Description
Tracks, for each person, the share of their network that has already adopted, and takes its value at the moment they adopt as their personal threshold.
A threshold is the number or share of others who must adopt before an individual does; critical mass is its system-level counterpart. Valente's procedure needs network ties and adoption dates for every member of a bounded system. For each person and period, exposure is the percentage of network partners who have adopted; exposure at the person's own adoption is their threshold. People are classified by innovativeness relative to the whole system and relative to their own network, which separates low-threshold early adopters from people who adopt late simply because their contacts did. Threshold lag (exposed but not yet adopting) is distinguished from lack of exposure. The 'turbocharger' effect is the extra variance in adoption timing explained by adding network variables to individual ones. Few data sets have both full networks and dates, and innovations whose use is invisible to others weaken threshold effects.
Attributes
- Discipline
- Statistics & Modelling
- Loss type
- Unmigrated
- Stage
- Post-launch, Strategy
- Data
- Quantitative
- Effort
- Weeks
Steps
- Define a bounded system (a team, organisation or workspace) and collect its ties and each member's adoption date.
- For each person and period, compute exposure as the percentage of their network partners who have adopted.
- Record exposure at the period of adoption as the person's threshold.
- Classify each person's innovativeness relative to the system and relative to their own network.
- Identify people with threshold lag (highly exposed non-adopters) and unexposed non-adopters as separate groups.
- Test the distribution of thresholds and estimate how much variance in adoption timing network variables add beyond individual ones.
Sources
- Diffusion of Innovations (Rogers, 2003)
Also known as
- Network exposure and individual threshold analysis
- Adoption threshold analysis
- Valente threshold model
Categories
- Analytics & Behaviour
- Statistics & Modelling