It is not wrong as a budgeting method. It is wrong as a verification method. A sample describes a population. A file describes an asset. Provisioning, disposal, enforcement and audit all happen at asset level, where the sample has nothing to say.

How long is a field inspection good for?

One day. That is the shelf life of the observation. The status derived from it is then carried in the file for months, undated in practice even when dated on paper.

What does a physical inspection cost?

€228–342 per asset, per check, in Iberia. That price is why coverage gets rationed to a sample rather than applied to the whole book.

What happens when the evidence disagrees?

The asset is returned UNKNOWN. The specific missing source is named and a resolving pass is scheduled automatically. No verdict is issued on an asset where the layers do not agree.

Can you tell us who is inside?

No. The question answered is whether the asset is occupied, not by whom. We verify the asset, never the occupant.

DV-I5 · Insight

Why sampling a loan book does not verify it

Sampling tells you what a subset of assets looked like on the day someone visited them. It does not tell you the status of the asset in front of you, which is the only question an auditor, a buyer or a court ever asks.

A sample is a budget decision wearing the clothes of a method. It exists because a physical inspection in Iberia costs €228–342 per asset, per check, and a book of thousands cannot absorb that. So a percentage gets visited, the rest gets inferred, and the inference is carried in the file as though it were a finding.

Inference is not evidence. It is a statement about a population. The file is about an asset.

Where the sample breaks

It answers at portfolio level. Every downstream question — provisioning, disposal, enforcement, audit — is asked at asset level.
It ages immediately. A field visit is accurate for one day. The status it produces is then carried for months.
It is not evenly wrong. Vacancy clusters. The assets most likely to have changed status are the ones least likely to be visited.
It cannot be re-run. Asking what an unsampled asset looked like last quarter means sending someone now, and now is not last quarter.

The question underneath

An auditor does not ask what percentage of the book was inspected. The auditor points at one line and asks how you know this asset is vacant. "The servicer said so" is not evidence. Neither is "it fell outside the sample."

Reframe: the problem is not vacant assets. The problem is not knowing which ones are vacant, and carrying a status you cannot date.

What full coverage requires

Coverage stops being a budget question when the per-asset cost of a check stops being a truck roll. That is the substitution. No hardware, no installs, no field agent — an address list in, a dated verdict out, ten assets or ten thousand.

01Submit the listAddress, cadastral reference if held, and the status currently carried against it.
02Run three independent evidence layersEach sourced, timestamped and geolocated independently of the others.
03ReconcileSources must agree before an asset is certified.
04Return a verdictVERIFIED, UNKNOWN or OCCUPIED. Dated, referenced, signed.

UNKNOWN is the honest cell

Where the layers conflict, the asset comes back UNKNOWN, the specific missing source is named, and the resolving pass is scheduled automatically. That is not a gap in the method. It is the method refusing to guess where a sample would have quietly extrapolated.

What lands in the file

Artefact	Purpose
PDF certificate per asset	The document handed across the table when one line is queried.
Machine-readable twin	XML or JSON, with fields shaped to the core system that has to hold them.
Portfolio roll-up	Every asset sorted by verdict, so the exceptions are the work list.
Method note	Sources, dates, coordinates, chain of custody. Full methodology disclosed under NDA.

Cadence

Monthly, quarterly, or event-driven — pre-sale, pre-filing, pre-audit. Live in Spain and Portugal. Italy, Greece, Ireland, Cyprus, France and the Netherlands on instruction, with lead time agreed before the scan is accepted. We will tell you what we cannot reach rather than quote for it.

Want an example?

Questions we get asked

Is sampling wrong?: It is not wrong as a budgeting method. It is wrong as a verification method. A sample describes a population. A file describes an asset. Provisioning, disposal, enforcement and audit all happen at asset level, where the sample has nothing to say.
How long is a field inspection good for?: One day. That is the shelf life of the observation. The status derived from it is then carried in the file for months, undated in practice even when dated on paper.
What does a physical inspection cost?: €228–342 per asset, per check, in Iberia. That price is why coverage gets rationed to a sample rather than applied to the whole book.
What happens when the evidence disagrees?: The asset is returned UNKNOWN. The specific missing source is named and a resolving pass is scheduled automatically. No verdict is issued on an asset where the layers do not agree.
Can you tell us who is inside?: No. The question answered is whether the asset is occupied, not by whom. We verify the asset, never the occupant.

Want an example?