Job Board pFR Deep Dive | SEA Markets

0 Impact Sizing: What Happens If 1–2h FR Reaches 96%?

The 1–2h bucket is 11% of all Normal AB URs but is the only bucket below 92% FR

SEA overall pFR (all Normal AB) is 95.6%. The 1–2h bucket sits at 87.3% — an 8.3pp gap vs the rest. Every other bucket fulfils at 92–97%. If the 1–2h bucket matched each market's own overall pFR, SEA total would lift to 96.5% (+0.9pp). Market-level detail below.

1–2h Bucket pFR (Current)

87.3%

SEA, Normal AB (Airport DO + Others) — lowest FR of any bucket

Overall pFR (Current)

95.6%

SEA, all Normal AB — May 1–25 2026

Overall pFR if 1–2h Matches Overall

96.5%

+0.9pp SEA uplift — Vietnam gains +2.1pp, Malaysia +1.0pp

Market-Level Uplift — If 1–2h FR Matches Each Market's Own Overall pFR

🇻🇳 Vietnam — biggest prize

78.3% → 91.6%

Overall pFR: 91.6% → 93.7% (+2.1pp)

🇲🇾 Malaysia

87.0% → 96.2%

Overall pFR: 96.2% → 97.2% (+1.0pp)

🇮🇩 Indonesia

87.5% → 94.7%

Overall pFR: 94.7% → 95.6% (+0.9pp)

🇵🇭 Philippines

90.3% → 95.2%

Overall pFR: 95.2% → 95.8% (+0.6pp)

🇹🇭 Thailand

90.4% → 95.9%

Overall pFR: 95.9% → 96.5% (+0.6pp)

🇸🇬 Singapore

90.3% → 97.5%

Overall pFR: 97.5% → 98.0% (+0.5pp)

Vietnam is the biggest prize — its 1–2h bucket alone contributes +3.2pp gap to Vietnam's overall pFR

Vietnam's 1–2h FR is only 78.3% (lowest SEA), with an overall pFR of 91.6%. Closing the 1–2h gap alone would move Vietnam's overall pFR to 93.7% (+2.1pp) — the largest market-level uplift opportunity. Malaysia (87.0% → 96.2% target, +1.0pp overall) and Indonesia (87.5% → 94.7%, +0.9pp) follow.

1 Executive Summary

Problem Statement

Job Board (JB) Fulfilment Rate is consistently lower for 1–2 hour lead-time bookings vs 2–6 hour bookings across all SEA markets. The table below sizes every stage of the funnel that explains the gap and highlights the key insights at both regional and market level.

1–2h Bucket pFR (Current)

87.3%

Lowest FR of any lead-time bucket across SEA

Overall pFR (Current)

95.6%

All Normal AB, 6 SEA markets, May 1–25 2026

Overall pFR if 1–2h Matches Overall

96.5%

+0.95pp lift — 1–2h bucket is 11% of URs but 100% of the bottom

Key Findings: 1 JB Post-Alloc No-Show 2 FA: DAX Offline / Invalid State (Morning Peak) 3 Fallback Recovery (DAX Post-Cancel)

