An evidence-backed view of what was examined, what stands out, and where to explore next.
Run
15 October 2026, 15:49 UTC
Created
15 October 2026, 15:50 UTC
Output version
0.1
Profile scope
Full
Contents
Orient
Overview
What was examined, the retained evidence scope, and the boundary at a glance.
Records represented9,309
Sources7
Fields13
Rows observed in profile9,309
ProfilingPerformed
Population terrainPerformed where eligible
Operational markersPerformed where eligible
MatchingPerformed
PII assessmentPerformed where eligible
7 sources examined
Source
Kind
Records
Fields
Profile
Format
Findings
File 1
HMRC_spending_over_25000_for_April_2026.csv
File
1,691
10 fieldsSchema bookkeeping
10 shared · 0 source-specific · 3 not present relative to the examination field union. Absence from another structure is not a missing value in this source.
10 shared · 0 source-specific · 3 not present relative to the examination field union. Absence from another structure is not a missing value in this source.
10 shared · 0 source-specific · 3 not present relative to the examination field union. Absence from another structure is not a missing value in this source.
10 shared · 0 source-specific · 3 not present relative to the examination field union. Absence from another structure is not a missing value in this source.
10 shared · 0 source-specific · 3 not present relative to the examination field union. Absence from another structure is not a missing value in this source.
10 shared · 0 source-specific · 3 not present relative to the examination field union. Absence from another structure is not a missing value in this source.
10 shared · 3 source-specific · 0 not present relative to the examination field union. Absence from another structure is not a missing value in this source.
What stands out in the retained examination. Notability describes prominence, not severity or error.
Medium notabilityScope
Input sources have non-identical field sets
Across 7 sources, 10 fields are shared and at least 3 fields occur in only a subset of sources. Affected sources: File 1, File 2, File 3, File 4, File 5, File 6 and 1 more. Fields present in only a subset: Contract Number, Project Code, Supplier Type.
Explore: Assess whether the source-specific fields reflect expected source coverage or a schema difference that changes the examination scope.
Why this was noticed
Input sources have non-identical field sets
Across 7 sources, 10 fields are shared and at least 3 fields occur in only a subset of sources. Affected sources: File 1, File 2, File 3, File 4, File 5, File 6 and 1 more. Fields present in only a subset: Contract Number, Project Code, Supplier Type.
Several retained values recur unusually often within the same observed structural-group scope.
2 underlying findings · retained examples, not an exhaustive list. Counts belong to each individual population.
Amount · recurrent value “43792.19” · 3 records · retained example · Observation scope 1
Medium notabilityPopulation
Amount · recurrent value “43792.19”
3records· 0.3% of 1,100 usable Amount values examined
Exact over the full observed source population
No single pattern stood out strongly; some patterns are worth examining below.
Explore: Does this recurrence represent an intentional shared identifier, a source-system convention, or records that warrant closer investigation?
Why this was noticed
Amount has a recurrent identifier-like value
Character43792.19 occurs in 3 records. Amount is 98.9% distinct overall; the value occurs about 3.0x the field-average multiplicity implied by that distinctness.
ExplanationNo retained source/category dimension strongly characterised where this population occurs.
Amount · recurrent value “28652.40” · 4 records · retained example · Observation scope 1
Medium notabilityPopulation
Amount · recurrent value “28652.40”
4records· 0.4% of 1,100 usable Amount values examined
Exact over the full observed source population
No single pattern stood out strongly; some patterns are worth examining below.
Explore: Does this recurrence represent an intentional shared identifier, a source-system convention, or records that warrant closer investigation?
Why this was noticed
Amount has a recurrent identifier-like value
Character28652.40 occurs in 4 records. Amount is 98.9% distinct overall; the value occurs about 4.0x the field-average multiplicity implied by that distinctness.
ExplanationNo retained source/category dimension strongly characterised where this population occurs.
27records· 0.3% of 8,209 usable Amount values examined
Exact over the full observed source population
100% Expense type=employee education (27/27 eligible population members) · 100% Expense area=hr (27/27 eligible population members)
Explore: What process or record population explains why this concrete value recurs across these records?
Why this was noticed
Amount has an unexpectedly recurrent value
Character£24,104.13 occurs in 27 records. Amount is 90.1% distinct overall; the value occurs about 24.3x the field-average multiplicity implied by that distinctness.
Explanation100% of 27 population members with usable Expense type evidence were employee education; 1.0% of 8,209 records with usable Expense type evidence globally overall (105.24× lift)
Several retained values recur unusually often within the same observed structural-group scope.
2 underlying findings · retained examples, not an exhaustive list. Counts belong to each individual population.
Transaction number · recurrent value “5100012634” · 18 records · retained example · Observation scope 2
Medium notabilityPopulation
Transaction number · recurrent value “5100012634”
18records· 0.2% of 8,209 usable Transaction number values examined
Exact over the full observed source population
No single pattern stood out strongly; some patterns are worth examining below.
Explore: Does this recurrence represent an intentional shared identifier, a source-system convention, or records that warrant closer investigation?
Why this was noticed
Transaction number has a recurrent identifier-like value
Character5100012634 occurs in 18 records. Transaction number is 87.2% distinct overall; the value occurs about 15.7x the field-average multiplicity implied by that distinctness.
ExplanationNo retained source/category dimension strongly characterised where this population occurs.
Fields involved Transaction number · Analysis Population
Transaction number · recurrent value “5100070190” · 18 records · retained example · Observation scope 2
Medium notabilityPopulation
Transaction number · recurrent value “5100070190”
18records· 0.2% of 8,209 usable Transaction number values examined
Exact over the full observed source population
No single pattern stood out strongly; some patterns are worth examining below.
Explore: Does this recurrence represent an intentional shared identifier, a source-system convention, or records that warrant closer investigation?
Why this was noticed
Transaction number has a recurrent identifier-like value
Character5100070190 occurs in 18 records. Transaction number is 87.2% distinct overall; the value occurs about 15.7x the field-average multiplicity implied by that distinctness.
ExplanationNo retained source/category dimension strongly characterised where this population occurs.
Fields involved Transaction number · Analysis Population
Several retained values recur unusually often within the same observed structural-group scope.
2 underlying findings · retained examples, not an exhaustive list. Counts belong to each individual population.
Transaction number · recurrent value “5100008310” · 7 records · retained example · Observation scope 1
Medium notabilityPopulation
Transaction number · recurrent value “5100008310”
7records· 0.6% of 1,100 usable Transaction number values examined
Exact over the full observed source population
No single pattern stood out strongly; some patterns are worth examining below.
