Skip to content

The Literature

The science underneath the product.

A literature audit in three layers: the canonical work that shaped the science of team performance, what Performix reflects of it today, and what the research says we don't yet measure. Audience: anyone evaluating whether we take the literature seriously.

A note on honesty: this page names the works and their established contributions. It does not assert effect sizes or pooled estimates — those, with their source grades, live in the Performix's evidence registry, the science-of-record behind the diagnostic. Numbers are cited there or omitted here, never invented. And where a synthesized prior exists, the diagnostic now shows it behind the number — the binding-constraint result carries its evidence prior (r, interval, study count, grade) with a link back to its provenance, and says so plainly when no prior is available rather than inventing one.

Different from /learn. That surface teaches the product; this page audits the science.

R1 · Reading list

The canonical work on team performance.

Curated, not exhaustive — the works that actually shaped how the field understands team performance, organized by the conjunction conditions (Capability, Alignment, Motivation, Support) they speak to, plus the measurement methods that make the diagnostic possible.

See the proof

The diagnostic logic

  • Goldratt, The Goal (1984) — Theory of Constraintsbinding constraint

    A system's throughput is set by its single binding constraint; relieve that one and the next becomes binding. The logic Performix imports wholesale: find the one condition starving the team, act there, remeasure.

  • J. R. Hackman, Leading Teams (2002); Hackman & Wagemanenabling conditions

    Team performance is driven by a small set of enabling conditions, not by the traits of individual members. The unit of analysis is the team-as-system — which is why Performix scores teams, not people.

  • Salas, Sims & Burke (2005), “Is there a Big Five in teamwork?”coordination

    Teamwork decomposes into a small set of coordinating mechanisms. Performance loss is often a coordination property, not a competence deficit.

Capability (C)

  • Hackman & Oldham (1976); Work Redesign (1980) — Job Characteristics Modelwork design

    Capability becomes performance only when the work is designed to use it — skill variety, task identity and significance, autonomy, feedback. The C dimension scores capability against the work, not in the abstract.

  • Schmidt & Hunter (1998) — validity of selection methodscapability prediction

    Structured measures predict job performance far better than intuition — but they predict the person, not the system around them. Selection is necessary, not sufficient; the other three conditions decide whether capability converts.

Alignment (A)

  • Locke & Latham (1990; 2002) — Goal-Setting Theorygoal clarity

    Specific, difficult, accepted goals outperform vague ones. Alignment is whether a team shares a clear goal it has actually accepted — not whether a goal exists somewhere in a deck.

Motivation (M)

  • Deci & Ryan (1985; 2000) — Self-Determination Theoryintrinsic motivation

    Durable motivation is intrinsic, resting on autonomy, competence, and relatedness; controlling environments crowd it out. Motivation is a condition of the environment as much as a property of the person.

  • Colquitt (2001) — organizational justice dimensionsperceived fairness

    Perceived fairness of how decisions and rewards are made drives discretionary effort. A motivation problem is frequently a fairness problem wearing a different label.

Support (S)

  • Edmondson (1999) — Psychological Safety and Learning Behavior in Work Teamspsychological safety

    Teams learn and perform when members can take interpersonal risks without fear. This is precisely the condition protected feedback engineers for — the diagnostic is impossible until people can safely say what they see.

  • Eisenberger et al. (1986) — Perceived Organizational SupportPOS

    The belief that the organization values one's contribution and has one's back predicts engagement and retention. Support is whether the environment lets the team do the work it is capable of.

Measurement & method (the analytical spine)

  • Lord & Novick (1968); Embretson & Reise (2000) — Item Response Theoryadaptive measurement

    Items carry different information at different trait levels, so you can ask fewer, better questions and adapt to what you already know. This is the engine behind adaptive narrowing.

  • Louviere & Woodworth (1991); Marley & Louviere (2005) — Best-Worst Scaling / MaxDiffforced choice

    Choosing the best and worst among options is harder to game and more discriminating than agree/disagree rating scales — which is why the survey feels like real trade-offs, not a Likert grid.

