I turn messy operations into systems that scale.

Context & problem

Pokus ran across a long, fragmented chain — customer → CX → RM → vendor → picker/runner → rider → customer. Every handoff was a place for delay and information loss, and nothing was owned end to end.

Decisions & actions

• —Mapped the full chain and assigned a single owner per stage
• —Designed request intake, customer communication, cart drop-off and escalation workflows
• —Built RM workflows: task allocation, vendor coordination, customer updates, escalation handling
• —Defined procurement flow: sourcing, vendor mapping, availability checks, procurement SLA
• —Defined fleet flow: rider allocation, pickup SLA, delivery SLA, shift planning
• —Gave management KPI dashboards, SLA monitoring and escalation tracking on top

Outcome

• —Execution moved from people-dependent to system-dependent
• —Clear ownership and defined SLAs at each stage
• —Escalations triggered by rules instead of by noise
• —Founders stopped being the escalation path of last resort

Context & problem

Allocation was a manual judgement call every time: which vendor, which zone, which rider. It scaled with headcount and broke during peak windows when the same high-throughput vendors got overloaded.

Decisions & actions

• —Built a daily-updated capacity database across 10+ vendors
• —Turned allocation into a decision tree using capacity, location, price and turnaround
• —Introduced zone-based hotspot dispatch instead of first-available assignment
• —Integrated Pro Routing and Porter as third-party logistics (3PL) overflow partners
• —Used AI-assisted workflows for research, routing checks and repetitive allocation decisions

Outcome

• —55% reduction in pickup-to-delivery time
• —Peak-window vendor overloading eliminated
• —Sharply less manual decision-making per order
• —Allocation quality no longer depended on who was on shift

Context & problem

The laundry pilot launched with no riders and no fleet function. In the first phase I did deliveries myself to learn where the real constraints were.

Decisions & actions

• —Onboarded gig partners and scaled the fleet from 0 to 15 riders
• —Designed onboarding, verification and training flow
• —Built shift design and incentive structures around demand windows
• —Set pickup and delivery SLAs with monitoring and breach escalation
• —Built rider-level performance tracking and review cadence
• —Added 3PL partners as overflow capacity for spikes

Outcome

• —0→15 rider fleet running on a repeatable onboarding system
• —Predictable coverage across peak windows
• —Performance visible per rider instead of per anecdote

Context & problem

Customers could request products that no conventional catalogue guaranteed. Fulfilling arbitrary requests fast while holding service levels meant availability, price, location and speed all had to be weighed at once.

Decisions & actions

• —Structured the flow as request → assignment → procurement → verification → delivery → feedback
• —Built structured vendor mapping across brands including MAC, Estée Lauder, La Roche-Posay, The Ordinary, L'Oréal, Lakmé, Kiehl's and Bioderma
• —Defined procurement SLA and availability confirmation steps before customer commitment
• —Added verification and invoice/payment workflows to close the loop

Outcome

• —Arbitrary requests fulfilled through a defined path, not improvisation
• —Clear ownership and escalation triggers at each stage
• —More predictable fulfilment and better customer communication

Context & problem

Completing the order was table stakes. Nothing in the experience made a customer choose us again, and nothing gave them a reason to increase basket size.

Decisions & actions

• —Designed post-purchase delight: personalised communication, handwritten notes, packaging, gifting
• —Built a tiered reward framework using physical utility rather than pure discounting
• —Wired loyalty and post-purchase engagement into the CX workflow

Outcome

• —Rewards ladder tied directly to order value thresholds
• —Discounting reserved for the highest tiers, protecting unit economics
• —A transaction turned into a relationship worth repeating

Context & problem

Operational decisions were being made on recollection. There was no shared definition of a good day, and no way to see where time was actually going.

Decisions & actions

• —Defined the core metric set: first response, assignment, execution, delivery and total task time
• —Instrumented turnaround time (TAT), pickup service-level agreement (SLA), delivery SLA, vendor delay and breach rate
• —Added relationship manager (RM), customer experience (CX) and rider performance views
• —Built a daily → weekly → monthly reporting cadence with SLA alerts

Outcome

• —Bottlenecks identified by data rather than debate
• —Team-level and vendor-level reporting on one cadence
• —Interventions could be measured after the fact