Explore: Does this recurrence represent an intentional shared identifier, a source-system convention, or records that warrant closer investigation?
Why this was noticed
Transaction number has a recurrent identifier-like value
Character5100008310 occurs in 7 records. Transaction number is 91.7% distinct overall; the value occurs about 6.4x the field-average multiplicity implied by that distinctness.
ExplanationNo retained source/category dimension strongly characterised where this population occurs.
Fields involved Transaction number · Analysis Population
Transaction number · recurrent value “5100006717” · 6 records · retained example · Observation scope 1
Medium notabilityPopulation
Transaction number · recurrent value “5100006717”
6records· 0.5% of 1,100 usable Transaction number values examined
Exact over the full observed source population
No single pattern stood out strongly; some patterns are worth examining below.
Explore: Does this recurrence represent an intentional shared identifier, a source-system convention, or records that warrant closer investigation?
Why this was noticed
Transaction number has a recurrent identifier-like value
Character5100006717 occurs in 6 records. Transaction number is 91.7% distinct overall; the value occurs about 5.5x the field-average multiplicity implied by that distinctness.
ExplanationNo retained source/category dimension strongly characterised where this population occurs.
Fields involved Transaction number · Analysis Population
Identical records observed: 21 repeated-row groups, with 25 surplus occurrences.
Compatible complete source structure: 10 canonical fields across 6 sources. engine-included records after configured scope filters; invalid/recovered or unreliable decoded records excluded.
6 sources examined · 21 within-source and 0 cross-source repeated groups (may overlap). Confirmation: Confirmed. Identical records are not necessarily errors.
Why this was noticed
Complete-row confirmation evidence
21 confirmed repeated-row groups · 46 participating records · 25 surplus occurrences. Exact within the observed eligible population.
Scope: engine-included records after configured scope filters; invalid/recovered or unreliable decoded records excluded. 8,209 eligible records; 0 excluded.
Oversized blocking bins limited the operational search space
74 oversized route bins were observed across the assessed routes. Excluded opportunities were not enumerated; this is a search boundary, not a matching conclusion. Route-specific record membership remains in each route's evidence.
Explore: Explore whether the oversized-bin populations have retained exact refinement evidence that can describe a safer bounded search.
Why this was noticed
Oversized blocking bins limited the operational search space
74 oversized route bins were observed across the assessed routes. Excluded opportunities were not enumerated; this is a search boundary, not a matching conclusion. Route-specific record membership remains in each route's evidence.
Groups of records that stood out during the examination. They may reveal structure, repeated patterns or useful distinctions in the data. A record may belong to more than one population.
Population scope: All records · 9,309 records observed · all profiled
Scope evidence
Record population represented by the configured source scope after input parsing; retained profile evidence may describe a sample of this population.
Observed
Retained population scope
Profile scope: Full · 9,309 rows observed in profile
Population quantity 27 records · 0.3% of 8,209 usable Amount values examined
Exact over the full observed source population
Recurrence eligibility excludes absent, malformed, null, blank and null-like field values.
Record predicate · Exact observed count
amount · value_equals
£24,104.13 occurs in 27 records. Amount is 90.1% distinct overall; the value occurs about 24.3x the field-average multiplicity implied by that distinctness.
Records where Amount equals the retained comparison value under canonical_category_v0; recurrence is observational evidence and does not establish duplicate records.
100% of 27 population members with usable Expense type evidence were employee education; 1.0% of 8,209 records with usable Expense type evidence globally overall (105.24× lift)
100% of 27 population members with usable Expense area evidence were hr; 1.6% of 8,209 records with usable Expense area evidence globally overall (64.13× lift)
Population quantity 45 records · 4.1% of 1,100 valid Amount values examined
Exact over the full observed source population
Structure eligibility requires a present, valid, non-null, non-blank field value. EMPTY means no retained core character family, not physical missingness.
Record predicate · Exact observed count
amount · structure_equals · core · D{9}
Values in this population contain 9 digits.
Records where Amount has core structure signature D{9}; explanation dimensions are evidence about this population, not membership conditions.
Population quantity 4 records · 0.4% of 1,100 usable Amount values examined
Exact over the full observed source population
Recurrence eligibility excludes absent, malformed, null, blank and null-like field values.
Record predicate · Exact observed count
amount · value_equals
28652.40 occurs in 4 records. Amount is 98.9% distinct overall; the value occurs about 4.0x the field-average multiplicity implied by that distinctness.
Records where Amount equals the retained comparison value under canonical_category_v0; recurrence is observational evidence and does not establish duplicate records.
Population quantity 361 records · 32.8% of 1,100 valid Amount values examined
Exact over the full observed source population
Structure eligibility requires a present, valid, non-null, non-blank field value. EMPTY means no retained core character family, not physical missingness.
Record predicate · Exact observed count
amount · structure_equals · core · D{8}
Values in this population contain 8 digits.
Records where Amount has core structure signature D{8}; explanation dimensions are evidence about this population, not membership conditions.
Population quantity 20 records · 1.8% of 1,100 usable Amount values examined
Exact over the full observed source population
Recurrence-cohort eligibility excludes absent, malformed, null, blank and null-like field values. Membership uses the complete bounded exact-confirmed recurrent-value set.
Record predicate · Exact observed count
amount · value_in_set
Amount is 98.9% distinct overall. Exact confirmation found 8 recurrent values involving 20 records. This population contains records whose Amount value belongs to that exact recurrent-value set.
Records whose canonical Amount value is one of the 8 exact-confirmed recurrent values in an otherwise near-unique field. Recurrence is an observed field-value property and does not establish duplicate records, shared entities or cause.
Population quantity 3 records · 0.3% of 1,100 usable Amount values examined
Exact over the full observed source population
Recurrence eligibility excludes absent, malformed, null, blank and null-like field values.
Record predicate · Exact observed count
amount · value_equals
43792.19 occurs in 3 records. Amount is 98.9% distinct overall; the value occurs about 3.0x the field-average multiplicity implied by that distinctness.
Records where Amount equals the retained comparison value under canonical_category_v0; recurrence is observational evidence and does not establish duplicate records.
LDLD: Letters → digits → letters → digits L = letters · D = digits · O = other core characters. Codes describe character-family order; exact lengths are available in Evidence.
