Vivameda

Research Portfolio 2026

Five Studies, One Panel

Predicting growth. Identifying structural vulnerability. Measuring transformation. Reconstructing long term organisational readiness. Detecting capability decay before failure. Five question types asked of the same longitudinal workforce panel.

Isometric visualisation of five layered research panels connected by data nodes

The panel behind the studies

4,000,000+

Companies observed

48,000,000

Observed company years, 1950 to 2026

1.88 bn

Skill rows on one taxonomy

100+

Countries covered

Why one panel answers five questions

Five applications of one dataset

An annual report gives a headcount. It does not say how many of those people were technical, how many were on the shop floor, how many were in buying. No reporting standard requires that breakdown, and for privately held companies no figure is published at all. Workforce composition is one of the few variables observable at company level, comparable across peers, available for firms that have since disappeared, and deep enough in time to test whether a pattern actually preceded an outcome.

Each study takes a state of the workforce at a point in the past, fixes the classification before looking at what followed, and then checks the outcome. Growth data ranks companies. Workforce structure explains the ranking, and it is visible earlier.

The portfolio

What each study detects

The studies are independent and can be read in any sequence. The order below runs from the earliest signal in a company life to the longest horizon.

Study I

The Unicorn Signature

Pre hypergrowth patterns

The top 1 percent of scored companies enter hypergrowth at 27 times the base rate. Of six unicorns held entirely out of training, 100 percent scored in the top 30 percent in the year before they scaled.

Show test design and decisions
Test design
12,285 held out companies, 90,233 observations, time safe features, company level split, AUC 0.889.
Relevant decisions
Sourcing, investment screening, account prioritisation

Study II

The Defensible Shelf

Structural business vulnerability

Of 225 large retailers, 72 are gone. Those with an online substitutable assortment failed at 42.1 percent against 7.6 percent for the rest, a factor of 5.5.

Show test design and decisions
Test design
Signals drawn only from 2012 to 2016, outcomes read from the public record for 2017 to 2026. Seven independent workforce signals all pointed the same way.
Relevant decisions
Due diligence, credit risk, portfolio review

Study III

The Closed Circuit

Capability and talent network lock in

Across seventy analysed personnel exchange relationships spanning six companies, not one software house appears. Eight of nine automotive units moved by less than 0.10 index points in the decade they spent describing themselves as software businesses.

Show test design and decisions
Test design
A capability scale validated against nine reference companies whose expected values were written down before the query was run.
Relevant decisions
Transformation, recruiting strategy, competitive analysis

Study IV

The 1998 Head Start

Long term organisational readiness

Across 41 multinational corporations, those already distributed in 1998 reduced workforce concentration almost ten times as much as concentrated ones over the following twenty years, at an identical rate of entity growth.

Show test design and decisions
Test design
Three classes formed from a single base year, tested with a permutation test across 100,000 draws.
Relevant decisions
Resilience, internationalisation, operating model assessment

Study V

The Portable Asset

Capability decay before organisational failure

Four AI companies that later contracted were already thinning their engineering share between 2017 and 2020. The two that persisted were thickening theirs. Six of six fell in the expected direction. Primer lost 14.5 points of engineering share before the contraction that followed.

Show test design and decisions
Test design
Cohort and classification fixed in writing before measurement, identical 2017 to 2020 window for every company, one role taxonomy, role coverage above 78 percent. Composition and subsequent presence are reported on separate layers and never netted.
Relevant decisions
Acquisition diligence, portfolio review

Standards

How these papers are written

Each study states its headline finding first, shows the evidence, then consolidates its limitations into a single section before the appendix. Those sections name the constraint that most restricts the headline claim, the alternative explanations that were tested and survived, and the approaches that failed the test and were discarded before publication.

Every classification rule and query behind these studies is reproducible and is disclosed on request. Where a figure comes from a second source with a different collection methodology, the paper says so and does not net it against panel figures.

Apply the research to your own universe

The same instrument can be applied to a specific company list, market, portfolio or acquisition universe. Engagements start with a short data discussion to identify the right starting point.

Discuss your use case

SYSTEM REF · VMD.21921.120826.224