	31%

West 2 (Gujarat) logs 31% of its files after 18h against 18–19% elsewhere — one desk person, evening-loaded. Zone profiles are otherwise identical, which says the pattern is the desk's, not the RMs'.

4 Desk hours by person

Login desk person	Zone	Logins	Before 12h	After 18h	After 20h	Read
Shubham Pawar	WEST 1	709	7%	33%	6%	Evening-loaded — candidate for the late shift, needs a morning counterpart
Rikin Tiwari	WEST 2	2,423	16%	31%	8%	Evening-loaded — candidate for the late shift, needs a morning counterpart
Haripriyan B	SOUTH	2,326	17%	27%	4%	Evening-loaded — candidate for the late shift, needs a morning counterpart
Keshav Verma	WEST 1	2,101	11%	21%	1%	Afternoon-only pattern; nothing logged in the morning
Naresh Baraskar	WEST 1	2,141	12%	21%	2%	Afternoon-only pattern; nothing logged in the morning
Naveen	NORTH	2,495	13%	21%	6%	Afternoon-only pattern; nothing logged in the morning
Sanket Kshirsagar	WEST 1	1,879	15%	20%	1%	Afternoon-only pattern; nothing logged in the morning
Kapil	NORTH	2,542	13%	18%	3%	Afternoon-only pattern; nothing logged in the morning
Vamsikrishna Nalla	SOUTH	1,241	19%	18%	3%	Afternoon-only pattern; nothing logged in the morning
Saidavali Miriyala	SOUTH	3,026	19%	17%	6%	Afternoon-only pattern; nothing logged in the morning
Suraj Patil	SOUTH	2,780	12%	16%	3%	Afternoon-only pattern; nothing logged in the morning
Shivam Shukla	NORTH	1,771	17%	15%	1%	Afternoon-only pattern; nothing logged in the morning
Sagar Holkar	WEST 1	2,345	25%	9%	1%	Balanced

5 U2 hours by person

U2 (regional credit)	Zone	Decisions	Before 12h	After 18h	After 20h	Weekend
Ravi Sharma	NORTH	1,677	18%	10%	1%	16%
Deepak Dixit	NORTH	1,362	10%	29%	8%	15%
Ramchandra Sarvankar	WEST 1	1,131	15%	19%	6%	17%
Ajay Kumar	WEST 1	1,107	28%	12%	4%	17%
Sudheer Kumar	SOUTH	1,107	13%	38%	23%	19%
Jay Limbad	WEST 2	1,080	17%	24%	6%	16%
Amol Deshmukh	WEST 1	971	18%	11%	0%	17%
Sudarshan Reddy	SOUTH	775	15%	28%	7%	19%
Safik S	SOUTH	725	15%	17%	3%	11%
Amol Wakode	SOUTH	593	9%	6%	0%	15%
Sachin Choudhari	WEST 1	576	20%	23%	7%	21%
Krishna A	SOUTH	535	14%	38%	14%	11%
Pranav Jogdand	SOUTH	460	31%	3%	0%	18%
Keyur Makwana	WEST 2	366	22%	23%	10%	14%
Kesab Pal	SOUTH	325	22%	2%	0%	19%
Harshal Pardhi	WEST 1	317	30%	11%	0%	13%
Aniket Modi	WEST 1	268	28%	6%	0%	14%
Pradeep Jadhav	WEST 1	255	18%	5%	0%	8%
Holagunda Bharateesha	SOUTH	204	23%	24%	5%	17%

Weekend decisions run 11–21% for every U2 — Saturday is a working day in credit, which is fine, but 8–23% after 20h is queue pressure showing as overtime. The three with the longest medians in the main report (Amol Deshmukh, Ajay Kumar, Sachin Choudhari) are not the late-night ones; they are the ones whose queue simply exceeds the day.

Share of week	Mon	Tue	Wed	Thu	Fri	Sat	Sun
Initiated	17	17	16	15	15	13	7
Login	19	18	18	16	16	12	2
U1	16	17	17	17	17	13	3
U2	17	16	18	16	17	13	3
Sanction	17	16	18	17	17	13	2
Payment auth	12	16	17	17	19	17	2
Disbursed	12	16	17	17	19	17	2

Logins front-load the week (Monday 19%), disbursals back-load it (Friday 19%, Saturday 17%). Money leaves the building on Friday evening and Saturday for files that were logged on Monday — the week is the unit of time the pipeline actually runs on.