49 records · 0.6% of 8,209 eligible
Context found
Expense type = business rates40 / 49 records · 81.6%0.8% overall · 97.12× lift
Expense area = estates46 / 49 records · 93.9%9.1% overall · 10.26× lift
Examples (5)
Nottingham Regional Centre. Business rates for period 01/04/26 to 31/03/27· File 1 · record 836
Croydon Regional Centre - Business Rates for the period 01/04/2026 to 31/03/2027· File 1 · record 770
Waterview Park, Newcastle. Business rates for period 01/04/26 to 31/03/27· File 1 · record 831
Queensway House, East Kilbride. Business rates for period 01/04/26 to 31/03/27· File 1 · record 830
Manchester Regional Centre. Business Rates for the period 01/04/2026 to 31/03/2027· File 1 · record 817
Population quantity 49 records · 0.6% of 8,209 valid Description values examined
Exact over the full observed source population
Structure eligibility requires a present, valid, non-null, non-blank field value. EMPTY means no retained core character family, not physical missingness.
Record predicate · Exact observed count
description · structure_equals · family · LDLD
Values in this population contain letters followed by digits followed by letters followed by digits.
Records where Description has family structure signature LDLD; explanation dimensions are evidence about this population, not membership conditions.
81.6% of 49 population members with usable Expense type evidence were business rates; 0.8% of 8,209 records with usable Expense type evidence globally overall (97.12× lift)
93.9% of 49 population members with usable Expense area evidence were estates; 9.1% of 8,209 records with usable Expense area evidence globally overall (10.26× lift)
93.9% of 49 population members with usable Date evidence were 2026-04; 20.6% of 8,209 records with usable Date evidence globally overall (4.56× lift)
LD: Letters → digits L = letters · D = digits · O = other core characters. Codes describe character-family order; exact lengths are available in Evidence.
Population quantity 28 records · 0.3% of 8,209 valid Supplier values examined
Exact over the full observed source population
Structure eligibility requires a present, valid, non-null, non-blank field value. EMPTY means no retained core character family, not physical missingness.
Record predicate · Exact observed count
supplier · structure_equals · family · LD
Values in this population contain letters followed by digits.
Records where Supplier has family structure signature LD; explanation dimensions are evidence about this population, not membership conditions.
82.1% of 28 population members with usable Description evidence were advertising campaign services; 2.6% of 8,209 records with usable Description evidence globally overall (32.11× lift)
82.1% of 28 population members with usable Expense area evidence were hmrc communications; 2.7% of 8,209 records with usable Expense area evidence globally overall (30.79× lift)
82.1% of 28 population members with usable Date evidence were 2026-07; 17.0% of 8,209 records with usable Date evidence globally overall (4.82× lift)
LDL: Letters → digits → letters L = letters · D = digits · O = other core characters. Codes describe character-family order; exact lengths are available in Evidence.
Population quantity 45 records · 0.5% of 8,209 valid Supplier values examined
Exact over the full observed source population
Structure eligibility requires a present, valid, non-null, non-blank field value. EMPTY means no retained core character family, not physical missingness.
Record predicate · Exact observed count
supplier · structure_equals · family · LDL
Values in this population contain letters with an embedded digit sequence.
Records where Supplier has family structure signature LDL; explanation dimensions are evidence about this population, not membership conditions.
LDLDL: Letters → digits → letters → digits → letters L = letters · D = digits · O = other core characters. Codes describe character-family order; exact lengths are available in Evidence.
Structure eligibility requires a present, valid, non-null, non-blank field value. EMPTY means no retained core character family, not physical missingness.
Record predicate · Exact observed count
supplierpostcode · structure_equals · family · LDLDL
Values in this population contain letters followed by digits followed by letters followed by digits followed by letters.
Records where Supplier Postcode has family structure signature LDLDL; explanation dimensions are evidence about this population, not membership conditions.
LDLDL: Letters → digits → letters → digits → letters L = letters · D = digits · O = other core characters. Codes describe character-family order; exact lengths are available in Evidence.
Structure eligibility requires a present, valid, non-null, non-blank field value. EMPTY means no retained core character family, not physical missingness.
Record predicate · Exact observed count
supplierpostcode · structure_equals · family · LDLDL
Values in this population contain letters followed by digits followed by letters followed by digits followed by letters.
Records where Supplier Postcode has family structure signature LDLDL; explanation dimensions are evidence about this population, not membership conditions.
Population quantity 18 records · 0.2% of 8,209 usable Transaction number values examined
Exact over the full observed source population
Recurrence eligibility excludes absent, malformed, null, blank and null-like field values.
Record predicate · Exact observed count
transactionnumber · value_equals
5100070190 occurs in 18 records. Transaction number is 87.2% distinct overall; the value occurs about 15.7x the field-average multiplicity implied by that distinctness.
Records where Transaction number equals the retained comparison value under canonical_category_v0; recurrence is observational evidence and does not establish duplicate records.
Population quantity 6 records · 0.5% of 1,100 usable Transaction number values examined
Exact over the full observed source population
Recurrence eligibility excludes absent, malformed, null, blank and null-like field values.
Record predicate · Exact observed count
transactionnumber · value_equals
5100006717 occurs in 6 records. Transaction number is 91.7% distinct overall; the value occurs about 5.5x the field-average multiplicity implied by that distinctness.
Records where Transaction number equals the retained comparison value under canonical_category_v0; recurrence is observational evidence and does not establish duplicate records.
Population quantity 7 records · 0.6% of 1,100 usable Transaction number values examined
Exact over the full observed source population
Recurrence eligibility excludes absent, malformed, null, blank and null-like field values.
Record predicate · Exact observed count
transactionnumber · value_equals
5100008310 occurs in 7 records. Transaction number is 91.7% distinct overall; the value occurs about 6.4x the field-average multiplicity implied by that distinctness.
Records where Transaction number equals the retained comparison value under canonical_category_v0; recurrence is observational evidence and does not establish duplicate records.
Population quantity 18 records · 0.2% of 8,209 usable Transaction number values examined
Exact over the full observed source population
Recurrence eligibility excludes absent, malformed, null, blank and null-like field values.
Record predicate · Exact observed count
transactionnumber · value_equals
5100012634 occurs in 18 records. Transaction number is 87.2% distinct overall; the value occurs about 15.7x the field-average multiplicity implied by that distinctness.
Records where Transaction number equals the retained comparison value under canonical_category_v0; recurrence is observational evidence and does not establish duplicate records.