  • Wilson (1927) — confidence interval for a proportionconfidence range

    With small samples, a score is a range, not a point. The right interval keeps the diagnostic honest about what it does and does not know.

  • Cronbach (1951) — coefficient alphareliability

    A measure is only as trustworthy as its internal consistency. Reliability tracking is what tells the system when an item has stopped earning its place.

  • Raiffa & Schlaifer (1961) — applied statistical decision theoryvalue of information

    Whether to gather more data is itself a decision with an expected value (EVPI/EVSI). Ask the next question only when the information it buys is worth its cost — the discipline behind VOI gating.

The engagement critique

  • Macey & Schneider (2008) — The Meaning of Employee Engagementengagement (and its ambiguity)

    “Engagement” conflates trait, state, and behavior. Measuring a feeling is not measuring the conditions that produce performance — the distinction the whole product rests on.

R2 · Reflected in the product

What we measure, and which research it traces to.

Each row maps a research result to the Performix surface that operationalizes it — and states honestly whether the reflection is complete, partial, or still planned. Partial means we use the construct without the full battery, and we say so.

See the proof
Research
Performix surface
State
Theory of Constraints (Goldratt)
The binding-constraint rule — lowest CAMS dimension is what to act on
Core
Job Characteristics Model (Hackman & Oldham)
Capability-dimension items + job-spec-authoring
Partial
Goal-Setting Theory (Locke & Latham)
Alignment-dimension items
Reflected
Self-Determination Theory + justice (Deci & Ryan; Colquitt)
Motivation-dimension items
Partial
Psychological safety + POS (Edmondson; Eisenberger)
Support-dimension items + the protected-feedback rationale
Reflected
Item Response Theory (Lord; Embretson & Reise)
Adaptive narrowing engine
Reflected
Best-Worst Scaling / MaxDiff (Louviere)
Forced-choice item format
Reflected
Wilson interval (1927)
Confidence ranges on dimension scores
Reflected
Decision theory / VOI (Raiffa & Schlaifer)
Value-of-information gating in adaptive narrowing
Reflected
Synthesized Bayesian priors (meta-analytic)
Evidence prior + provenance behind the binding-constraint result (/diagnose)
Partial
Coefficient alpha (Cronbach)
Reliability tracking (Adaptive Measurement)
Planned

The discipline behind this table is the same one behind the whole site: a claim of reflection is backed by a surface you can see, or it is marked Partial or Planned. Nothing here claims a battery we have not built.

R3 · Roadmap from the research

Findings not yet reflected — and the plan for each.

The load-bearing honesty move. Where the literature says something we don't yet measure, we name it here, with the planned addition. Readers see both the gap and the path; a marked gap is a credibility asset, not a weakness.

See the proof
  • Longitudinal team dynamics — how conditions evolve over time, not just a cross-section.

    planRemeasurement is the V0 falsifier today; formal longitudinal modeling lands with MVP 4 (Adaptive Measurement).

  • Relational structure — who depends on whom inside the team and across functions.

    planNot modeled yet. A future org-graph spoke would let the diagnostic reason over dependencies, not just dimensions.

  • Measurement invariance — whether items mean the same thing across cultures and segments.

    planRequired before cross-segment benchmarking; sequenced ahead of the V1 benchmark layer rather than asserted prematurely.

  • Causal inference — did relieving the constraint cause the lift, or did something else move?

    planRemeasurement against the same dimension is the honest V0 test; formal causal design is on the research roadmap.

  • Capability scored against the specific basket of work, not generic competence.

    planClosed by the Capability Modeling Engine (MVP 3) + job-spec-authoring; partially built today.

  • Graded effect-size provenance — the precise pooled effects behind each construct.

    planNow being closed: the diagnostic reads the synthesized evidence prior from Performix's research registry and renders it behind the binding-constraint result — r, interval, study count, grade, and a provenance link — with “no prior available” as a first-class state. Shipped as the affordance on demo data today; flips to the live registry once connected. The graded, annotated bibliography is generated from the same source.

Next

The science says what to measure. The vision is how the platform gets there.