Funnel Stage	Gap Contribution — Regional	Gap by Market	Overall pFR Uplift if fully fixed (SEA)	Key Insights
Priority Finding 1 JB Post-Allocation No-Show Problem PAX cancels after allocation, within 30 min of pickup — no-show rate is 2–4× higher in 1–2h vs 2–6h	17–47% of FR gap across SEA markets SEA avg (ex-SG): ~21%	🇸🇬 SG 47% 🇵🇭 PH 31% 🇻🇳 VN 20% 🇲🇾 MY 19% 🇮🇩 ID 22% 🇹🇭 TH 17% Contribution = (NS Rate 1–2h − NS Rate 2–6h)	+0.38pp SEA overall pFR uplift 🇻🇳 VN +0.73pp 🇵🇭 PH +0.68pp 🇸🇬 SG +0.46pp 🇮🇩 ID +0.40pp 🇲🇾 MY +0.39pp 🇹🇭 TH +0.17pp	SG is the outlier (47%) — supply barely fails in SG; almost all of SG's gap is demand-side abandonment. For all other markets, no-show is a consistent but secondary contributor (~17–22%). No-show rate in 1–2h is 2–4× the 2–6h rate (e.g. MY: 2.5% vs 0.7%) — passengers who book close to their ride time are far more likely to cancel.
⚡ FAILED ALLOCATION — Driver fails the pre-award check 30 min before pickup. A driver always signed up; the failure is at the T−30min broadcast check, not at sign-up.
FA: Overall Net FA contribution = FA rate gap × (1 − fallback recovery). Portion of FR gap not rescued by fallback	4–66% of FR gap across SEA markets MY: 5.4pp (49%) \| TH: 3.3pp (66%)	🇹🇭 TH 66% 🇵🇭 PH 62% 🇲🇾 MY 49% 🇻🇳 VN 31% 🇮🇩 ID 28% 🇸🇬 SG 4% % of total FR gap attributable to net unrecovered FA failures	+0.63pp SEA overall pFR uplift 🇵🇭 PH +1.04pp 🇻🇳 VN +1.00pp 🇲🇾 MY +0.91pp 🇹🇭 TH +0.50pp 🇮🇩 ID +0.42pp 🇸🇬 SG +0.03pp	Dominant lever in supply-first markets. TH (66%), PH (62%), and MY (49%) have FA as the biggest single contributor to their FR gap. FA rate in the 1–2h bucket is 2–3× higher than 2–6h: TH 11.7% vs 3.5%; MY 25.2% vs 7.8%; PH 14.0% vs 6.8%. SG is the structural outlier — FA contributes only 4% of SG's gap; the gap is almost entirely PAX No-Show (58%) and Fallback recovery (38%). Vietnam's high FA rate (20.5%) is split evenly between driver unavailability and Too Far — Bangkok/Vietnam geography amplifies TooFar failures more than other markets.
Priority Finding 2 Failed Allocation — DAX Offline / Invalid State Driver offline, suspended, or system-disqualified after JB acceptance · Morning peak amplifies Invalid State 2× (53% of FAs at 04–07h vs 27% off-peak)	3–52% of FR gap across markets MY: 4.6pp (41%) \| TH: 2.6pp (52%) Offline rate 1–2h vs 2–6h: MY 20.9% vs 6.3%	🇹🇭 TH 52% 🇵🇭 PH 46% 🇲🇾 MY 41% 🇻🇳 VN 23% 🇮🇩 ID 23% 🇸🇬 SG 3% 80–85% of FA failures in most markets are Offline/Unavailable (excl. VN at 55%)	+0.53pp SEA est. pFR uplift 🇲🇾 MY +0.77pp 🇵🇭 PH +0.78pp 🇻🇳 VN +0.74pp 🇹🇭 TH +0.42pp 🇮🇩 ID +0.34pp 🇸🇬 SG +0.02pp	Dominant FA sub-category everywhere except Vietnam. 80–85% of all FA failures in MY, ID, PH, TH are from drivers who went offline or entered an invalid state after accepting the JB. FA offline rate is 3–4× higher in 1–2h than 2–6h (MY: 20.9% vs 6.3%). Vietnam anomaly: Only 55% of VN FA failures are Offline/Unavailable; the remaining 45% are Too Far — reflecting greater geographic spread of drivers relative to pickup zones in Vietnam cities. Lever: Validate driver eligibility at JB sign-up; re-alert at T−45min; penalise post-acceptance offline transitions.
FA: Too Far Driver moved too far from the pickup zone by the T−30min allocation check	1–15% of FR gap across markets MY: 0.9pp (8%) \| VN: 1.2pp (8%) TooFar rate 1–2h vs 2–6h: VN 9.2% vs 6.7%	🇵🇭 PH 15% 🇹🇭 TH 14% 🇻🇳 VN 8% 🇲🇾 MY 8% 🇮🇩 ID 6% 🇸🇬 SG 1% % of total FR gap attributable to Too Far failures	+0.10pp SEA est. pFR uplift 🇻🇳 VN +0.26pp 🇵🇭 PH +0.25pp 🇹🇭 TH +0.08pp 🇲🇾 MY +0.08pp 🇮🇩 ID +0.07pp 🇸🇬 SG +0.01pp	Secondary lever; geography-driven. Too Far accounts for 15–25% of all FA failures across markets. The share is highest in Vietnam (45% of FA) because driver coverage zones are more spread out; lowest in SG (27%) but SG's overall FA rate is negligible. TooFar rate is relatively flat between 1–2h and 2–6h buckets — it contributes to the FR gap mainly because the overall FA rate is higher in 1–2h. In VN and TH, the TooFar gap is more pronounced (rate: VN 9.2% vs 6.7%; TH 2.8% vs 1.1%). Lever: Geofence enforcement on JB sign-up; progressive unsubscription warnings when driver moves beyond service radius.
Priority Finding 3 Fallback Recovery Gap Fallback rescues failed JB bookings at a lower rate in 1–2h vs 2–6h · Root cause: DAX post-cancellation after fallback acceptance (17–33% rate vs 1–6% in 2–6h)	35–60% of FR gap across SEA markets MY: 4.1pp (44%) of gap MY rescue rate: 77.5% vs 92.6%	🇮🇩 ID 60% 🇸🇬 SG 51% 🇻🇳 VN 47% 🇲🇾 MY 44% 🇹🇭 TH 35% 🇵🇭 PH ~0% Share of total FR gap explained by fallback rescue gap	+0.49pp SEA overall pFR uplift 🇻🇳 VN +1.45pp 🇮🇩 ID +0.68pp 🇲🇾 MY +0.55pp 🇸🇬 SG +0.30pp 🇹🇭 TH +0.09pp 🇵🇭 PH ~0pp	Single largest lever — root cause identified. Fallback recovery drops 15pp in MY (77.5% vs 92.6%) and 16pp in ID (69.9% vs 85.3%) for 1–2h bookings. 🔴 Root Cause: Driver post-cancellation after fallback acceptance. Drivers accept the fallback booking but then cancel at 17–33% rates for 1–2h vs only 1–6% for 2–6h. This gap exists at every hour of the day — it is not a morning peak issue. Fallback allocation rate is actually better in 1–2h; PAX behaviour is not the issue. Lever: Fallback commitment enforcement (cancellation penalty), fare uplift on fallback assignment, ETA visibility before acceptance. See Section 6 for full deep-dive.

Market contribution % = component's pp contribution ÷ total FR gap for that market. Market-level deep-dives in Section 4.

