Where the numbers come from, and what they're allowed to say.

The rule under everything

Every claim traces to something you can check.

A client-facing statement in an Arristal report points back to one of six things: a driver score, a detected signal, a piece of evidence, a named source, a deterministic calculation, or a reviewer's judgement. If a claim cannot be traced to one of those, it does not go in the report.

• Score
• Signal
• Evidence
• Source
• Calculation
• Reviewer judgement

The evidence we trust, in order

Better evidence wins. Always.

When two sources disagree, the higher one decides. A self-reported figure does not override an audited statement, and a sector reference never overrides a number you can stand behind.

1. 01Audited public financialsFiled statements. The strongest base for any € figure.
2. 02Verified client dataOperational KPIs and figures you supply and stand behind.
3. 03Self-reported rangesWhat you tell us during the assessment, used as a range, not a point.
4. 04External public contextSector, trade, and logistics references. Context only — never the source of a value.
5. 05Disclosed assumptionA conservative placeholder, always labelled as an assumption in the report.

Your history outranks the sector

Your own trend is the better benchmark.

Where your own history is reliable, it takes priority over a broad sector reference. How your inventory turns or cash conversion moved over three years says more about you than a peer-set median ever could. Sector references fill the gaps; they do not replace your own record.

What benchmarks can and can't do

A benchmark can frame a number. It cannot create one.

One distinction worth holding: external public context is not peer benchmarking. Peer benchmarking would compare you against anonymised Arristal assessment data, and that begins only once the dataset is large enough to be fair. We do not sell the diagnostic as a benchmark product, and we do not imply peer comparison we cannot yet support.

External intelligence stays in its lane

Public context informs the reading. It does not move the score.

In the paid modes, and only where there is dated evidence behind it, Arristal draws on public context — country, trade, logistics, and sector conditions. It contextualises, validates, challenges, or reduces our confidence in a finding. It does not create scores, create signals, create financial values, or override your input by default. No external claim appears in a report without a source and a date.

The limits that keep a number honest

Hard ceilings, so an estimate can't run away.

The financial translation runs deterministically, before any interpretation. Fixed ceilings cap what the model is allowed to claim, so a long chain of signals can never compound into a figure no CFO would believe.

• ≤ 30%

  of inventory

  Working capital opportunity

• ≤ 5%

  of revenue

  Revenue at risk

• ≤ 15%

  of revenue

  Total mid-case estimate

• ≤ 40%

  cumulative

  Cost-signal uplift

The same economic effect is never counted twice through two different signals. Where signals overlap, the higher-confidence one is used and the rest are recorded but excluded from the figure.

Confidence, stated plainly

We tell you how sure we are, and act on it.

Every score, signal, and value range carries a confidence view. It decides how strongly we state a conclusion and how wide the range becomes. Low confidence does not get dressed up as certainty.

High

Multiple sources agree and the data is current. We proceed after a standard review.

Moderate

One evidence class is partial. The report states the limitation and widens the range.

Low

Evidence is thin, old, or contradictory. We avoid strong language and route it to a person.

Escalated

A reviewer signs off before the report is released. Nothing goes out on autopilot.

A person reviews the paid reports before they reach you. External Validation always carries human review, and no report is released automatically.