LUMINO · DATA QUALITY INDEX

How the DQI is calculated

Every company in Lumino carries a Data Quality Index from 0 to 100. It doesn’t grade the company — it grades our evidence about the company: how it’s proven, how complete it is, and how rich the picture is. This page shows exactly how the number is built, because a quality score you can’t inspect isn’t worth trusting.

methodology version: dqi_v2
STRUCTURE

Two halves make the whole

Marketing and IT each earn their own score from the same three elements. The company DQI is simply both halves of one ring — each function fills its side, and the center shows what they make together. A function with no published data leaves its half empty, and the whole ring shows it.

DQI · Marketing
94
fills the left half
Evidence 50Coverage 30Depth 20
MKTGIT
Company DQI
84
Primarily disclosed
DQI · IT
74
fills the right half
Evidence 50Coverage 30Depth 20

Shown here: a typical US public discloser — strong Reported marketing on the left, externally-modelled IT on the right, meeting in the middle at 84.

THE THREE ELEMENTS

Each half is built from three questions

Same three questions for both functions. What counts toward each answer differs slightly — because marketing evidence and IT evidence come in different shapes.

50of 100

Evidence

How is each number proven?

Every published figure carries a tier. Evidence scores the tier mix of a company’s data — weighted so its most important metrics count most.

Marketing

The tier of the marketing spend figures.

IT

The tiers of revenue and IT spend together.

30of 100

Coverage

How much of the expected picture exists?

Each company is expected to have its core metrics for the 3 most recent fiscal years — including the latest one. Gaps and honest refusals lower Coverage, and only Coverage.

Marketing

Spend figure present per expected year.

IT

Revenue + IT spend present per expected year.

20of 100

Depth

How rich and coherent is the picture?

Beyond the headline numbers: corroborating sources, supporting signals, and internal consistency. Two companies can share a tier mix and still differ in how much surrounds the numbers.

Marketing

Scope precision, corroboration, year-over-year consistency.

IT

Grounding specificity, confirmed signals (tech capex, transformation programs), segment detail.

EVIDENCE CREDITS

Not all evidence is equal — and the score says so

Inside the Evidence element, each published figure earns credit based on its tier. A disclosed number is worth more than a derived one, which is worth more than a peer-based estimate.

Reported
1.00
Inferred
0.70
Modelled
0.40

Modelled data earns real credit deliberately — a reviewed, versioned, benchmark-grounded estimate is genuine value, just weaker evidence than a disclosure. Zero is reserved for absence, not honesty.

READING THE SCORE

Bands describe the evidence — they don’t judge it

Instead of grades like “good” or “poor,” DQI bands tell you what kind of evidence dominates. A luxury house that discloses nothing isn’t failing; our data about it is simply built differently.

Primarily modelled
Mixed evidence
Primarily disclosed
05080100
Partial coverage sits outside the 0–100 evidence scale. When one function has no published data at all, the company isn’t “primarily modelled” — its covered function may be excellent. We say so plainly (e.g. “Marketing only — IT not yet covered”) and hold the empty half open until real data fills it. The three evidence bands above describe a company only once both functions are covered.
TRY IT

Build a score yourself

Assemble one function’s evidence below — click each cell to cycle its tier — then choose what the company’s other function looks like, and watch both halves of the ring make the company score.

Your function — two core metrics × three fiscal years

Click a cell to cycle its tier: missing → Modelled → Inferred → Reported

FY2023
FY2024
FY2025
Metric A
Metric B
Reported = 1.0 credit · Inferred = 0.7 · Modelled = 0.4 · missing = no credit, no coverage
The company’s other function
MKTGIT
Company DQI
75
Mixed evidence
Evidence
35.0/50
Coverage
30.0/30
Depth
12.0/20

Left half: the function you’re building. Right half: the preset. The center is their meeting point.

IN PRACTICE

What real profiles look like

Three archetypes from the actual data landscape — the same company can carry very different evidence on its two functions, and the ring is designed to show that, not smooth it over.

US public discloser

Discloses advertising spend in its 10-K every year. Revenue Reported; IT spend externally modelled — like most of the industry.

Company DQI
84
Mktg 94IT 74
Primarily disclosed

European luxury house

Doesn’t break out marketing — an honest non-disclosure with a peer-modelled estimate. Revenue fully Reported with real segments.

Company DQI
64
Mktg 56IT 72
Mixed evidence

Private company

No public filings exist. Everything is honestly modelled from peers and benchmarks — or honestly refused where even that isn’t grounded.

Company DQI
33
Mktg 34IT 31
Primarily modelled
THE RULES

What the score lives by

Published data only

Only human-reviewed, published figures — and reviewer-confirmed signals — ever influence a score. Nothing unreviewed touches it.

Versioned & snapshotted

Every score records the methodology version and the exact inputs it was computed from. When a score changes, the reason is inspectable.

Drillable everywhere

The score is never a bare number. Every place it appears, it opens into its three elements — and each element opens into the actual data behind it.

Honesty is never penalized twice

A refused estimate lowers Coverage — there’s no number — but never Evidence. Declining to guess is the highest-quality choice available.

← Back to Data Foundations