FR Gap Reference — All Markets

Market	FR (1–2h)	FR (2–6h)	FR Gap ▼	FA Rate (1–2h)	FA Rate (2–6h)	FB Recovery (1–2h)	FB Recovery (2–6h)	Archetype
🇻🇳 Vietnam	78.3%	92.2%	+13.9pp	20.6%	8.9%	60.5%	65.0%	Fallback + FA
🇲🇾 Malaysia	87.0%	96.4%	+9.4pp	23.6%	8.0%	77.5%	92.6%	FA Supply
🇮🇩 Indonesia	87.5%	95.2%	+7.7pp	10.4%	5.7%	69.9%	85.3%	Fallback dominant
🇸🇬 Singapore	90.3%	97.3%	+7.0pp	2.8%	2.3%	72.8%	90.7%	PAX No-Show
🇰🇭 Cambodia	88.0%	94.3%	+6.3pp	9.0%	12.1%	—	—	Low volume
🇹🇭 Thailand	90.4%	95.2%	+4.8pp	10.2%	3.4%	66.2%	80.3%	FA Supply
🇵🇭 Philippines	90.3%	93.8%	+3.5pp	21.4%	7.3%	82.8%	93.6%	FA Supply

JB FA Rate = jb_failed_allocation / total_request | FB Recovery = fallback_completed / fallback_request

2 Gap Decomposition

The FR gap decomposes into three additive components. Bars below show the proportional contribution of each for every market.

PAX No-Show FA: Offline / Invalid State FA: Too Far Fallback Recovery Gap

🇻🇳 Vietnam — 13.9pp gap

2.8pp
20%

3.2pp
23%

1.5pp
11%

6.5pp
47%

🇲🇾 Malaysia — 9.4pp gap

1.8pp
19%

3.0pp
32%

0.6pp
6%

4.1pp
44%

🇮🇩 Indonesia — 7.7pp gap

1.7pp
22%

1.0pp
13%

0.4
5%

4.6pp
60%

🇸🇬 Singapore — 7.0pp gap

3.3pp
47%

3.6pp
51%

🇹🇭 Thailand — 4.8pp gap

0.8pp
17%

1.7pp
36%

0.5pp
11%

1.7pp
35%

🇵🇭 Philippines — 3.5pp gap

1.1pp
31%

2.0pp
58%

0.4pp
11%

🔴 Supply-Driven Markets

Malaysia · Thailand · Philippines · Vietnam

FA Offline/Invalid State is the dominant JB-stage driver. Fixing driver state at the pre-award check directly closes the gap. FA rate in 1–2h is 2–3× the 2–6h rate.

🟡 Fallback-Driven Markets

Indonesia · Singapore (+ Vietnam partially)

The JB booking itself barely fails differently between buckets. The gap is driven by fallback rescuing at a much lower rate in the 1–2h window. Requires a separate investigation.

3 Failed Allocation Deep Dive

Key Structural Finding

100% of failed alloc URs had a driver sign up for the JB. The failure always occurs at the T-30min pre-award broadcast check — never at sign-up. No intervention at the sign-up stage alone would help; the intervention must target what happens to the driver in the 30-minute window before pickup.

Failed Allocation Rate — All Markets, 1–2h vs 2–6h

All 6 SEA Markets · May 1–25 2026 · Normal AB, Non-Airport | Red = 1–2h · Green = 2–6h · Annotations show the excess gap (pp) for 1–2h over 2–6h

What the chart shows

Left — Total FA Rate: Malaysia (25.2% vs 7.8%) and Vietnam (20.5% vs 10.7%) have the widest 1–2h FA rate gaps. Singapore is structurally flat at ~2–3% — supply is not the constraint there.

Centre — Offline/Unavailable: This sub-reason dominates in every market except Vietnam. MY offline rate is 3.3× higher in 1–2h (20.9% vs 6.3%). In TH and PH the gap is similarly pronounced.

Right — Too Far: Vietnam is the clear geography outlier — 9.2% vs 6.7%, and Too Far accounts for 45% of Vietnam's FA failures. For all other markets Too Far is a secondary lever (1–4%).

Key takeaway: For MY/TH/PH/ID, fix driver availability and state at the pre-award check. For VN, combine availability fixes with geofencing — Too Far is a co-equal contributor.

Hypothesis Testing

DAX Behavior — Are Drivers Busy on Concurrent OD Trips?

✗ Rejected

HYPOTHESIS

InvalidState failures spike at morning peak (06–08h) because drivers who accepted a JB booking are occupied on a concurrent on-demand (OD) trip when the pre-award allocation check runs — preventing them from being available for the JB pickup.

HOW WE TESTED IT

For every failed allocation flagged as InvalidState, we checked whether the signed driver had an active non-AB booking within a 90-minute window around the allocation timestamp. This directly measures OD-trip overlap as a share of all InvalidState failures — including at peak hours.

EVIDENCE

InvalidState failures with a verifiable active OD trip at allocation time — across all hours

~6%

OD-Busy share at 06–08h peak — no material increase vs off-peak hours

~46%

InvalidState failures caused by driver suspension / system disqualification post-acceptance — the dominant root cause

92.8%

Malaysia drivers with only 1 failed allocation in 25 days — failures are distributed, not concentrated in bad actors

CONCLUSION

The OD-busy overlap is negligible and constant across all time slots. Driver suspension/disqualification after JB acceptance (~46%) is the real driver of InvalidState failures. Targeting concurrent OD trips would not materially reduce FA failures.

Pickup Time of Day — Is the FA Gap Driven by Morning Peak Timing?

