AInora
Back to reports

AInora Report · April 2026

EU Debt Collection Statistics

European collections is widely quoted and thinly measured. This report keeps the supervisory figure we could open the source document and read, quotes it from the table it sits in, explains why recovery rates and cost per recovered euro have no public measurement, and lists the figures an earlier version of this page carried together with what happened when each was traced.

1.9%
EU and EEA bank NPL ratio, December 2024, weighted average (EBA Risk Dashboard Q4 2024)
1.0% / 2.5%
The 25th and 75th percentile across countries. The EU average hides a spread of more than two to one
No figure
Recovery rate, cost per recovered euro, or agency headcount. No reachable source publishes these
1 of 12
Figures on the earlier version of this page that survived a source check

What the supervisor actually publishes

The European Banking Authority is the one body in this subject that publishes on a fixed schedule, with a defined population and a downloadable document. We quote it from the table rather than from a summary of it, and we state what the document does not contain as carefully as what it does.

1.9%
Share of non-performing loans and advances across EU and EEA banks, December 2024, weighted average
EBA Risk Dashboard Q4 2024, indicator 8, "Ratio of non-performing loans and advances (NPL ratio)". Read from the dispersion table row "Dec-24", which gives 1.9% weighted average, 1.0% at the 25th percentile, 1.8% median and 2.5% at the 75th percentile by country.
Flat
The ratio has been between 1.8% and 1.9% every quarter from March 2022 to December 2024. There is no deterioration trend in this series to point at
Same table, quarterly rows Mar-22 through Dec-24. Worth stating because the capacity-crisis framing usually assumes a rising line.
EUR 375.0bn
Non-performing loans and advances outstanding across EU and EEA banks, December 2024
EBA Risk Dashboard Q4 2024, statistical annex, table headed "Volumes bn EUR", row "EU/EEA", column "Dec-24". Corroborated by the EBA press release of 21 March 2025, which states that NPLs decreased 1.1% quarter on quarter, "amounting to EUR 375bn". Worth knowing where it lives: the headline indicator chart expresses the NPL numerator as an index set to 100 at December 2014, so a reader who checks only the chart concludes the document carries no euro total. It does, in the annex. The EUR 373bn that circulates for this figure is close to it but is not the published number.
Banks only
The population is supervised credit institutions. Receivables held by utilities, telecoms, landlords, healthcare providers and collection agencies are outside it entirely
Any figure describing "European receivables in collections" is therefore measuring something the EBA does not measure, and cannot be sourced to it.

What is not measured, and why

Recovery rates, cost per recovered euro and agency headcount are the three numbers buyers ask for, and all three are commercially sensitive figures held by private firms. No trade body we could reach publishes them with a sampling frame. This report carries none of them rather than an estimate presented as a measurement.

No figure
Recovery rate by delinquency age, for consumer or B2B debt
Recovery is a function of debt age, documentation quality, segment and jurisdiction. A single European range would have to average across all four, and no published study we could reach constructs one.
No figure
Cost per euro recovered, or fully loaded cost per outbound dial
These are agency margin figures. Agencies do not publish them and no analyst document we could open contains them.
No figure
Number of licensed collections professionals in the EU
Licensing regimes differ by member state and several do not licence collectors as a distinct profession, so a single EU headcount would need a definition no source states.
Not 24/7
Permitted contact hours for collections calls. An earlier version of this page stated that most EU jurisdictions allow a 24/7 reminder window, which is wrong as well as unsourced
Member states restrict debt-collection contact hours, and the restrictions differ by country. Treat permitted hours as a per-jurisdiction legal question, never as a European default.

What you can compute from your own ledger

The figures this report will not supply are all derivable from your own receivables data, and they will be more useful than a European average would have been because they carry your own segment mix.

Step 1
Recovery rate by cohort. Take one month of accounts entering delinquency, follow that same cohort forward, and report recovery at 30, 60, 90 and 180 days
Cohort tracking rather than a period snapshot. A snapshot mixes new and aged accounts and flatters the result.
Step 2
Cost per euro recovered. Total collections operating cost for a period, divided by euros actually collected in that period, including the accounts you never reached
Excluding unreached accounts is the most common way this number gets quietly improved.
Step 3
Contact rate by attempt number. Share of accounts reached on the first, second and third attempt, split by time of day within your permitted window
This is where automation either helps or does not, and it is measurable before you buy anything.
Step 4
Right-party contact rate specifically, not connection rate. Reaching a number is not reaching the debtor, and only the second one recovers anything
Keeping these two apart is the single most useful discipline in collections reporting.

