Modeling Neighborhood-Scale Retail Dynamics

This project simulates how anchor strength and small-business vitality influence pedestrian movement and lingering behavior in Central Square, Cambridge. The workflow integrates network science and agent-based modeling to estimate how people might move and pause in response to changes in neighborhood retail ecosystems.

Methodology

1. Data Pipeline

  • Central Square BID business dataset

  • OpenStreetMap street network

  • Scenario modifications

2. Scenario Construction

  • Baseline Case

    • Represent case as of December 2025

    • Designate anchors ( range from 0.4 to 1.0)

  • Scenario 1 — Public Space Intervention

    • Design intention: enhance Jill Brown-Rhone Park with amenities that meaningfully draw people

      • seating, greenery, programming, lighting, etc.

    • Model implementation: increased anchor weight for the park

  • Scenario 2 — Small Business Boost

    • Behavioral assumption: small, independent businesses become more attractive relative to chains

      • reflecting trends such as “shop local,” curated retail experiences, and human-scale service

    • Model implementation: increased anchor weight for non-chain stores

System Design

Agent-Based Modeling Rules

Each agent represents a simulated pedestrian.

Initialization

  • 300 agents seeded at random network nodes

  • 120 movement steps each

Movement Rule

At each step, an agent chooses among neighbors using:

Where:

  • A(v) = node attractiveness from retail + anchor effects

  • d(u,v) = edge distance

  • β = sensitivity to attractiveness

  • c = distance penalty

P(move to v) ∝ exp( β · A(v) − c · d(u,v) )

Lingering Rule

At each node:

dwell time+ = 1 + linger_scale × A(n)

Output Metrics

  1. Movement Flow: how often edges are traversed

  2. Lingering Heat: accumulated dwell time at nodes

  3. Anchor Influence Map: visualized anchor weights

Maps

  1. Significant increase in lingering near the park:
    Agents now treat it as a destination rather than a pass-through.

  2. Adjacent edges show higher movement flow, as agents divert toward the upgraded node.

  3. The park becomes a third anchor balancing MIT/Harvard influence.

  4. Slight reduction of lingering at nearby mid-tier anchors due to competition for attention.

  1. More distributed lingering—a network of micro-anchors emerges.

  2. Flow becomes less centralized, spreading further into smaller side streets.

  3. Stronger reinforcement of “local loop” behavior:
    Agents stay within a walkable cluster of interesting small businesses.

  4. Chain stores do not receive boosts → local retail becomes more competitive.

  1. Strongest anchors: MIT and Harvard → pull agents north and south.

  2. Moderate anchors (restaurants, cafés) create a central ribbon of activity along Mass Ave.

  3. Lingering clusters appear around food + cultural venues.

Previous

Research: Retail Vitality
Next

Prediction: Retail Vitality