Shows retained exact value distributions for fields that met the current bounded eligibility rules for useful categorical examination.
Why these fields?
Eligibility considers bounded cardinality, repeated values, and retained field characteristics and roles. The report does not retain an individual exclusion reason for every field, so fields that are absent should not be read as sharing one reason.
Description
Exact counts over the full observed source population; shares of 9,309 eligible values.
monthly electricity for hmrc buildings across uk269 · 2.9%
crossrail brs (polar) payments253 · 2.7%
advertising campaign services247 · 2.7%
project development228 · 2.4%
physical hosting and infrastructure204 · 2.2%
Other retained values1,736 · 18.6%
24 additional retained values
project mandays supp135 · 1.5%
virtual hosting and infrastructure120 · 1.3%
software licence support - off the shelf107 · 1.1%
it contractors106 · 1.1%
software licence support98 · 1.1%
software licences - off the shelf & not integral to it infrastructure95 · 1.0%
project support94 · 1.0%
desktop services90 · 1.0%
contractor service85 · 0.9%
employee education76 · 0.8%
utility payments - electricity60 · 0.6%
project - contract mandays57 · 0.6%
system maintenance57 · 0.6%
property rent54 · 0.6%
professional services - property management51 · 0.5%
rent (irrecoverable vat)43 · 0.5%
costs awarded against department41 · 0.4%
counsel fees41 · 0.4%
non-it contingent labour40 · 0.4%
consultancy - technical38 · 0.4%
contracted services38 · 0.4%
contractor serv rn28 · 0.3%
grant funding26 · 0.3%
insolvency services (vat recoverable).26 · 0.3%
Expense area
Exact counts over the full observed source population; shares of 9,309 eligible values.
71 distinct · 0.8% · exact · 100% populated
cdio - core2,895 · 31.1%
change delivery group1,535 · 16.5%
estates823 · 8.8%
customs change delivery498 · 5.3%
business rates for diplomatic missions447 · 4.8%
b&t customs change delivery350 · 3.8%
hmrc communications259 · 2.8%
change delivery making tax digital229 · 2.5%
63 additional retained values
ccg transformation programmes214 · 2.3%
pdc customer strategy & tax design213 · 2.3%
hmrc legal group156 · 1.7%
operational excellence & output mgmt153 · 1.6%
unity business services150 · 1.6%
risk & intelligence145 · 1.6%
polar ame expenditure144 · 1.5%
hr136 · 1.5%
hmrc payments135 · 1.5%
customs services & operations88 · 0.9%
fraud investigation service82 · 0.9%
debt management51 · 0.5%
cdio transformation programmes48 · 0.5%
knowledge, analysis & intelligence44 · 0.5%
ccg transformation43 · 0.5%
csg transformation37 · 0.4%
chief executive officer36 · 0.4%
pt operations30 · 0.3%
csg strategy and change29 · 0.3%
locations programme29 · 0.3%
compliance operations directorate27 · 0.3%
commercial directorate22 · 0.2%
intermediaries22 · 0.2%
government banking19 · 0.2%
cs director general16 · 0.2%
hmrc15 · 0.2%
large business service15 · 0.2%
ubs15 · 0.2%
chief valuer group14 · 0.2%
customs policy and strategy14 · 0.2%
psg portfolio management team13 · 0.1%
etg portfolio management11 · 0.1%
psg transformation unit11 · 0.1%
individuals policy8 · 0.1%
internal audit7 · 0.1%
wealthy & mid-sized business compliance7 · 0.1%
ccg central services6 · 0.1%
business, assets & international policy5 · 0.1%
strategic policy into delivery5 · 0.1%
finance planning & performance4 · <0.1%
hr cx4 · <0.1%
income tax administration4 · <0.1%
strategic investment and delivery4 · <0.1%
strategy and transformation group4 · <0.1%
surge & odp4 · <0.1%
cdg enterprise design3 · <0.1%
etg customer process design3 · <0.1%
etg enterprise design3 · <0.1%
etg insight, ai & innovation3 · <0.1%
hmrc strategies3 · <0.1%
chief operating officer group2 · <0.1%
indirect tax directorate2 · <0.1%
risk, control and financial accounting2 · <0.1%
small business & individuals2 · <0.1%
tax administration2 · <0.1%
voa cc's not used for over 7 years2 · <0.1%
b&t dg, finance, planning & performance1 · <0.1%
business & specialist taxes1 · <0.1%
counter avoidance directorate1 · <0.1%
etg digital design1 · <0.1%
finance data & analysis1 · <0.1%
hmrc security1 · <0.1%
individuals & small business compliance1 · <0.1%
Expense type
Exact counts over the full observed source population; shares of 9,309 eligible values.
