Plub held a 30-minute grocery SLA in LATAM with Shipsy's AI clubbing engine

Plub held a 30-minute grocery SLA in LATAM with Shipsy’s AI clubbing engine

Plub, a quick-commerce grocery operator in Latin America, now hits 90%+ adherence to its marketed 30-minute SLA, auto-allocates 98.5%+ of orders, saved $168K per year in resource costs, and lifted rider utilization by 10–25% — all on Shipsy’s AI-native order-clubbing and allocation engine. In a market where a missed SLA means a lost customer, Plub’s dispatch layer now runs itself.

Customer: Plub. Industry: Quick-commerce grocery. Region: Latin America (~$430M USD online grocery TAM). Shipsy modules deployed: AI Order-Clubbing Engine, Allocation Engine with time-constraint logic, Dynamic ETA, Live Tracking, Operational Alerts. Headline metric: 90%+ adherence to 30-min SLA, 98.5%+ auto-allocated orders, $168K/year saved, +10–25% rider utilization.

The challenge: a 30-minute promise and a mixed fleet

Plub operates in a ~$430M USD LATAM online grocery market, serving 150,000+ users through 400+ drivers across two hubs. The brand’s positioning is built on a 30-minute SLA — the threshold at which quick-commerce grocery economics work.

Holding that SLA turned out to be the hardest operational problem on the roster. Dispatch and fulfilment windows were razor-thin; every incremental minute of planning latency ate into delivery margin. Order clubbing — grouping multiple orders into a single rider trip — was inefficient: either batches breached the SLA, or riders waited on a delayed second order, killing productivity.

The fleet model was mixed: dedicated riders on fixed hours plus floating riders activated during peak windows, spread across two hubs. Every clubbing decision had to account for rider type, hub proximity, order readiness, and SLA remaining — all in real time. Manual dispatchers couldn’t keep up, and earlier rules-based engines missed the nuance.

When clubbed orders stalled, rider productivity collapsed. When they didn’t club, per-order costs spiked. Plub needed an engine that could thread both needles.

The solution: AI clubbing + time-aware allocation

Shipsy’s AI-native order-clubbing engine became the backbone of Plub’s dispatch stack. Unlike static batching heuristics, Shipsy’s engine learns from historical clubbing outcomes — which order pairs actually delivered on time, which hub-rider combinations minimize wait, which item categories tolerate a brief pickup delay — and uses that pattern library to propose clubs in real time.

On top of the clubbing engine sits an allocation engine with time-constraint logic. Every assignment decision explicitly evaluates SLA remaining, rider type (dedicated vs floating), hub proximity, and current rider load. Dedicated riders on fixed hours get first pick of low-variance trips; floating riders get surge-appropriate assignments during peaks. The system doesn’t just allocate — it allocates with the clock in the decision.

Dynamic ETA and live tracking close the customer-experience loop: customers see an ETA that actually reflects dispatch-stage reality, not an optimistic static promise. When a club is at risk, the system alerts ops proactively.

Operational visibility and alerts surface SLA-risk signals to ops managers before breaches happen — not after. That shift from reactive to proactive is the difference between saving a delivery and losing a customer.

The result: a dispatch layer where 98.5%+ of orders get auto-allocated without human touch. Ops teams are reserved for genuine edge cases, not routine routing decisions.

The outcome: margin reclaimed, SLAs held

The headline: 90%+ adherence to Plub’s marketed 30-minute SLA. For a quick-commerce grocery brand, that’s the difference between retaining customers and churning them — the 30-minute promise is the reason the customer chose Plub over a traditional grocery channel in the first place.

98.5%+ orders auto-allocated. The dispatch layer runs effectively on autopilot, with human ops teams handling the 1.5% that need judgment calls — not the other 98.5%.

Rider utilization up 10–25%. Better clubbing means riders carry more orders per active hour, and tighter allocation means less idle time between trips. For a mixed-fleet operator, that directly improves unit economics on both dedicated and floating labor pools.

$168K/year in resource cost savings. Fewer dispatcher hours, less manual intervention, fewer SLA breach refunds. For a 400-rider, two-hub operator, that savings figure is a meaningful slice of operating margin.

Together, these outcomes mean Plub can hold a 30-minute SLA profitably — not as a loss-leading promise, but as the anchor of a margin-positive unit-economics model.

What’s next

Plub continues expanding Shipsy’s footprint — more hubs, more rider cohorts, tighter integration between demand signals and clubbing decisions. Near-term priorities include deeper predictive-demand inputs to the allocation engine and expanded incentive automation for floating riders during peak windows.