✓ Validated

HYPOTHESIS

The 1–2h bucket's higher Failed Allocation rate is concentrated at morning peak hours (06–08h), where supply is tightest and 1–2h bookings made during the pre-dawn window (04–06h) land squarely in the commute surge — unlike 2–6h bookings which are distributed earlier in the day.

HOW WE TESTED IT

We computed the FA rate by hour of pickup across all 6 SEA markets, split by 1–2h and 2–6h lead-time buckets. We also measured each market's share of FAs that fall within the 06–08h window and calculated a concentration ratio (1–2h ÷ 2–6h).

EVIDENCE

2–5×

Higher FA rate for 1–2h vs 2–6h during Morning Peak (06–08h) across MY, TH, PH, ID

20–40%

1–2h FA rate at peak hours in key markets — i.e. 1 in 3 JB bookings in MY & TH fail allocation

2.3×

Malaysia: 44% of 1–2h FAs in 06–08h vs 19% for 2–6h (concentration ratio)

2.8×

Thailand: 42% of 1–2h FAs in 06–08h vs 15% for 2–6h (sharpest peak)

2–3%

Singapore FA rate — flat across all hours, consistent with supply-surplus market structure

Stable mix

Offline/Suspended/OD-Busy sub-type mix unchanged at peak — timing amplifies all sub-types equally, not selectively

SUPPORTING CHART

🔴 Morning Peak FA Rate is 2–5× Higher for 1–2h Bookings Across All Markets

The FA rate for 1–2h bookings surges during Morning Peak (06–08h) and Wee Hours (00–05h) to 20–40% in key markets — vs only 4–10% for 2–6h bookings at the same hours. Gap annotations highlight the 1–2h excess FA rate for periods where the gap exceeds 8pp.

All 6 SEA Markets · May 1–25 2026 · Non-airport Normal AB | Red line = 1–2h FA Rate · Blue dashed line = 2–6h FA Rate · Red shading = gap ≥10pp · Amber shading = gap 5–10pp · Yellow band = morning peak (06–08h)

CONCLUSION

The timing hypothesis is confirmed. 1–2h bookings systematically collide with peak-hour supply constraints because their lead time places pickup squarely in the commute window. The failure sub-type mix is identical at peak and off-peak, confirming this is a structural timing mismatch — not a new failure mode. The primary lever is supply pre-positioning ahead of the 06–08h window for 1–2h bookings.

BONUS FINDING: WHICH REASON DRIVES THE PEAK?

🔴 Invalid State Doubles Its Share at Morning Peak — It's Not Just "All Reasons Rising"

During 04–07h, Invalid State accounts for 53% of 1–2h FAs — vs only 27% off-peak (+26pp shift). Offline/Unavailable is completely stable at ~35% all day. Too Far actually drops at peak (24% → 6%). This means the morning peak amplifies Invalid State specifically — drivers who accepted a JB booking but are then disqualified by the system (suspension, state change) in the window leading up to a pre-dawn pickup. The implication: fixing driver state management before the pre-award check is the highest-leverage intervention for the morning peak.

Malaysia · May 1–25 2026 · 1–2h Non-airport Normal AB | Lines show % share of each reason within that hour's FAs · Grey dashed = overall 1–2h FA rate (right axis) · Yellow = morning peak (04–07h)

FA Reason Share by Pickup Hour — Malaysia

4 Fallback Recovery Deep Dive

Root Cause Found: Driver Post-Cancellation After Fallback Acceptance

The fallback funnel was decomposed into four stages across all markets. In every single market, the fallback recovery gap is explained by one factor: drivers accept the fallback booking then cancel at 17–33% rates for 1–2h bookings vs 1–6% for 2–6h bookings. Supply (unallocation) is not the issue. PAX behaviour is not the issue. The problem is driver commitment after accepting the fallback.

DAX Post-Cancel (MY, 1–2h)

18.8%

vs 1.6% for 2–6h → +17pp gap

DAX Post-Cancel (PH, 1–2h)

32.5%

vs 0.0% for 2–6h → +32.5pp gap

Fallback Alloc Rate (MY, 1–2h)

97.4%

Actually BETTER than 2–6h (94.5%)

PAX Pre-Cancel (MY, 1–2h)