200 distinct · 2.1% · exact · 100% populated
project mandays supp4,452 · 47.8%
non domestic rates expenditure (polar)336 · 3.6%
utility payments - electricity329 · 3.5%
project mandays hmrc320 · 3.4%
crossrail brs (polar) payments253 · 2.7%
project development223 · 2.4%
physical hosting and infrastructure164 · 1.8%
sw lic sup off shlf118 · 1.3%
188 additional retained values
license or reg fee112 · 1.2%
buying agent's fees108 · 1.2%
contin labor build106 · 1.1%
rent (irrec vat)105 · 1.1%
sw lic sup mngd serv103 · 1.1%
project support101 · 1.1%
soft licen off shelf100 · 1.1%
desktop services90 · 1.0%
virtual hosting and infrastructure84 · 0.9%
employee education83 · 0.9%
business rates79 · 0.8%
cons ad imp (ir vat)61 · 0.7%
prop. man. (ir vat)59 · 0.6%
system maintenance57 · 0.6%
consultancy – it50 · 0.5%
lifecycle replace49 · 0.5%
cross government ser46 · 0.5%
contracted services43 · 0.5%
it software licenses and support42 · 0.5%
research (vat rec)42 · 0.5%
tribunal appellant costs42 · 0.5%
counsel fee vat irre41 · 0.4%
non-it contingent labour40 · 0.4%
consultancy - techni38 · 0.4%
contractor serv rn38 · 0.4%
ad campaign ser.37 · 0.4%
soft licen integral37 · 0.4%
virtual hosting36 · 0.4%
physical hosting35 · 0.4%
property management34 · 0.4%
contact platform ser31 · 0.3%
legal solicitors fee31 · 0.3%
contingnt labr-other30 · 0.3%
sw lic sup bespkd30 · 0.3%
utility payments - gas30 · 0.3%
govdept serv(vat ex)28 · 0.3%
grant in aid funding26 · 0.3%
insolvency services26 · 0.3%
govt dept services25 · 0.3%
eu supply - irr vat24 · 0.3%
subscriptions to pub24 · 0.3%
conferences23 · 0.2%
national postal evat23 · 0.2%
recruitment services23 · 0.2%
block charges22 · 0.2%
file archive storage21 · 0.2%
soft licen off bespo21 · 0.2%
furniture & fixtures20 · 0.2%
it consultancy20 · 0.2%
debt coll serv (del)19 · 0.2%
desktop managed serv19 · 0.2%
procurement contract19 · 0.2%
disposal seized gds18 · 0.2%
maintenance fees18 · 0.2%
ohs18 · 0.2%
prof membership16 · 0.2%
cycle purchase for staff - cycle to work scheme15 · 0.2%
facility rental15 · 0.2%
hardware maintenance15 · 0.2%
office fit irrec vat15 · 0.2%
govt commercial org14 · 0.2%
nat pst del ser(vat)14 · 0.2%
national postal zvat14 · 0.2%
projects14 · 0.2%
telephony services14 · 0.2%
automobiles or cars13 · 0.1%
accommodation security services12 · 0.1%
physical data transf12 · 0.1%
training and tuition11 · 0.1%
gco costs10 · 0.1%
office fit rec vat10 · 0.1%
property mgmt. ser.10 · 0.1%
sys software maint10 · 0.1%
utility payments - water10 · 0.1%
wan maint or supp10 · 0.1%
computer accessories9 · 0.1%
consultancy - organi9 · 0.1%
gbs operational bank9 · 0.1%
internat. courier9 · 0.1%
minor wrks irrec vat9 · 0.1%
building mngment8 · 0.1%
courier services8 · 0.1%
cs government recrut8 · 0.1%
electricity8 · 0.1%
fast strm vat exmpt8 · 0.1%
tax research8 · 0.1%
business continuity & disaster recovery7 · 0.1%
forensic services7 · 0.1%
govt auditing serv7 · 0.1%
mail/pickup/del serv7 · 0.1%
management of agreement & admin costs7 · 0.1%
minor wrks rec vat7 · 0.1%
mobile call charges7 · 0.1%
mobile tel charges7 · 0.1%
oth pfi unitary chrg7 · 0.1%
phones - ll rental7 · 0.1%
scanning services7 · 0.1%
telephones-landline7 · 0.1%
consultancy - strtgy6 · 0.1%
other it hardware6 · 0.1%
projects rec vat6 · 0.1%
training/tuition/edu6 · 0.1%
cs learning5 · 0.1%
data purchases5 · 0.1%
desktop computers5 · 0.1%
government procur.5 · 0.1%
legal services5 · 0.1%
property services5 · 0.1%
staff transfer - rep5 · 0.1%
consultancy - p&c4 · <0.1%
fis spec. it equip4 · <0.1%
it security vetting4 · <0.1%
vehicle servicing ma4 · <0.1%
voa4 · <0.1%
construction fit out3 · <0.1%
contin labor run3 · <0.1%
contingent labr-ops3 · <0.1%
fis spec req - res3 · <0.1%
id authorisation payments3 · <0.1%
legal ser. (zr vat)3 · <0.1%
other miscellaneous expenditure3 · <0.1%
printing3 · <0.1%
t&s other3 · <0.1%
utility - elec. zvat3 · <0.1%
utility - gas zvat3 · <0.1%
water3 · <0.1%
bids (polar) payments2 · <0.1%
bus int consultancy2 · <0.1%
chairs2 · <0.1%
computer hardware2 · <0.1%
consultancy - legal2 · <0.1%
consultancy non it2 · <0.1%
electronic newspaper2 · <0.1%
equipment & machinery2 · <0.1%
fis pwe equipment2 · <0.1%
govt inf services2 · <0.1%
insurance irrec vat2 · <0.1%
it delivery charges2 · <0.1%
lifecycle replacement2 · <0.1%
mobile phones2 · <0.1%
ogd costs2 · <0.1%
project admin/plan2 · <0.1%
publications2 · <0.1%
recruitment advert.2 · <0.1%
translation services2 · <0.1%
camera lens adapters1 · <0.1%
chartered bus ser.1 · <0.1%
computer servers1 · <0.1%
consultancy - learni1 · <0.1%
consultancy - market1 · <0.1%
consultancy not vat1 · <0.1%
county court costs1 · <0.1%
cs org dev & des.1 · <0.1%
desks1 · <0.1%
dilapidations1 · <0.1%
dilipidations1 · <0.1%
e2 other ex-gratia special payments1 · <0.1%
equipment & machiner1 · <0.1%
events management1 · <0.1%
facilities mngmnt1 · <0.1%
gov bank serv charge1 · <0.1%
health assesment1 · <0.1%
hmrc shared(rec-vat)1 · <0.1%
hotels1 · <0.1%
internet services1 · <0.1%
inwd secondmnt costs1 · <0.1%
it software & supply1 · <0.1%
lcd monitors1 · <0.1%
legal expert witness1 · <0.1%
library services1 · <0.1%
lockers1 · <0.1%
media monitoring1 · <0.1%
mens shirts1 · <0.1%
mens trousers1 · <0.1%
non-eu supply - irr1 · <0.1%
non-eu supply - rec1 · <0.1%
occupational health1 · <0.1%
professional fees and subscriptions1 · <0.1%
research conslt serv1 · <0.1%
safety/security syst1 · <0.1%
tablet computers1 · <0.1%
telephony accs.1 · <0.1%
training costs1 · <0.1%
vehicle body repair1 · <0.1%
vehicle repairs1 · <0.1%
video prod services1 · <0.1%
website development1 · <0.1%
wide area network1 · <0.1%
Supplier
Exact counts over the full observed source population; shares of 9,309 eligible values.