Figures this report no longer uses

These appeared on earlier versions of this page. Each was traced and each failed. They are listed with their former attribution so the trace is not repeated and none is quietly re-added.

1 trillion euro
"European receivables in collections at any given time", formerly attributed to FENCA 2024 and carried in the page subtitle.
Cited only to the FENCA home page, never to a document. It also sat in the subtitle, which this component passes straight into the Dataset.description of the page structured data, so an unsourced number was shipping as machine-readable data. Removed from both at once.
373 billion euro
"Outstanding EU bank NPLs at end of 2024", attributed to the EBA. The attribution was to a real and correct publication. The figure is not in it.
We downloaded the EBA Risk Dashboard Q4 2024 PDF and its Data Annex and searched both. The string 373 does not occur in the dashboard, and the NPL numerator is published as an index rather than in euros. The ratio from the same document is kept above. This is the failure worth naming: a real source, correctly cited, that does not contain the number attributed to it.
67,000+
"Licensed collections professionals in the EU", formerly attributed to FENCA 2024.
Same home-page citation, no document, and no stated definition of what counts as licensed. Removed.
45-65% / 20-35%
Consumer debt recovery rates by delinquency age, formerly attributed to "ACA Europe 2024".
There is no European debt-collection body of that name. ACA International is the United States trade association, and ACA Europe is the association of the supreme administrative courts of the EU member states, which has nothing to do with collections. An attribution to a body that does not exist in this field is a generated citation, not a mistaken one.
12-22%
B2B recovery on debt over 180 days, formerly attributed to the Atradius Payment Practices Barometer.
The Barometer is a real annual publication, but it reports days sales outstanding, overdue shares and write-off shares, not a recovery-rate range by delinquency age. Another case of a real source cited for a quantity it does not report.
5-12%
Recovery on debt sold to NPL portfolio buyers, formerly attributed to "Prelios / Cerved 2024".
Cited to a company home page with no document named. Removed.
55-70% / 0.22-0.35 euro / 3.50-7 euro
Labour share of operating cost, cost per euro recovered, and fully loaded cost per human outbound dial, all formerly attributed to "Everest Group 2024".
Three separate agency-economics figures, all citing the same analyst home page, none naming a report. Everest Group attributions failed the same trace on our voice AI adoption report and were removed there too.
0.27 euro / 12-28%
Headline "average cost per euro recovered, pre-AI" and "typical 90-day recovery rate across segments". Neither carried any source at all.
An unattributed number in a headline tile is the strongest claim on the page and the weakest sourced. Both removed.
3-5x
"Outbound dial rate increase with AI voice", formerly attributed to "ACA Europe 2026".
Future-dated, attributed to the same non-existent body as the recovery rates above, and a performance claim about our own product category besides. Removed.

Methodology

One figure on the earlier version of this page survived a source check. We downloaded the EBA Risk Dashboard for Q4 2024 and its Data Annex, read the non-performing loan table directly, and quote the December 2024 row with its country dispersion rather than a rounded restatement. Everything else was traced and removed: figures attributed to FENCA, Atradius, Prelios, Cerved and Everest Group named no document and linked only to an organisation home page; figures attributed to ACA Europe named a body that does not operate in this field at all; and the 373 billion euro NPL total, though attributed to the correct and genuine EBA publication, does not appear anywhere in that publication or its annex, which reports the numerator as an index rather than in euros. Where a real source was cited for a quantity it does not report, we say so on the figure rather than swapping in a different source for the same number. No recovery, cost or headcount figure has been replaced with a substitute. No figure here is our own measurement, and this report claims no client results of any kind.

Sources

Turn these numbers into operations

AINORA is the AI voice agent that answers the missed calls, closes the debt collection gap, and scales human-quality conversations without hiring.

See AINORA for debt collection