Current metrics: what the selection step loses today, and on whose clock

This page measures the loss at the last step of the gig funnel: an applicant exists, nobody selects them, and the job expires. It is the first page of the auto-select timing initiative. Every number comes from the nightly copy of production, jodgig_ai. Unless a table says otherwise, the window is shifts starting 22 June to 13 September 2026, twelve full weeks, the period since the outlet flag was switched on for every outlet. Every key number carries one of three tags: fact from data, replay estimate, or assumption.

Terms used on this page:

term	meaning
no-selection job	a jod_jobs row with status: 8: it expired, it had applicants, nobody selected any of them. One row is one worker position.
live eligible applicant	a job_user row still in status IN (1, 10) at expiry, whose user is enabled, not deleted, and not user 100641, which the code always skips
a miss	a no-selection job at an outlet with locations.is_auto_select_applicant = 1 that had a live eligible applicant. This is the set auto-select could have filled.
normal job	posted 24 hours or more before the shift. It expires 8 hours before the shift.
short-notice job	posted less than 24 hours before the shift. It expires at the shift start.
manager-selected	hired_by_manager_id set and no genuine auto-select marker. A repost's auto_select_user_id is a copy of its parent's and does not count.
worked slot	a slot with a valid paid payment record. The company charge on it averaged SGD 90.89 in the window.

1. The loss: about 59 jobs a week at outlets with the flag on, worth up to SGD 119,000 a year

measure	value	tag
no-selection jobs, all outlets	1,009 jobs, 1,062 slots, about 88 slots a week	fact from data
of which had a live eligible applicant	927	fact from data
misses, meaning at an outlet with the flag on	703 jobs, about 59 a week	fact from data
no-selection jobs at outlets with the flag off	224, of which 81 had an applicant waiting at the run time	fact from data
upper bound of the loss: every miss selected and worked, at SGD 90.89 a slot	about SGD 5,300 a week, SGD 277,000 a year	assumption
what the proposed rule recovers: 452 of the 703, worked at the 66.8% auto-pick completion rate	about SGD 2,300 a week, SGD 119,000 a year	replay estimate

The impact page compares the options. The solution page explains the 452.

1.1 The number fell once, in July, and then stayed flat

No-selection slots a week, all outlets, by shift week (fact from data):

week of	slots	week of	slots	week of	slots	week of	slots
27 Apr	82	1 Jun	164	6 Jul	121	10 Aug	67
4 May	116	8 Jun	218	13 Jul	100	17 Aug	89
11 May	110	15 Jun	177	20 Jul	84	24 Aug	77
18 May	135	22 Jun	157	27 Jul	61	31 Aug	98
25 May	133	29 Jun	86	3 Aug	74	7 Sep	89

The flag was switched on for every outlet on 22 June. Auto-selected jobs jumped from about 70 a week to 222 that week and drifted down to 90 to 100 as outlets switched it off. At platform level the no-selection rate, no-selection jobs as a share of jobs with applicants, was 8.6% before auto-select existed (June to November 2025), 9.2% during the opt-in months, and 4.7% since 22 June 2026. The same summer weeks were 8.2% in 2024 and 7.7% in 2025, so the drop is not the season (fact from data). What did not move is the completion of the jobs that were selected, which is why the credits did not follow; the impact page has that comparison.

2. Who loses it: two companies carry 42% of the no-selection jobs

The ten companies with the most no-selection jobs in the twelve weeks (fact from data). "Outlets off" is the number of the company's outlets with the flag off on 15 September 2026. "After 24h" is the share of the company's manager selections made more than 24 hours after the first applicant.

company	outlets (off)	jobs with applicants	no-selection jobs	share	of which short-notice	manager picks after 24h
McDonald's	43 (0)	4,816	232	4.8%	126	33%
NTUC FairPrice (Hyper)	5 (1)	2,272	191	8.4%	103	70%
NTUC FairPrice Supermarkets (North)	12 (0)	1,820	81	4.5%	64	30%
NTUC Foodfare	16 (1)	1,646	46	2.8%	31	34%
NTUC FairPrice Supermarkets (Northwest)	11 (1)	1,314	42	3.2%	30	36%
NTUC FairPrice Supermarkets (Central)	9 (0)	944	41	4.3%	37	29%
NTUC FairPrice Supermarkets (West)	8 (0)	405	36	8.9%	32	15%
Winter Time	9 (6)	122	31	25.4%	26	9%
Good Idea Concepts	6 (0)	444	27	6.1%	25	5%
NTUC FairPrice Supermarkets (Central East)	4 (0)	694	27	3.9%	19	34%

The ten outlets with the most no-selection jobs (fact from data). "Applicant 6h or more" is the number of the outlet's misses where the first live applicant arrived 6 hours or more before the shift, so a rule with a 6-hour floor could have selected them.

