Research
Our models come from labor economics and the economics of organizations. This page explains the main ideas we use and why they matter for workforce decisions.
A worker's productivity in a specific role is not known in advance. It is revealed gradually as the worker gains tenure and both sides learn whether the fit is good. This view of work, developed in the job matching literature, explains why turnover happens, why experience raises wages, and why moving a person to a new role carries short term risk while the new match is still being learned.
We use this to model staffing and backfill. When people leave or deploy, the twin ranks candidate moves by expected match quality and the time it takes to get up to speed, rather than filling seats by seniority alone. The goal is to protect output during transitions.
Organizational structure is, in economic terms, an assignment problem. Who works on what, who manages whom, and which people form a team are all choices that determine output. When skills are complements, the best assignment is assortative: high skill paired with high skill, demanding roles paired with capable people. When tasks differ in how substitutable skills are, the pattern is more subtle, and the same role can hold very different types of workers.
We use sorting and team assignment models to evaluate structure. The twin measures how well a given hierarchy lets an organization place the right skills in the right roles, and where mismatch is quietly costing output.
Hierarchy exists partly to manage knowledge and attention. Managers handle the problems that frontline workers cannot, and layers let an organization apply scarce expertise across many people. But every layer adds communication cost, and every manager has a limited span of control. The right number of layers and the right span differ by industry: parallel, low coordination work favors flat and wide structures, while complex problem solving rewards deeper delegation.
The simulator encodes these field specific relationships. It computes the span of control implied by your headcount and layers, compares it to the field's effective range, and searches for the structure that balances coordination benefit against communication cost.
Skills are not fixed. They are produced through education, training, and experience, and they depreciate when they go unused. A labor market view of skills treats the workforce as a stock of capabilities that an organization invests in and deploys. This matters for automation: the value of a new tool depends on whether people have, or can build, the complementary skills to use it well.
We model human capital so that restructuring and automation plans account for the learning that has to happen, not only the headcount that changes.
Because real personnel records are sensitive, we use these models as data generators. Calibrated to an organization's patterns, they produce synthetic workforce data that reproduces realistic distributions of tenure, mismatch, and team output, without exposing individuals. Synthetic data lets leaders test scenarios safely and lets us validate the twin against held out outcomes.
Full citations are available in our technical materials on request.