Case Study: a Zero-App Coordination System for Office Meals

I didn’t fix people. I fixed the system and suddenly no one needed to talk anymore. A real-world coordination experiment.
society
systems
Author

Y Sekhan Althaf

Published

April 27, 2026

Context

We have a small ecosystem:

  • 3 companies (sister companies + holding)
  • ~24 people
  • Shared office space
  • Free lunch provided daily by the owner
  • Hybrid work (WFA/WFO)

The core problem:

No predictability → chef doesn’t know how many people will eat → waste, shortage, confusion → people sometimes blame each other (including the chef)

Important note: There was no formal rule or policy about food waste or attendance. Yet social tension still emerged.

This was not a technical problem. It was a coordination + social problem under uncertainty.

Before vs After

Before

  • Admin walking around office asking
  • “Who eating tomorrow?”
  • “Maybe 5–6? John and Doe?”
  • “Wait let me chat…”
  • …forgot

After

  • No discussion
  • No reminders
  • No coordination needed

The topic of “who is eating tomorrow” effectively disappeared.

Phase 0 — Pre-System (6–7 months ago)

Process:

  • People DM office admin
  • Admin forwards to chef

Problems:

  • No visibility
  • No structure
  • High friction
  • Constant back-and-forth

Phase 1 — WhatsApp Poll (First Intervention)

At first a simple WhatsApp poll.

No explanation. No rollout.

Result:

  • People used it immediately

Cons:

  • Poll gets buried quickly in chat
  • People forget to check

Insight:

Lower friction beats formal process

Phase 2 — HRIS / ERP Form (Formalization)

Moved to ESS (HRIS mobile app).

Pros:

  • Structured data

Cons:

  • People forget
  • Access friction
  • Weekly rotating PIC assigned just to manage this

Insight:

Formal systems increase correctness, but reduce usability

Phase 3 — Partial Automation (Still Heavy)

Automation added:

  • Daily triggers

Improvements:

  • No more PIC

Still broken:

  • People still lazy to open app
  • No shared truth + no follow-up → low compliance
  • Attendance still manual

Phase 4 — WhatsApp + n8n + ERP (Breakthrough)

Timeline:

  • Summary (group visibility) → Feb 3
  • WhatsApp broadcast → Apr 2

Before April:

  • Hybrid (ESS + WhatsApp)

After:

  • Fully WhatsApp-based interaction

Result:

Instant adoption without training or socialization

Phase 5 — Physical Verification (TV + Code)

Added:

  • TV display
  • Rotating 3-digit code

Result:

  • ~60–70% compliance
  • Emergent ritual: people start entering the code together while waiting for food

Design decision:

  • No shaming
  • Only neutral display

Reason:

  • Social psychology → avoid resistance

Phase 6 — System Collapse (Unintentional Experiment)

After long holiday:

  • System off
  • TV gone (It get repurposed by management)

Result:

  • Chaos returns
  • People confused
  • Admin complains

CEO asked to turn system back on.

Insight:

The system was doing invisible coordination work

Phase 7 — Reconstruction (Independent System)

Removed:

  • TV dependency
  • HRIS dependency

Final:

  • Pure WhatsApp + backend

Final Architecture

3 phases (grouped workflows):

Pre-Event

  • 2 Days Before — Broadcast Registration
  • 1 Day Before — Summary
  • Registration Listener

During Event

  • On The Day — Code Sender
  • Check-in Listener (validation + walk-in)

Post Event

  • Cleanup (6 PM finalization)

Key Properties

1. Independence

Each workflow:

  • self-contained
  • can fail independently

2. Graceful Degradation

  • Registration fails → attendance still works
  • Attendance fails → registration still useful

3. Emergent Behavior

Example:

  • Cleanup at 18:00
  • Late check-in still allowed until 23:59

No explicit design for this.

Key Design Principles

1. Stateless Core

  • Validate at the edge, duplicated form don’t affect system

2. Loose Coupling

  • No tight dependencies

3. Late Validation

  • Accept → validate later

4. Human-Compatible Interface

  • WhatsApp only

5. Tiered Communication (Anti-Spam)

  • Minor action → emoji reaction
  • Important action → message
  • Repeat → reaction only

6. Social Layer > Enforcement

  • No punishment
  • No rules

Behavioral & Social Layer

Relevant lenses (to be cited later):

  • Nudge Theory → reduce friction, guide behavior subtly
  • Social Proof / Collectivism → seeing others influences decision
  • Tribal Coordination → small group self-regulates
  • Cognitive Load Reduction → remove need to think/remember

Important observation:

No rule was introduced. Behavior changed anyway.

Emergent Behaviors

1. Self-Regulation (TV era)

  • People reminded each other

2. Invisible Operation (current)

  • No discussion
  • Everyone still follows system

3. Social Alignment

  • Summary (posted in shared WhatsApp group at 6 PM the day before) influences attendance
  • People sometimes change their decision after seeing friends attending, and register for the next day

4. Soft Deadline

  • 18:00 → analytical truth
  • 23:59 → physical truth

Outcomes

Observable

  • Registration → ~100% adoption
  • Code → improved after iteration

Not Quantified (but real)

  • Admin saves ~5–30 minutes daily
  • No more walking around asking people
  • No more group chat noise

System Quality

  • Low-code (n8n + ERP + WhatsApp)
  • Extremely maintainable and malleable

What I Didn’t Do

  • No app
  • No onboarding
  • No presentation
  • No mandate
  • No change management
  • No strict security (low abuse surface)

Bottom-up adoption, not top-down enforcement

What This System Is

A lightweight coordination infrastructure embedded in existing tools

Simple interface. Complex backend.

Hypothesis

  1. Coordination shifts from conversation to observation when shared truth is present.

  2. Communication is often a symptom of missing structure.

Key Insight

The system didn’t make people more disciplined. It removed friction and made coordination obvious.

Final Reflection

When the system stopped → people noticed. When it runs → no one talks about it.

That is what good infrastructure feels like.