1.9%

Also BETTER than 2–6h (4.0%) — pax more committed

Fallback Funnel Decomposition by Market

Market	FB Recovery Gap	Unalloc Rate (1–2h / 2–6h)	PAX Pre-Cancel (1–2h / 2–6h)	Alloc Rate (1–2h / 2–6h)	DAX Post-Cancel (1–2h / 2–6h) 🔑	PAX Post-Cancel (1–2h / 2–6h)	Verdict
🇲🇾 Malaysia	-15.0pp	1.5% / 2.4% ✅ Not the issue	1.9% / 4.0% ✅ Better in 1–2h	97.4% / 94.5% ✅ Better in 1–2h	18.8% / 1.6% 🔴 +17.2pp gap	3.4% / 2.2% Minor	DAX post-cancel
🇮🇩 Indonesia	-13.7pp	4.3% / 4.7% ✅ Flat	4.7% / 8.3% ✅ Better in 1–2h	93.8% / 90.6% ✅ Better	20.3% / 2.3% 🔴 +18.0pp gap	8.7% / 8.8% Flat	DAX post-cancel
🇹🇭 Thailand	-8.1pp	8.9% / 17.4% ✅ Better in 1–2h	7.1% / 16.6% ✅ Much better in 1–2h	90.4% / 77.5% ✅ Better	26.3% / 5.7% 🔴 +20.6pp gap	3.1% / 1.7% Minor	DAX post-cancel
🇵🇭 Philippines	-28.6pp	1.3% / 2.3% ✅ Minimal	2.1% / 4.6% ✅ Better in 1–2h	97.1% / 94.3% ✅ Better	32.5% / 0.0% 🔴 +32.5pp gap	2.2% / 2.4% Flat	DAX post-cancel
🇸🇬 Singapore	-9.2pp	2.4% / 4.5% ✅ Better in 1–2h	3.4% / 6.4% ✅ Better in 1–2h	94.7% / 90.0% ✅ Better	13.3% / 3.0% 🔴 +10.3pp gap	6.1% / 2.0% +4.1pp, secondary	DAX post-cancel
🇻🇳 Vietnam	-2.9pp	42.8% / 35.6% ⚠️ Supply shortage	3.2% / 6.2% ✅ Better in 1–2h	69.7% / 61.8% Better but still low	13.9% / 1.5% 🔴 +12.4pp gap	6.2% / 3.6% Minor	DAX post-cancel + Supply shortage

Is It a Morning Peak Issue? — No.

The gap is consistent at every hour of the day

DAX post-cancel rate in Malaysia is ~17–21% for 1–2h bookings across all time periods — Wee Hours, Morning Peak, Day Time, Evening Peak, Night Time. For 2–6h it stays at 0–4% across all hours. This is not a morning peak OD competition problem. It is structural to how 1–2h fallback bookings are presented and committed to by drivers.

Malaysia · May 1–25 2026 · Non-airport Advance Bookings | Shaded area = FR gap between buckets | Amber bands = Morning/Evening peak hours

Time of Day	MY 1–2h DAX Post-Cancel	MY 2–6h DAX Post-Cancel	MY Gap	TH 1–2h DAX Post-Cancel	TH 2–6h DAX Post-Cancel	TH Gap
Wee Hours (00–05h)	20.7%	3.0%	+17.7pp	31.2%	11.1%	+20.1pp
Morning Peak (06–08h)	18.3%	3.9%	+14.4pp	27.9%	0.0%	+27.9pp
Day Time (09–16h)	17.3%	0.0%	+17.3pp	20.5%	5.1%	+15.4pp
Evening Peak (17–20h)	18.4%	1.0%	+17.4pp	20.2%	1.5%	+18.7pp
Night Time (21–23h)	17.2%	0.7%	+16.5pp	23.6%	7.5%	+16.1pp

What Does This Mean? — Structural Causes

Fare / Incentive Gap

1–2h fallback fares may be less attractive relative to what drivers can earn from competing bookings at the moment of assignment. 2–6h fallbacks inherit the scheduled booking premium.

ETA Pressure

Driver accepts, then realises they cannot physically reach pickup in 30 min and cancels rather than arriving late. More prevalent in 1–2h because the booking is near-instant.

Commitment Asymmetry

2–6h fallbacks inherit a longer-window AB booking context with stronger acceptance norms. 1–2h fallback feels like a short-notice on-demand booking — drivers treat it with lower commitment.

Revised Action Priorities

Action	Mechanism	Gap Addressable	Markets
Fallback commitment enforcement	Cancellation penalty for driver post-accept cancellation on fallback bookings; increase within first 10 min of acceptance	MY: ~4.1pp (44% of total gap)	MY, ID, TH, PH, SG
Fare uplift on fallback assignment	Increase driver-side fare for 1–2h fallback to make it competitive with OD alternatives at point of assignment	Reduces 17–33pp dax post-cancel	All supply-driven markets
ETA display at fallback accept	Show driver ETA to pickup clearly before acceptance to reduce "accept then cancel" behaviour	Reduces panic cancellations	MY, TH, ID
Vietnam supply intervention	Vietnam has structural supply shortage (42.8% unalloc rate) — incentives for drivers to accept AB fallback in low-supply periods	Addresses unallocation separately from post-cancel	VN only

Remaining Open Question

Open Question	Why It Matters	Status
Final Waterfall Sizing UR → JB-stage losses → fallback-stage losses → entirely unrecovered URs, per market	Quantifies absolute ride count recoverable per lever for business case sizing	Pending

A Appendix — Sample Booking Codes for Engineering Deep Dive

How to use these samples

Each table provides 5 representative bookings per market for the three priority failure buckets (1–2h lead-time, Non-airport Normal AB, May 1–25 2026). Always use the "Last Booking Code" column for system lookups — this is the last booking code in the UR chain (unique_request_id format: UR-{first}-XXXXXXXXXXXXXXXXXX-{last}) and is the canonical reference used in all funnels and certified tables. All timestamps are local time unless noted.

