VERITAS · the measurement methodology

Prove the P&L — against what would have happened anyway.

Plan-vs-actuals isn’t proof.
VERITAS isolates the causal effect of the decision
and attributes the value to your P&L.

See it on your data How it works

⚡ Out-of-fold · counterfactual · direction-aware — the number that backs outcome-based pricing

P&L · counterfactual vs realizedlive
PAV baselinedecisiontoday
Realized Counterfactual Pragma-Attributed Value

Operating partner to teams at

Nidec PepsiCo Danone Globo Electrolux Philips Eversendai

The problem with “impact”

Most reported value is a story. VERITAS makes it a number.

A consultancy tells you the initiative “drove” a result. But the market moved, prices shifted, a competitor stumbled — and the number conveniently absorbs all of it. VERITAS separates the effect you caused from everything that would have happened anyway.

 
Reported impact
VERITAS · proven impact
Baseline
Plan, or last year — whatever flatters the result
Counterfactual — the world where the decision wasn’t taken
Validation
In-sample — fit on the same data it claims to explain
Out-of-fold — held-out, leakage dropped, direction-aware
Attribution
Credit for the whole topline swing, market noise included
Isolated effect — market, seasonality & price stripped out
Cadence
A slide in the final report, then never revisited
Tracked over time — re-checked automatically as data arrives
What it buys
A number nobody can defend to the CFO
A defensible, attributable figure — the base of outcome pricing

If you can’t name the counterfactual, you don’t have a result — you have a coincidence.

The methodology

Every decision ships with a measurement contract.

Before a workflow touches revenue, margin, cost or working capital, it must declare three things up front — pre-registered, so the result can’t be reverse-engineered to look good after the fact. This is what turns “we think it helped” into proof.

01

Pre-registered

The estimand

What outcome does this decision claim to move — named and frozen before you act, so there’s no shopping for a flattering metric later.

e.g. incremental gross margin on the 8 repriced SKUs, NE region, next 12 weeks

02

Counterfactual

The baseline source

What would have happened anyway — a held-out comparison built from untreated units, not the plan and not last year.

e.g. a synthetic control of comparable SKUs × stores that were left at the old price

03

Attribution

The estimation method

How the causal effect is isolated — the econometric method is declared with the decision, matched to its design.

e.g. difference-in-differences, with Shapley allocation across shared levers

Recorded on every decision — a decision passport that travels with the value

Decision passport · #4821 pricing · fresh dairy · NE region · signed by committee
Estimand
Incremental gross margin
8 SKUs · 12 weeks
Counterfactual
Synthetic control
42 matched SKU×store cells
Method
Difference-in-differences
+ Shapley allocation
Confidence
94%
p < 0.05 · out-of-fold
Projected (pre)
+R$ 8.2M
modeled at decision time
Realized (post)
+R$ 7.6M
measured vs counterfactual
Market (excluded)
−R$ 1.1M
stripped out as noise
Status
Attributed
tracked · re-checked weekly

Breaking down the P&L

We don’t claim the swing. We decompose it.

The topline moved +R$ 11.4M. The honest question is how much of that you caused. VERITAS bridges from the counterfactual baseline to today — crediting each lever, and pushing market noise out of the result entirely.

P&L bridge · baseline → today · gross margin (R$ M) attribution · out-of-fold
R$ 84.0
Counterfactual baseline
+3.1
Pricing
+2.2
Trade & promo
+1.7
Allocation
+0.6
Service triage
+3.8
Market & season
(excluded)
R$ 95.4
Realized today
Attributed to decisions · +R$ 7.6M Market & seasonality · +R$ 3.8M — not ours to claim

Shapley allocation across shared levers. When pricing, promo and allocation all touch the same revenue, we split the credit fairly — no double-counting, no lever taking the whole win.

Market noise pushed out, not absorbed. The +R$ 3.8M from a rising category and seasonality is identified and excluded — so the figure you report is only the part you can defend.

Getting the causal effect

Four ways to build the world where you didn’t act.

There’s no single right method — the design of the decision dictates the estimator. VERITAS picks the one that fits, declares it on the passport, and validates out-of-fold. Market-mix-modeling lineage throughout.

Method 01when you can’t randomize

Synthetic control

Build a weighted clone of the treated unit from untreated ones. The gap that opens after the decision is the effect — clean even with a single treated region.

intervention
Method 02treated vs control group

Difference-in-differences

Compare the change in the treated group to the change in a control over the same window. Shared trends cancel; what remains is the lift you caused.

Method 03when you can switch on/off

Switchback experiments

Turn the intervention on and off across time blocks and measure the swing. Ideal for pricing, routing and supply decisions where A/B isn’t clean.

ONOFF
Method 04many levers, one outcome

Shapley allocation

When several decisions move the same number, game theory splits the credit fairly across them — every lever gets exactly its marginal contribution.

Pricing 41% Promo 29% Allocation 22% Triage 8%

Simulate before you commit

Move the levers. Watch the P&L respond.

The same causal model that measures the past projects the future. Drag a lever and VERITAS re-estimates the outcome live — with a confidence band, not a point guess — so the committee decides on the trade-off curve, not a hunch.

What-if leverslive model
Price change+4%
elasticity −1.6 · from 142 weeks of history
Trade & promo depth28%
diminishing returns past ~35%
Service level held97%
below 95% churn risk rises sharply

Modeled out-of-fold · re-estimated on every change

Projected P&L · next 12 weeks
nowwk 12
Projected PAV
+R$ 8.2M
90% band
±1.4M
Confidence
94%

Illustrative model. On your data, the curve and band are fit to your own history — and the chosen scenario is written straight onto the decision passport.