437 distinct · 4.7% · exact · 100% populated
capgemini1,598 · 17.2%
accenture (uk) ltd878 · 9.4%
cognizant technology solutions594 · 6.4%
fujitsu services ltd549 · 5.9%
equal experts uk ltd487 · 5.2%
inspired energy plc369 · 4.0%
city of westminster - polar338 · 3.6%
eviden technology services limited289 · 3.1%
Other retained values2,033 · 21.8%
24 additional retained values
ibm united kingdom ltd199 · 2.1%
cgi it uk ltd191 · 2.1%
manning gottlieb omd, a division of190 · 2.0%
tecknuovo ltd178 · 1.9%
alexander mann solutions ltd152 · 1.6%
royal borough of k & c - polar122 · 1.3%
netcompany uk limited108 · 1.2%
coforge98 · 1.1%
pegasystems ltd98 · 1.1%
credera ltd95 · 1.0%
csl - kpmg llp70 · 0.8%
restore digital ltd69 · 0.7%
workman holding account66 · 0.7%
cabinet office65 · 0.7%
arvato limited61 · 0.7%
deloitte llp (l)60 · 0.6%
royal mail group ltd59 · 0.6%
bytes software services ltd58 · 0.6%
british telecommunications plc51 · 0.5%
computacenter (uk) ltd48 · 0.5%
mitie fm limited46 · 0.5%
london borough of camden-polar33 · 0.4%
kinto uk limited31 · 0.3%
city of london-nd rates-polar26 · 0.3%
Temporal terrain
Date
Exact counts over the full observed source population; bucket shares of 9,309 parseable values.
2026-01-02 to 2026-07-31
9,309 parseable of 9,309 · 100% coverage · month buckets
View 7 time buckets
Start
Count
Share
2026-01
1,022
11.0%
2026-02
1,117
12.0%
2026-03
1,583
17.0%
2026-04
1,691
18.2%
2026-05
1,100
11.8%
2026-06
1,397
15.0%
2026-07
1,399
15.0%
Field evidence
A compact posture for each retained field, using existing structure, cardinality, and role evidence.
Distinct evidence: Exact means all distinct values in the retained scope were counted. Lower bound means exact retention stopped at a configured bound; an estimate, where present, indicates likely total cardinality.
2026-01-02–2026-07-31 · 9,309 parsed eligible records among 9,309 examined records. 0 populated records were unparseable. Bounds describe parsed observations, not allowed business dates.
9,309 eligible records in this retained run-level distribution; Complete. Complete for its observed eligible population, not a permitted list of values.
cdio - core · 2,895 occurrences
change delivery group · 1,535 occurrences
estates · 823 occurrences
customs change delivery · 498 occurrences
business rates for diplomatic missions · 447 occurrences
b&t customs change delivery · 350 occurrences
hmrc communications · 259 occurrences
change delivery making tax digital · 229 occurrences
Showing 8 of 71 retained values. Full distribution remains available through the field evidence links.
9,309 eligible records in this retained run-level distribution; Complete. Complete for its observed eligible population, not a permitted list of values.
project mandays supp · 4,452 occurrences
non domestic rates expenditure (polar) · 336 occurrences
utility payments - electricity · 329 occurrences
project mandays hmrc · 320 occurrences
crossrail brs (polar) payments · 253 occurrences
project development · 223 occurrences
physical hosting and infrastructure · 164 occurrences
sw lic sup off shlf · 118 occurrences
Showing 8 of 196 retained values. Full distribution remains available through the field evidence links.
9.31k records → 11 routes produced candidates → 252.8k unique candidate pairs → 1.59k accepted relationships → 349 clusters containing 1.08k records.
Operational blocking routes
In this run, eligible bins contain 2–100 records. A pair requires two records; singletons cannot generate a candidate. Bins above a configured maximum are excluded. Eligibility describes size, not whether a route was activated.
Each fingerprint shows bucket ranges of records per bin → relative number of bins. Open a route for exact reconciliation. Route record populations can overlap; their counts are not added into a globally unique total.
Bin fingerprint: below eligibility · eligible · above safety limit. Stripes mark a bucket that crosses the safety limit.
Of 1,311 repeated bins, 1,310 were within the safety limit of 100 records per bin. The oversized bins contained 125 records.
57,235 theoretical pair opportunities comprise 49,485 safe pair opportunities and 7,750 excluded opportunities. Excluded pairs were not enumerated.
Safe pair opportunities describe the assessed route bins. Hierarchical execution activates this compound route only where its required direct-parent bins are oversized; nominations describe that activated parent scope. Opportunities outside that scope were not emitted by this route. These are overlapping route opportunities, not globally unique pairs.
Of 454 repeated bins, 442 were within the safety limit of 100 records per bin. The oversized bins contained 4,698 records.
1,671,749 theoretical pair opportunities comprise 63,804 safe pair opportunities and 1,607,945 excluded opportunities. Excluded pairs were not enumerated.
Selected token route: 123.9k nominations. Exact-bin eligibility above does not describe token-signature eligibility.
Selected token route: 99.8k nominations. Exact-bin eligibility above does not describe token-signature eligibility.
Selected token route: 123.7k nominations. Exact-bin eligibility above does not describe token-signature eligibility.
Selected token route: 43k nominations. Exact-bin eligibility above does not describe token-signature eligibility.
Combined candidate-pair funnel
Candidate nominations731.2k
Unique candidate pairs252.8k
Pairs scored252.8k
Met threshold1.59k
Relationship edges1.59k
Nominations may overlap across routes; unique candidate pairs count each pair once. Candidate pairs are not confirmed matches.
Configured threshold: 0.85.
Exact candidate reconciliation
Records represented
9,309
Candidate nominations
731,154
Unique candidate pairs
252,766
Pairs scored
252,766
Met threshold
1,591
Relationship edges
1,591
Candidate nominations count emissions from selected exact and token routes before global pair de-duplication. A pair may be nominated more than once. Unique candidate pairs count each nominated pair once. Route contribution is overlapping attribution; route totals may exceed the unique pair population.
Some assessed blocking routes encountered bins above their safety limit. Route summaries identify them.
Assessed-route safety evidence
16,102,259 theoretical route-pair opportunities across assessed routes were excluded. Route populations overlap; this is not globally unique pair space. Refinement routes may be included. Excluded opportunities were not enumerated.
Matching result
1.59k accepted relationships formed 349 clusters containing 1.08k records.
8.23k records remain outside multi-record clusters under the current matching rules.
Cluster-size distribution
Records per cluster → relative number of clusters.
151001+Exact cluster and record reconciliation
1,591 accepted relationships formed 349 clusters containing 1,081 records. Largest cluster: 66 records.
8,228 records outside multi-record clusters.
