Product Adoption and Diffusion Models

Summary

Bonabeau (2002) presents a formal product adoption model where a product’s value depends on the fraction of adopters, creating positive network externalities. The model demonstrates that aggregate system dynamics and ABM produce equivalent results for well-mixed populations but diverge when network structure introduces local information asymmetries. The adoption process follows an S-curve driven by a threshold effect at ~40% adoption.

Overview

Product adoption and diffusion is a classic application of ABM. The fundamental question is: how does a new product spread through a population? Traditional models (Bass diffusion model, system dynamics) treat the population as well-mixed. ABM reveals that network structure can qualitatively change diffusion patterns.

Main Content

The Value Function

Definition: Product Value Function (Bonabeau 2002)

$$V(N)=V(\rho )=\frac{(1+{\theta }^d){\rho }^d}{{\rho }^d+{\theta }^d}$$

where:

• $\rho =N/N_T$ is the fraction of adopters in a population of $N_T$ potential adopters
• $\theta =0.4$ is a characteristic value (adoption threshold)
• $d=4$ controls the steepness of the adoption curve

Properties:

• $V(0)=0$: no value when no one has adopted
• $V(N_T)=1$: maximum value when everyone has adopted
• $\theta$ acts as a threshold: the value curve takes off when user base approaches 40% of the population

Adoption Dynamics

The adoption rate follows the master equation:

$${\partial }_t\rho =V(\rho )(1-\rho )$$

This produces the characteristic S-curve:

1. Slow start: Few adopters → low value → low adoption rate
2. Tipping point: When $\rho$ approaches $\theta$ (~40%), the value curve takes off
3. Rapid growth: High value + many non-adopters → fast adoption
4. Saturation: Few remaining non-adopters → slow completion

Agent-Level Transition Rule

Each individual agent has a transition probability:

$$P({\text{adopt}}_k)=V({\hat{\rho }}_k)\text{ per time unit}$$

where ${\hat{\rho }}_k=n_k/n$ is person $k$‘s local estimate of the adoption fraction based on their $n$ neighbors.

Network Effects on Diffusion

The critical insight is that ${\hat{\rho }}_k\ne \rho$ when the network is not well-mixed:

• Random network (30 random neighbors per agent): ${\hat{\rho }}_k\approx \rho$ → Smooth S-curve, consistent with differential equation
• Clustered network (two clusters): ${\hat{\rho }}_k$ varies dramatically between clusters → Two-wave adoption pattern

See Network Topology Effects on Diffusion for detailed analysis.

Connection to Rogers’ Diffusion Theory

The ABM approach connects to Rogers’ (1983) diffusion of innovations theory:

• Innovators adopt early with low ${\hat{\rho }}_k$ (high innovativeness threshold)
• Early adopters follow once some neighbors have adopted
• Early majority wait for substantial neighborhood adoption
• Late majority and laggards adopt only under strong social pressure

In CUBES, these adopter categories are formalized through the five-group classification of consumers based on adoption speed (Ben Said et al. 2002, footnote).

Examples

Example: Adoption Dynamics with $\theta =0.4$, $d=4$ (Bonabeau 2002, Fig. 3-4)

Setup: 100 agents, 5% initial adopters, time unit = 10 days.

Random network (30 neighbors): Adoption follows a smooth S-curve reaching saturation around day 100. Nearly identical to the differential equation solution.

Clustered network (2 clusters): Adoption proceeds in two distinct waves — the cluster containing the initial adopters saturates first (by ~day 50), then adoption jumps to the second cluster and proceeds to saturation (~day 120). The transition between waves produces a visible plateau in the adoption curve.

Connections

• This model formally demonstrates the comparison in ABM vs Equation-Based Modeling
• Network topology effects are detailed in Network Topology Effects on Diffusion
• The adoption model connects to the broader emergence concept
• In marketing contexts, diffusion interacts with Word of Mouth Mechanisms and opinion leader targeting

See Also

• ABM vs Equation-Based Modeling — formal comparison using this model
• Network Topology Effects on Diffusion — how topology shapes adoption
• Word of Mouth Mechanisms — the mechanism driving adoption