1 JB Post-Allocation PAX No-Show (1–2h bucket)

Driver was allocated at T−30min, arrived on time, but PAX did not board within the no-show window. Timeline: Booking Created → Allocation Broadcast (T−30min) → Driver Allocated → Driver Arrived → PAX No-Show → Driver Cancelled

Market	Last Booking Code ✓	Driver ID	Booking Created	Scheduled Pickup	Lead (min)	Alloc Broadcast (T−30)	Driver Allocated	Driver Arrived (UTC)	Cancel Time	DAX secs post-alloc	Cancel Reason
🇮🇩 INDONESIA
🇮🇩 ID	A-9BOTB3PWWAKHAV	853464	2026-05-17 17:47	19:15:00	87	18:45:00	18:45:00	12:13:42 UTC	19:30:08	2707s (~45min)	PAX_NO_SHOW_V2
🇮🇩 ID	A-99JN63RWWDU7AV	6269480	2026-04-30 22:43	00:00:00	76	23:05:00	23:05:00	16:55:41 UTC	00:18:00	4380s (~73min)	PAX_NO_SHOW_V2
🇮🇩 ID	A-9AEPTQ9GX7GKAV	7219349	2026-05-07 12:30	14:00:00	89	13:15:00	13:15:00	06:48:03 UTC	14:15:09	3609s (~60min)	PAX_NO_SHOW_V2
🇮🇩 ID	A-9AWXWNPWWFU6AV	8224179	2026-05-03 22:24	23:45:00	80	22:50:00	22:54:40	16:16:28 UTC	00:02:25	4065s (~68min)	PAX_NO_SHOW_V2
🇮🇩 ID	A-99K3QCPWWGXVAV	4350837	2026-05-01 01:01	02:30:00	88	01:45:00	01:45:00	19:41:07 UTC	02:56:41	4301s (~72min)	PAX_NO_SHOW_V2
🇲🇾 MALAYSIA
🇲🇾 MY	A-9BXIRI7WXDH7AV	7657974	2026-05-12 02:53	04:15:00	81	03:45:00	03:45:00	19:58:27 UTC	04:30:13	2712s (~45min)	PAX_NO_SHOW_V2
🇲🇾 MY	A-9CF7OMDWX3BFAV	8308383	2026-05-23 04:50	06:15:00	84	05:45:00	05:45:00	22:01:56 UTC	06:31:07	2767s (~46min)	PAX_NO_SHOW_V2
🇲🇾 MY	A-9CJJ7ELGW39CAV	3509161	2026-05-24 06:14	07:30:00	75	07:00:00	07:00:00	23:16:19 UTC	07:45:10	2709s (~45min)	PAX_NO_SHOW_V2
🇲🇾 MY	A-9AOXNMNWXWT9AV	4512562	2026-05-09 19:21	20:45:00	83	20:15:00	20:15:00	12:54:30 UTC	21:13:06	3485s (~58min)	PAX_NO_SHOW_V2
🇲🇾 MY	A-99K4TNOWXCUOAV	2988165	2026-05-01 02:13	03:30:00	76	03:00:00	03:00:00	19:20:22 UTC	03:45:42	2741s (~46min)	PAX_NO_SHOW_V2
🇵🇭 PHILIPPINES
🇵🇭 PH	A-9C96JE2GW6PRAV	5336265	2026-05-21 17:41	19:00:00	78	18:15:00	18:15:10	10:42:21 UTC	19:17:04	3713s (~62min)	PAX_NO_SHOW_V2
🇵🇭 PH	A-9BSI9PJWXALTAV	6633517	2026-05-18 16:05	17:30:00	84	16:45:00	16:45:00	09:28:08 UTC	17:46:05	3665s (~61min)	PAX_NO_SHOW_V2
🇵🇭 PH	A-9BP37COWWO2EAV	1314196	2026-05-17 19:51	21:15:00	83	20:30:00	20:30:00	13:02:28 UTC	21:30:26	3626s (~60min)	PAX_NO_SHOW_V2
🇵🇭 PH	A-9BNB6DGWX2PPAV	8938114	2026-05-17 09:39	11:00:00	80	10:15:00	10:15:00	02:45:41 UTC	11:17:01	3721s (~62min)	PAX_NO_SHOW_V2
🇵🇭 PH	A-9AQJ2FUGWUO9AV	3964130	2026-05-10 10:09	12:00:00	110	11:15:00	11:15:00	03:57:03 UTC	12:17:18	3738s (~62min)	PAX_NO_SHOW_V2
🇸🇬 SINGAPORE
🇸🇬 SG	A-99VUSFFGWUD6AV	932599	2026-05-03 23:01	00:30:00	88	00:00:00	00:00:00	16:21:06 UTC	00:45:29	2728s (~45min)	PAX_NO_SHOW_V2
🇸🇬 SG	A-9B42PJJGXABSAV	5960514	2026-05-12 17:27	18:45:00	77	18:35:00	18:35:00	10:42:04 UTC	18:50:12	912s (~15min)	PAX_NO_SHOW_V2
🇸🇬 SG	A-99TOUKOWWKU2AV	7425730	2026-05-03 10:17	12:00:00	102	11:50:00	11:50:00	03:56:07 UTC	12:06:30	989s (~16min)	PAX_NO_SHOW_V2
🇸🇬 SG	A-9CGGN72WXCP4AV	7406745	2026-05-23 12:18	13:45:00	86	13:35:00	13:35:00	05:41:23 UTC	13:50:09	909s (~15min)	PAX_NO_SHOW_V2
🇸🇬 SG	A-9BDNHSCWW4TTAV	319268	2026-05-15 01:39	03:00:00	80	02:30:00	02:30:00	18:51:53 UTC	03:15:42	2742s (~46min)	PAX_NO_SHOW_V2
🇹🇭 THAILAND
🇹🇭 TH	A-9AUFSM3WWWWGAV	6363942	2026-05-11 07:52	09:15:00	82	08:30:00	08:30:00	02:09:57 UTC	09:31:07	3667s (~61min)	PAX_NO_SHOW_V2
🇹🇭 TH	A-9BRV7FLGX3W9AV	7501075	2026-05-18 11:36	13:00:00	83	12:15:00	12:15:00	05:57:12 UTC	13:16:09	3669s (~61min)	PAX_NO_SHOW_V2
🇹🇭 TH	A-9A8KUDSW3W7XAV	7715500	2026-05-06 00:38	02:30:00	111	01:45:00	01:45:00	19:02:45 UTC	02:45:23	3623s (~60min)	PAX_NO_SHOW_V2
🇹🇭 TH	A-9CDWCE9WX3BFAV	7088886	2026-05-22 14:51	16:15:00	83	15:30:00	15:30:00	08:57:08 UTC	16:43:00	4379s (~73min)	PAX_NO_SHOW_V2
🇹🇭 TH	A-9CIO3PJWXWSPAV	8143349	2026-05-24 00:18	01:45:00	86	01:00:00	01:00:00	18:21:21 UTC	02:00:17	3616s (~60min)	PAX_NO_SHOW_V2
🇻🇳 VIETNAM
🇻🇳 VN	A-9BPALVNW2G84AV	3744088	2026-05-17 20:13	21:30:00	76	21:00:00	21:01:32	14:34:55 UTC	21:51:39	3007s (~50min)	PAX_NO_SHOW_V2
🇻🇳 VN	A-9BXHN5KWX5C6AV	7735104	2026-05-12 01:41	03:00:00	78	02:30:00	02:30:00	19:34:43 UTC	03:15:07	2706s (~45min)	PAX_NO_SHOW_V2
🇻🇳 VN	A-99VOL4FGWKU2AV	9496750	2026-05-03 20:53	22:15:00	81	21:45:00	21:45:00	15:02:30 UTC	22:34:21	2960s (~49min)	PAX_NO_SHOW_V2
🇻🇳 VN	A-9C6XJ62GWSERAV	8221902	2026-05-20 22:17	23:45:00	87	23:15:00	23:15:00	16:15:07 UTC	00:00:06	2706s (~45min)	PAX_NO_SHOW_V2
🇻🇳 VN	A-9AAMOJ5GXDFQAV	7060874	2026-05-06 12:37	14:00:00	82	13:30:00	13:30:00	06:43:56 UTC	14:15:10	2710s (~45min)	PAX_NO_SHOW_V2