Records per cluster
Clusters
Records
1
0
0
2
243
486
3
53
159
4
20
80
5
20
100
6-10
7
44
11-20
2
23
21-50
2
70
51-100
2
119
101-200
0
0
201-500
0
0
501-1000
0
0
1001+
0
0
Clusters reflect the current matching rules, not confirmed real-world entities. Records outside multi-record clusters are not necessarily unique entities.
5 additional assessed routes
These routes were assessed but have no retained separate candidate work. Refinement and redundant status remain visible; this does not establish that a route is unimportant.
Size-eligible bins were assessed, but this route emitted no separate candidate nominations in this run. Size eligibility alone does not activate a route.
Size-eligible bins were assessed, but this route emitted no separate candidate nominations in this run. Size eligibility alone does not activate a route.
Size-eligible bins were assessed, but this route emitted no separate candidate nominations in this run. Size eligibility alone does not activate a route.
Size-eligible bins were assessed, but this route emitted no separate candidate nominations in this run. Size eligibility alone does not activate a route.
Of 530 repeated bins, 518 were within the safety limit of 100 records per bin. The oversized bins contained 4,332 records.
1,421,750 theoretical pair opportunities comprise 64,524 safe pair opportunities and 1,357,226 excluded opportunities. Excluded pairs were not enumerated.
Safe pair opportunities describe the assessed route bins. Hierarchical execution activates this compound route only where its required direct-parent bins are oversized; nominations describe that activated parent scope. Opportunities outside that scope were not emitted by this route. These are overlapping route opportunities, not globally unique pairs.
Detailed route/score evidence
Blocking groups records with usable route values into bins. A singleton bin contains one record and creates no candidate pair. Repeated bins create pair opportunities; safety-bounded routes nominate pairs for comparison.
Blocking fields are used by one or more routes to nominate candidate record pairs. A route may combine several fields; each field does not necessarily define an independent route.
A blocking route uses one or more field values together to nominate candidate record pairs. The same pair may be nominated by more than one route.
Route overlap describes candidate coverage, not field disagreement or false matches.
These populations count safety-bounded pairs nominated by a route inside a specific refinement context, not its total nominations across the entire run.
Parent-scoped route populations
Route
Parent context
Pairs
Evidence basis
Findings
Description + Expense type + Transaction number
Description + Expense type refinement
0
Exact parent-scoped count · safety-bounded pairs
—
Date + Description + Expense type + Supplier Postcode
Description + Expense type + Supplier Postcode refinement
39,362
Exact parent-scoped count · safety-bounded pairs
—
Description + Expense type + Supplier Postcode
Description + Supplier Postcode refinement
0
Exact parent-scoped count · safety-bounded pairs
—
Description + Expense type + Supplier Postcode
Expense type + Supplier Postcode refinement
0
Exact parent-scoped count · safety-bounded pairs
—
Description + Expense type + Supplier Postcode
Description + Expense type refinement
0
Exact parent-scoped count · safety-bounded pairs
—
Comparison fields
Comparison fields are fields whose retained comparison evidence is eligible to contribute to matching after candidate pairs have been nominated. Blocking narrows the pair search space; comparison evidence helps determine how strongly those pairs agree. A field may participate in both.
Per-field comparison method is not separately reported in this examination output.
What structurally or semantically distinguishes the retained population.
Population / field
Character
Technical basis
Findings
Transaction number · recurrent value Transaction number
5100006717 occurs in 6 records. Transaction number is 91.7% distinct overall; the value occurs about 5.5x the field-average multiplicity implied by that distinctness.
£24,104.13 occurs in 27 records. Amount is 90.1% distinct overall; the value occurs about 24.3x the field-average multiplicity implied by that distinctness.
Values in this population contain letters followed by digits.
Structure · family LD
—
Transaction number · recurrent value Transaction number
5100008310 occurs in 7 records. Transaction number is 91.7% distinct overall; the value occurs about 6.4x the field-average multiplicity implied by that distinctness.
Values in this population contain letters followed by digits followed by letters followed by digits.
Structure · family LDLD
—
Amount · recurrent value Amount
28652.40 occurs in 4 records. Amount is 98.9% distinct overall; the value occurs about 4.0x the field-average multiplicity implied by that distinctness.
Transaction number · recurrent value Transaction number
5100012634 occurs in 18 records. Transaction number is 87.2% distinct overall; the value occurs about 15.7x the field-average multiplicity implied by that distinctness.
Transaction number · recurrent value Transaction number
5100070190 occurs in 18 records. Transaction number is 87.2% distinct overall; the value occurs about 15.7x the field-average multiplicity implied by that distinctness.
43792.19 occurs in 3 records. Amount is 98.9% distinct overall; the value occurs about 3.0x the field-average multiplicity implied by that distinctness.
Values in this population contain letters with an embedded digit sequence.
Structure · family LDL
—
Amount · recurrence cohort (8 values) Amount
Amount is 98.9% distinct overall. Exact confirmation found 8 recurrent values involving 20 records. This population contains records whose Amount value belongs to that exact recurrent-value set.
lowest_xxhash64_occurrence_priority_v1 · maximum 5
Existing second pass Full observed structural population state
—
Record-population explanations
Whether the population is disproportionately associated with another observed field or source dimension. Lift is the population share divided by the corresponding global share.
Population / dimension
Population eligible
Leading named value
Population share
Global share
Lift
Evidence scope
Findings
Amount · recurrence cohort (8 values) Date · Temporal
20 / 20 usable
2026-05-05
4 / 20 · 20.0%
80 / 1,100 · 7.3%
2.75×
Complete
—
Amount · D{9} structure Supplier · Categorical
45 / 45 usable
capgemini
6 / 45 · 13.3%
243 / 1,100 · 22.1%
0.60×
Complete
—
Amount · D{9} structure Date · Temporal
45 / 45 usable
2026-05-06
5 / 45 · 11.1%
81 / 1,100 · 7.4%
1.51×
Complete
—
Description · LDLD structure Expense area · Categorical
Retained high-frequency values 82.1% named · 17.9% Other; not complete
—
Amount · recurrent value Date · Temporal
4 / 4 usable
2026-05-14
4 / 4 · 100%
59 / 1,100 · 5.4%
18.64×
Complete
—
Amount · D{8} structure Description · Categorical
361 / 361 usable
project - contract mandays (supplier controlled)
156 / 361 · 43.2%
543 / 1,100 · 49.4%
0.88×
Complete
—
Amount · recurrent value Description · Categorical
4 / 4 usable
consultancy - includes implementation (irrecoverable vat)
4 / 4 · 100%
6 / 1,100 · 0.5%
183.33×
Complete
—
Amount · recurrent value Description · Categorical
3 / 3 usable
project - contract mandays (supplier controlled)
3 / 3 · 100%
543 / 1,100 · 49.4%
2.03×
Complete
—
Transaction number · recurrent value Supplier · Categorical
18 / 18 usable
accenture (uk) ltd
18 / 18 · 100%
791 / 8,209 · 9.6%
10.38×
Retained high-frequency values 100% named · 0% Other; not complete
—
Pair-population evidence
How an additional field behaves across the retained record pairs in this population.
