Method · Technical note

How we decide where to intervene and how the result is measured.

This note explains the quantitative reasoning behind our six working phases. It is written for technical teams and operations or finance leadership who want to review the criteria before engaging.

You do not need to understand this document to work with us: the commercial proposal is always written in business language, with deliverables, acceptance criteria and risks.

The name is the methodSpectrum × scaling law

Eigen

J(x*) v₁ = λ₁ v₁
x(t) = Σₖ cₖ vₖ e^(λₖt)  ·  ρ(J) < 1 ⇒ stable

Business dynamics are linearized around the operating point: the dominant eigenpair (λ₁, v₁) of the Jacobian shows how fast the system grows or degrades and which variables carry that mode. That is where we intervene — not where the symptom is loudest.

Fractal

Y(sN) = s^β Y(N)
R[S] ≅ S  ·  β > 1 ⇒ increasing returns to scale

The same measure under change of scale: the exponent β shows whether a solution improves or degrades as volume, plants or business units multiply. We design modules the scaling operator leaves near-invariant: they hold in the cell, the plant and the group.

EigenFractal = where to intervene (spectrum of the dynamics) × how far it scales (power law).

Decision model

Value is not promised: it is formulated.

A company is not a deterministic function of one variable: it is a stochastic, multivariable system where each intervention moves several axes at once, under uncertainty and with coupling. So we do not optimize “cost” by default — we optimize the axis the diagnostic proves to be the priority, with the criterion written down and auditable.

Optimization model — validationEF-VAL · REV 07.26
01x ∈ ℝⁿ, ω ~ PState and uncertainty
The system and its noise: demand, prices, failures, people, market. P is estimated from data, not from convenient assumptions.
02θ ∈ ΘIntervention portfolio
Θ is the real feasible set: budget, calendar, team capacity, compliance and tolerated risk.
03ΔV(θ, ω) ∈ ℝ⁶Effect vector
Each intervention moves six coupled results at once: productivity, cost, revenue, innovation, engineering and knowledge.
04W ⪰ 0, ‖W‖₁ = 1Priority from the diagnostic
W weights the results by what your company needs today. It is estimated in the diagnostic and agreed in writing.
05V(θ) = 𝔼ω[⟨W, ΔV⟩] − λ·CVaRαRisk-adjusted expected value
Mean over scenarios minus the tail penalty: downside is paid in the criterion, not discovered in operation.
06ℙ(V > 0) ≥ 1 − εAcceptance criterion
Only what survives the confidence test is executed. Everything else goes back to model or experiment.
θ* = argmax θ∈Θ V(θ)The optimal portfolio maximizes risk-adjusted expected value inside the feasible set. Change W and θ* changes: there is no single right answer, there is a right answer for your priority.

Without W there is no optimization, only opinion. The diagnostic exists to estimate W, P and Θ before capital is committed; the measured effect in operation must fall inside the prediction interval — if it does not, the model is corrected, not the report.

Working principles
01Problems are formalized before they are solved.
02A solution that cannot be measured is not finished.
03Models are instruments, not substitutes for reality.
04Uncertainty is modeled; it is not removed with promises.
05Evidence prevails over authority and intuition.
06The whole system matters more than optimizing one part.

Does your problem justify this level of analysis?

Not every problem does. Bring us the case and we will say plainly whether it needs brief advisory, a study, a development or a full program.

Discuss a projectBack to home