⚠️ Driver Arrived is stored in UTC in the raw table. All other timestamps are local time. SG no-show window is noticeably shorter (~15min) vs other markets (~45–73min).

2 Failed Allocation — Driver Invalid State Only (1–2h bucket)

A driver signed up for the JB but was disqualified at the T−30min pre-award check specifically due to Invalid State (PreAwardCheckFailDriverInvalidState). Bookings where Offline/Unavailable or Too Far were also present for any driver are excluded.
Timeline: Booking Created → Driver Signs Up → Allocation Broadcast (T−30min) → Pre-Award Check: Driver Invalid State → Booking fails (no replacement found)
⚠️ This is a very rare failure mode: only MY, TH, VN have booking-level failures with a pure Invalid State signal across Apr–May 2026. ID, PH, SG have zero such cases — when Invalid State fires in those markets, another driver always recovers the booking.

Market	Last Booking Code ✓	JB Booking Code	Driver ID	Booking Created	Scheduled Pickup	Lead (min)	Alloc Broadcast (T−30)	Award Failure Reason	State
🇮🇩 INDONESIA · No pure Invalid State booking failures found in Apr–May 2026 — when Invalid State fires in ID, another driver recovers the booking
🇲🇾 MALAYSIA
🇲🇾 MY	A-99BE7OAWWGVCAV	same	6991893	2026-04-28 23:29	01:30:00	120	01:00:00	Invalid State	CANCELLED_PRE_ALLOC
🇲🇾 MY	A-9CH9A46GWDK5AV	same	8265555	2026-05-23 16:46	18:15:00	88	17:45:00	Invalid State	CANCELLED_PRE_ALLOC
🇵🇭 PHILIPPINES · No pure Invalid State booking failures found in Apr–May 2026 — when Invalid State fires in PH, another driver recovers the booking
🇸🇬 SINGAPORE · No pure Invalid State booking failures found in Apr–May 2026 — when Invalid State fires in SG, another driver recovers the booking
🇹🇭 THAILAND
🇹🇭 TH	A-95QD7SKGWQR7AV	same	7561360	2026-04-01 12:01	13:30:00	88	13:00:00	Invalid State	UNALLOCATED
🇹🇭 TH	A-95QINTQGW34VAV	same	8952027	2026-04-01 13:01	14:30:00	88	14:00:00	Invalid State	CANCELLED_PRE_ALLOC
🇻🇳 VIETNAM
🇻🇳 VN	A-967COO4WWX3SAV	same	1500173	2026-04-04 15:40	17:00:00	79	16:30:00	Invalid State	CANCELLED_PRE_ALLOC
🇻🇳 VN	A-9BNPT3DWWMQ8AV	same	8255052	2026-05-17 11:20	13:15:00	114	12:45:00	Invalid State	UNALLOCATED

✅ All samples verified against ocd_adw.f_transport_advance_jobboard: award_failure_reason = 'PreAwardCheckFailDriverInvalidState' exclusively, with no Unavailable or TooFar events on the same booking. Data scope: Apr 1 – May 28 2026. The generic cancel_reason = 'Cancel before allocation' in the raw booking table is the same for all pre-award failure subtypes; the exact sub-reason is only visible in the JB events table.