outlet	company	flag	jobs with applicants	no-selection jobs	share	short-notice misses	applicant 6h or more
FairPrice Xtra Changi Business Park	NTUC FairPrice (Hyper)	off, UKG	1,631	155	9.5%	74	123
McDonald's Hougang 1	McDonald's	on	442	57	12.9%	18	33
FairPrice Northpoint City	FairPrice Supermarkets (North)	on, UKG	353	34	9.6%	25	24
FairPrice Xtra Parkway Parade	NTUC FairPrice (Hyper)	on, UKG	470	27	5.7%	20	15
FairPrice Jurong West Blk 502	FairPrice Supermarkets (West)	on, UKG	134	20	14.9%	17	13
McDonald's Plaza Singapura	McDonald's	on	206	18	8.7%	14	12
FairPrice Lot 1 Shoppers' Mall	FairPrice Supermarkets (Northwest)	off, UKG	85	16	18.8%	10	13
FairPrice City Square Mall	FairPrice Supermarkets (Central)	on, UKG	181	15	8.3%	15	12
FairPrice Yishun MRT	FairPrice Supermarkets (North)	on, UKG	216	14	6.5%	11	5
Senoko Kitchen / ICM	NTUC Foodfare	on	529	13	2.5%	8	12

"UKG" marks outlets whose jobs come from the UKG workforce system, so a hire needs a live UKG call. Changi Business Park switched the flag off around 11 August 2026; the impact page shows what that cost. McDonald's Hougang 1 has 21 misses with no eligible applicant left, the pattern where one jodder applies to several positions of the same shift and, once hired on one, is removed from the others.

3. Jodders apply around the clock, with the peak from 15:00 to 17:00 and no drop until midnight

clock band	applications to normal jobs	applications to short-notice jobs
00:00 to 05:59	7,401 (10.8%)	736 (7.4%)
06:00 to 11:59	14,567 (21.3%)	2,211 (22.3%)
12:00 to 17:59	24,465 (35.8%)	3,909 (39.4%)
18:00 to 23:59	21,902 (32.1%)	3,061 (30.9%)
total	68,335	9,917

The busiest hours are 17:00, 15:00 and 16:00, but no hour holds more than 6.2% of applications. 42% of applications arrive outside 06:00 to 17:59. Monday and Tuesday are the busiest days. The one hour that matters most for auto-select is the one after the 17:00 run: 43,304 of the 78,252 applications (55%) arrive between 17:00 and 08:59, after the evening run and before the morning run (all fact from data).

Applications by weekday and hour, counts

day	00	01	02	03	04	05	06	07	08	09	10	11	12	13	14	15	16	17	18	19	20	21	22	23	total
Mon	406	319	147	97	79	109	126	244	371	530	695	666	763	724	806	934	692	709	696	773	648	767	670	622	12,593
Tue	410	233	254	164	81	123	212	366	329	389	752	677	698	684	749	733	880	864	723	752	893	824	800	575	13,165
Wed	513	253	175	93	79	122	149	311	468	544	676	700	754	770	703	640	703	877	670	585	778	596	711	596	12,466
Thu	322	250	153	126	71	110	171	287	301	403	497	658	624	855	616	658	625	704	552	528	556	544	517	516	10,644
Fri	484	308	185	66	98	78	138	282	343	435	547	658	719	609	713	701	697	555	455	535	422	469	542	585	10,624
Sat	493	247	153	117	79	73	108	238	271	325	385	546	578	542	548	630	504	545	558	458	538	473	547	502	9,458
Sun	336	226	167	139	109	90	114	168	328	346	499	525	436	539	494	540	637	622	414	447	415	441	617	653	9,302
total	2,964	1,836	1,234	802	596	705	1,018	1,896	2,411	2,972	4,051	4,430	4,572	4,723	4,629	4,836	4,738	4,876	4,068	4,078	4,250	4,114	4,404	4,049	78,252

4. Hiring managers keep office hours and take their time on normal jobs

Selection time is the earliest slot_user.created_at on the job. 17,754 manager-selected jobs starting in the window, 14,929 normal and 2,825 short-notice (fact from data).

clock band of the selection	normal jobs	short-notice jobs
00:00 to 05:59	509 (3.4%)	131 (4.6%)
06:00 to 11:59	4,777 (32.0%)	683 (24.2%)
12:00 to 17:59	6,103 (40.9%)	1,103 (39.0%)
18:00 to 23:59	3,540 (23.7%)	908 (32.1%)

Hours from the first application on the job to the selection:

job kind	lower quartile	median	upper quartile	within 1 hour	over 24 hours
normal	2.1 h	15.5 h	44.1 h	2,706 (18.1%)	5,390 (36.1%)
short-notice	0.03 h	0.5 h	2.5 h	1,688 (59.8%)	0

Hours from the selection to the shift start:

job kind	median	under 6 hours	6 to 24 hours	over 24 hours
normal	129 h	0	476 (3.2%)	14,453 (96.8%)
short-notice	8.7 h	1,244 (44.0%)	1,581 (56.0%)	0

Three things follow. Managers select when they are at work: 71% of selections fall between 06:00 and 17:59, while 42% of applications arrive outside those hours. On a normal job they wait: a median 15.5 hours after the first applicant, and 36% wait more than a day. Their speed on short-notice jobs comes from the short notice, not from a habit: 60% select within an hour because the shift is hours away. No normal job is ever selected inside the last 6 hours, because it expired 8 hours before the shift. Only 2.1% of normal-job selections happen after the day-before run has passed, so on normal jobs the run has almost nothing left to take from a manager. On short-notice jobs 97.7% of selections come after it, because the job did not exist when the run time passed (all fact from data).

5. The two clocks do not meet where the run is

Applications keep arriving until midnight. Managers stop selecting in the evening. The one automatic attempt runs at 09:00 or 17:00 the day before the shift, before 55% of the applications, and for every shift starting before 17:00 the next morning's run comes after the job has already expired. The problems page puts the three clocks on one timeline and counts where the 703 misses fall.