VERITAS · inside Signal

It isn’t a slide. It’s a measurement view that’s always live.

Live in PRAGMA OS

VERITAS runs inside Signal, the portfolio cockpit. Every decision carries its passport — estimand, counterfactual, method — and the realized value is re-estimated against the baseline as new data arrives. The P&L attribution, the confidence, and the executive narrative update themselves. Multi-tenant, board-ready, always current.

signal.pragma.do/macro

Governance · measurement

VERITAS · attributed value

⚡ Ask operator
PAV · year to date
R$ 65M
vs counterfactual
Decisions attributed
23/28
+6 this Q
Out-of-fold accuracy
92%
held-out
Avg. confidence
94%
p < 0.05
P&L · counterfactual vs realized · gross marginout-of-fold →
Measurement narrative

Pricing in the NE region is the strongest attributed lever at +R$ 31M, validated out-of-fold against a synthetic control. Trade & promo adds +R$ 22M; allocation +R$ 12M. A category tailwind of ~R$ 4M is identified and excluded — not counted as ours. Total PAV holds at +R$ 65M.

Auto-generated · re-estimated 2h ago

Decisions · measured23 attributed
Pricing · +4% on 8 SKUs
DiD · +R$ 31M
Attributed
Trade & promo reallocation
Switchback · +R$ 22M
Attributed
Scarce-SKU allocation
Synthetic control · +R$ 12M
Tracking
Service-triage copilot
Pre-registered · baseline set
Measuring

A live VERITAS view inside Signal — every decision’s realized value measured against its counterfactual, tracked to the P&L.

Three layers, working together

It isn’t a dashboard. It’s the decision layer, operated.

Portfolio

Every decision on one pipeline.

Opportunities → use cases → business cases → initiatives → roadmap. Scored on impact × feasibility, with value rolled up to the P&L.

↓
Context layer
your data · your IP · your decisions
↓

Context layer

Memory that compounds, not drifts.

Every engagement’s data, decisions, and in-flight IP kept warm across cycles. This is the layer that makes Signal appreciate.

Committee
VERITAS
Risk library
Roadmap
Business cases
Multi-tenant

Governance & proof

Approved in place. Proven out-of-fold.

The committee signs off; VERITAS measures against the counterfactual; risk is pre-classified. Multi-tenant and board-ready.

This is the decision layer, operated. Not a dashboard. Not a deck.

The proof object

One screen the CFO can actually sign.

Counterfactual, causal drivers, and the attributed total — on a single object that travels with the decision. This is the number that backs outcome-based pricing.

P&L · baseline vs realizedout-of-fold
PAV baselinepilottoday
Realized Counterfactual Attributed value
Causal driversattributed
Pricing+R$ 31M
Trade & promo+R$ 22M
Allocation+R$ 12M
Market (excluded)noise
Pragma-Attributed Value
+R$ 65M
proven vs the counterfactual — the base of outcome-based pricing.
Method

Attribution

Market-mix-modeling lineage. We isolate the real effect of the decision from market noise.

Counterfactual

Vs. what would have happened

Out-of-fold, direction-aware. We measure against the world where the decision wasn’t taken — not plan-vs-actuals.

PAV

Pragma-Attributed Value

Value proven and attributed to the P&L — the base of our pay, and the financeable moat.

Proof in production

Attributed outcomes, measured in real P&Ls.

NidecS&OP · Inventory
−35%
inventory, in 6 months

S&OP and inventory optimization, plus a decision-intelligence & KPI management platform.

PepsiCoIntegrated Business Plan
+R$ 35–85M
annual margin · S&OP cycle −40–50%

Integrated Business Plan across channels and processes — faster, evidence-based decisions.

DanoneCommercial ops
−R$ 70M
in write-offs & stock-outs

Optimized write-offs and stock-outs caused by commercial decisions.

EversendaiValue-chain orchestration
+15 p.p.
orchestration simulation

Procurement, planning, production and construction orchestrated as one simulated value chain.

GloboDecision Intelligence
85%
audience-growth accuracy

A decision-intelligence platform for what to build and who to engage.

ElectroluxEmployee experience
9.0
satisfaction · 6s response

“Luz”, a synthetic HR worker answering policy, benefits and process questions in PT/ES/EN.

AxiaCommercial intelligence
R$ 300M
potential return · 4× performance

“CassIA”, a synthetic support worker for pricing and discount intelligence.

PhilipsLogistics
−15%
logistics cost · ~US$30M

A demand-allocation engine that cut logistics cost across the network.

Read the case →

← drag · eight programs in production

Paid when the P&L moves — not when the deck lands.

Measured against what would have happened anyway. Proof, not attribution theater.

NidecPepsiCoDanoneEversendaiGloboElectroluxAxiaPhilips

Why it matters

A proven number changes how you can be paid.

Once value is attributable instead of asserted, the whole commercial model flips. VERITAS turns measurement reliability into the asset everything else is priced on — the moat no slide-deck consultancy or generic agent platform can copy.

PAV

Pragma-Attributed Value — value proven against the counterfactual, and attributed to your P&L.

01
Outcome-based pricing
Paid when the P&L moves, not when the deck lands — because the move is now measurable and defensible.
02
A CFO-grade audit trail
Every claim carries its estimand, counterfactual and method on a decision passport. No attribution theatre.
03
A financeable record
An accumulating, attributed measurement history — the asset that compounds and can be underwritten.

Reported impact is a story. VERITAS makes it a number you can bank.

Prove it on your P&L.

Bring one decision and one dataset. In a single session you’ll see the counterfactual, the isolated causal effect, and the value VERITAS can attribute — out-of-fold.

See it on your data