3 Fallback DAX Post-Allocation Cancel (1–2h original JB)

The original JB driver cancelled → system created a fallback booking → a new driver was allocated to the fallback → that fallback driver also cancelled, often within seconds. Timeline: JB Created → JB Driver Cancels → Fallback Created → Fallback Driver Allocated → Fallback Driver Cancels

🔴 Pattern Alert: Many fallback cancels happen within 2–15 seconds of allocation using ACCEPTED_IN_ERROR_FREE_CR — a penalty-free cancel reason. This is a strong signal of cancel-penalty evasion: drivers accept then immediately cancel with no financial consequence.

Market	Fallback Code (Last = ✓)	Original JB Code	JB Created	Sched. Pickup	JB Lead (min)	JB Alloc Check	FB Created	FB Driver Allocated	FB Driver Cancelled	DAX secs post-alloc	JB Cancel → FB Cancel Reason
🇮🇩 INDONESIA
🇮🇩 ID	A-9AIRE5AWXFMFAV	A-9AIITQ8GWSM7AV	2026-05-08 10:31	12:30:00	118	11:35:00	12:04:46	12:05:24	12:22:17	1013s (~17min)	ACCEPTED_IN_ERROR → TOO_MANY_PAX
🇮🇩 ID	A-9CO4H7NGWB9NAV	A-9COWIPEWW72DAV	2026-05-25 06:58	08:15:00	76	07:35:03	07:41:23	07:41:42	07:57:21	938s (~16min)	ACCEPTED_IN_ERROR → ACCEPTED_IN_ERROR
🇲🇾 MALAYSIA
🇲🇾 MY	A-9BVVSQPGWBMVAV	A-9BVQSOCGX6XJAV	2026-05-19 11:07	12:30:00	82	12:00:00	12:02:07	12:02:39	12:05:20	161s (~3min)	NOT_ENOUGH_TIME → NOT_ENOUGH_TIME
🇲🇾 MY	A-9B82RR2GW6LRAV	A-9B7TKOFWX72NAV	2026-05-13 15:48	17:15:00	86	16:45:00	16:45:56	16:46:06	16:46:11	4s ⚡	NOT_ENOUGH_TIME → ACCEPTED_IN_ERROR_FREE_CR
🇵🇭 PHILIPPINES
🇵🇭 PH	A-9AQRC65GX65TAV	A-9AQN4OWWXXFKAV	2026-05-10 10:54	12:15:00	80	11:40:06	11:40:18	11:40:51	11:40:57	5s ⚡	ACCEPTED_IN_ERROR_FREE_CR → ACCEPTED_IN_ERROR_FREE_CR
🇵🇭 PH	A-9ARPGO3WWI5FAV	A-9ARLTJAWX6JSAV	2026-05-10 16:30	17:45:00	74	17:00:00	17:09:32	17:09:35	17:09:50	15s ⚡	ACCEPTED_IN_ERROR → ACCEPTED_IN_ERROR_FREE_CR
🇸🇬 SINGAPORE
🇸🇬 SG	A-9BCGUOJGWDU4AV	A-9BC9RJVWX8SDAV	2026-05-14 17:20	18:45:00	84	18:35:00	18:37:57	18:38:37	18:39:02	24s ⚡	PROBLEM_CONTACT_CUSTOMER → PROBLEM_CONTACT_CUSTOMER
🇸🇬 SG	A-9A2OILBGW6L9AV	A-9A2J9BUGWAOUAV	2026-05-04 14:23	15:30:00	66	15:20:00	15:21:08	15:23:49	15:25:07	77s	PROBLEM_CONTACT_CUSTOMER → PROBLEM_CONTACT_CUSTOMER
🇹🇭 THAILAND
🇹🇭 TH	A-9C774T8GWGF7AV	A-9C6VSAEWX25EAV	2026-05-21 03:48	05:15:00	86	04:30:00	05:08:00	05:08:21	05:08:28	6s ⚡	ACCEPTED_IN_ERROR → ACCEPTED_IN_ERROR_FREE_CR
🇹🇭 TH	A-99UV67TGXBB9AV	A-99UR8XXGW6TOAV	2026-05-03 15:32	17:00:00	87	16:15:00	16:25:15	16:25:20	16:25:24	3s ⚡	ACCEPTED_IN_ERROR_FREE_CR → ACCEPTED_IN_ERROR_FREE_CR
🇻🇳 VIETNAM
🇻🇳 VN	A-9CHFGRLWW5JVAV	A-9CH8MU6GXEPVAV	2026-05-23 15:40	17:15:00	94	16:45:28	16:54:43	16:55:02	16:55:04	2s ⚡	ILLEGAL_PICKUP → ACCEPTED_IN_ERROR_FREE_CR
🇻🇳 VN	A-9AWAQHOWX2SAAV	A-9AW4KX9WWMHLAV	2026-05-03 23:04	00:30:00	85	00:01:27	00:11:52	00:16:19	00:16:25	6s ⚡	ACCEPTED_IN_ERROR_FREE_CR → ACCEPTED_IN_ERROR_FREE_CR

⚡ = fallback cancel within 15 seconds of allocation — strong indicator of cancel-penalty evasion via ACCEPTED_IN_ERROR_FREE_CR. All timestamps local time. Scope: 1–2h Non-airport Normal AB, May 1–25 2026.