7 The long view — what has changed about the clock since FY22

Same measures by financial year of the timestamp, real files only. FY27 (*) is April–August 2026.

Share of stage volume	FY22	FY23	FY24	FY25	FY26	FY27*
Login n: FY22 3,199 · FY23 6,323 · FY24 9,882 · FY25 14,937 · FY26 26,213 · FY27 11,218
Before 12h	6%	8%	9%	10%	13%	19%
After 18h	56%	49%	40%	28%	23%	15%
After 20h	17%	14%	8%	5%	5%	2%
U1 n: FY22 3,029 · FY23 5,552 · FY24 8,308 · FY25 13,052 · FY26 22,630 · FY27 10,499
Before 12h	15%	15%	15%	14%	14%	15%
After 18h	35%	31%	31%	31%	31%	30%
After 20h	11%	10%	9%	9%	10%	10%
U2 n: FY22 2,161 · FY23 3,739 · FY24 5,498 · FY25 8,043 · FY26 12,867 · FY27 6,451
Before 12h	11%	11%	11%	14%	17%	17%
After 18h	24%	22%	23%	21%	19%	20%
Payment auth n: FY22 967 · FY23 2,050 · FY24 3,170 · FY25 3,504 · FY26 7,378 · FY27 3,396
After 18h	72%	63%	44%	28%	33%	30%
After 20h	51%	45%	28%	10%	7%	4%
Sat+Sun	24%	28%	36%	23%	20%	18%
Disbursed n: FY22 539 · FY23 860 · FY24 3,011 · FY25 3,919 · FY26 7,349 · FY27 3,396
After 18h	28%	24%	17%	24%	30%	26%
Sat+Sun	13%	18%	18%	21%	20%	18%
Initiated n: FY22 3,746 · FY23 7,498 · FY24 12,650 · FY25 19,993 · FY26 33,961 · FY27 15,308
Before 12h	15%	16%	16%	20%	23%	24%
Login → U1 hand-off
U1 decided same day	2%	2%	3%	4%	5%	8%
U1 decided by next noon	3%	3%	6%	6%	7%	10%
Median hours login → U1	70 h	90 h	73 h	71 h	67 h	54 h

Two functions moved into the day; one never did. The login desk in FY22 did 55% of its logins after 18h and 17% after 20h; in FY27 it is 15% and under 2%, with morning logins up from 6% to 19% — the desk has already made the shift this report asks for, on its own. Payment authorisation has gone further: 51% of authorisations after 20h in FY22 (peak hour 22h in FY23–FY24), 4% now. U1 has not moved in five years: 15% before noon and 30–35% after 18h in every year, peak hour 18h throughout. The desk got earlier and U1 did not follow, so the same-day hand-off that could have opened up never did — same-day U1 is 2% → 8%, next-noon 3% → 10%, and the median wait is still 54 hours. The queue between login and U1 has been two to three days since FY22. It is not a symptom of FY26 volume; it is how the two teams have always been scheduled against each other. Weekend work has fallen in every credit stage (18% → 14%) but risen at disbursal (13% → 18–20%) — ops has moved out of the night and into Saturday.

What this changes in the recommendation. The desk timing fix is largely done; the remaining desk item is consistency between people (Section 4), not the shift. The lever is U1: a next-morning clear rule is the first change to U1's day in five years and it is the one that converts the desk's earlier hours into a shorter queue.

Timing changes, in one table

Change	Who	Metric
Initiation at the customer visit via app (geo check-in, e-KYC, fee) — not at the branch at noon	Sales / CTO	Initiated peak 12h → 10–11h; initiated → login median 1.6 → same day
Desk in two shifts (10–14, 14–20) with a 20:00 cut-off; one consolidated checklist query per file	COO / desk lead	Logins before 12h: 6% → 30%; rounds per file 2.6 → <1.5
U1 next-morning clear: everything logged by 20:00 decided by 12:00 — the first change to U1's day since FY22	CRO / BCMs	U1 by next noon 10% → 80%; U1 after 18h 30% → <10%
Remove Biz UW; U2 starts the day at 10:00 on yesterday's U1 output	CRO	U2 decisions before 12h 15% → 40%
Two payment batches, 13:00 and 18:00; no Saturday-only disbursals	COO / Treasury	Disbursals after 18h 29% → 15%; Fri+Sat share 36% → 28%
Add time-of-day to the ADR report so raise/resolve hours can be measured	CTO	—