Population / field
Base pairs
Exact agreement
Usable but different
Unavailable
Findings
Date + Description + Expense type + Supplier Postcode Transaction number
39,362
999 2.5%
38,363 97.5%
0 0%
—
Date + Description + Expense type + Supplier Postcode Expense area
39,362
25,050 63.6%
14,312 36.4%
0 0%
—
Date + Description + Expense type + Supplier Postcode Supplier
39,362
39,350 ≈100%
12 <0.1%
0 0%
—
Date + Description + Expense type + Supplier Postcode Amount
39,362
83 0.2%
39,279 99.8%
0 0%
—
Field role evidence
Why Preflight thinks a field may play this role, based on retained profile and value evidence. Likely roles remain hypotheses. Role inference uses bounded semantic compatibility checks; it is not a PII assessment.
Amount: Unresolved
No role was promoted from the retained evidence.
Full observed source population. Rule: field_role_semantic_bundle.v1
@ rate 0%; digit ratio 83.4%; alpha ratio 0%; average length 8.954345257277904; temporal rate 0%.
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
£170,000.00 File 2; record 473; Representative distinct sample
£96,887.78 File 6; record 698; Representative distinct sample
506070 File 3; record 58; Representative distinct sample
23825.9 File 5; record 254; Representative distinct sample
61818.75 File 4; record 251; Representative distinct sample
High-frequency candidates
Frequently observed retained values used as supporting context. Counts may be lower bounds where retention is bounded.
Profile scope above applies. These are canonical category spellings, not retained original raw values.
£24,104.13 — at least 27 retained occurrences
400.00 — at least 12 retained occurrences
24104.13 — at least 10 retained occurrences
£100,000.00 — at least 5 retained occurrences
271950.00 — at least 4 retained occurrences
Existing structural examples
Supplemental minority evidence; these examples do not describe the whole field and do not veto the profile-time role.
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
01/05/2026 — Temporal-compatible File 7; record 1; Representative distinct sample
12/03/2026 — Temporal-compatible File 6; record 753; Representative distinct sample
13/02/2026 — Temporal-compatible File 2; record 552; Representative distinct sample
26/06/2026 — Temporal-compatible File 5; record 17; Representative distinct sample
26/03/2026 — Temporal-compatible File 6; record 1,433; Representative distinct sample
High-frequency candidates
Frequently observed retained values used as supporting context. Counts may be lower bounds where retention is bounded.
Profile scope above applies. These are canonical category spellings, not retained original raw values.
22/04/2026 — at least 241 retained occurrences; Temporal-compatible
11/03/2026 — at least 226 retained occurrences; Temporal-compatible
21/04/2026 — at least 224 retained occurrences; Temporal-compatible
20/04/2026 — at least 198 retained occurrences; Temporal-compatible
11/06/2026 — at least 197 retained occurrences; Temporal-compatible
Department family: Unresolved
No role was promoted from the retained evidence.
Full observed source population. Rule: field_role_semantic_bundle.v1
@ rate 0%; digit ratio 0%; alpha ratio 100%; average length 4; temporal rate 0%.
Contradictions / guards: Very low-cardinality field
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
HMRC File 1; record 1; Representative distinct sample
High-frequency candidates
Frequently observed retained values used as supporting context. Counts may be lower bounds where retention is bounded.
Profile scope above applies. These are canonical category spellings, not retained original raw values.
hmrc — at least 9,309 retained occurrences
Description: Unresolved
No role was promoted from the retained evidence.
Full observed source population. Rule: field_role_semantic_bundle.v1
@ rate 0%; digit ratio 0.1%; alpha ratio 86.1%; average length 38.69148136212268; temporal rate 0%.
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
Leeds, 7 & 8 Wellington Place - Door Closer upgrade / replacement works. File 1; record 740; Representative distinct sample
Computer servers File 3; record 465; Representative distinct sample
Subscriptions to Publications (Irrecoverable VAT) File 1; record 317; Representative distinct sample
Fast Streamer Recharge File 6; record 1,536; Representative distinct sample
Pre-employment checks File 3; record 1,018; Representative distinct sample
High-frequency candidates
Frequently observed retained values used as supporting context. Counts may be lower bounds where retention is bounded.
Profile scope above applies. These are canonical category spellings, not retained original raw values.
project - contract mandays (supplier controlled) — at least 4,133 retained occurrences
non domestic rates expenditure (polar) — at least 335 retained occurrences
project - contract mandays (hmrc controlled) — at least 297 retained occurrences
monthly electricity for hmrc buildings across uk — at least 269 retained occurrences
crossrail brs (polar) payments — at least 252 retained occurrences
Existing structural examples
Supplemental minority evidence; these examples do not describe the whole field and do not veto the profile-time role.
@ rate 0%; digit ratio 0%; alpha ratio 100%; average length 4; temporal rate 0%.
Contradictions / guards: Very low-cardinality field
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
HMRC File 1; record 1; Representative distinct sample
High-frequency candidates
Frequently observed retained values used as supporting context. Counts may be lower bounds where retention is bounded.
Profile scope above applies. These are canonical category spellings, not retained original raw values.
hmrc — at least 9,309 retained occurrences
Expense area: Unresolved
No role was promoted from the retained evidence.
Full observed source population. Rule: field_role_semantic_bundle.v1
@ rate 0%; digit ratio <0.1%; alpha ratio 85.6%; average length 18.495112256955636; temporal rate 0%.
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
Income Tax Administration File 1; record 591; Representative distinct sample
@ rate 0%; digit ratio <0.1%; alpha ratio 88.6%; average length 21.14663229133097; temporal rate 0%.
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
Computer servers File 3; record 465; Representative distinct sample
Fast Strm VAT Exmpt File 1; record 884; Representative distinct sample
Consultancy - Organi File 1; record 57; Representative distinct sample
Grant in Aid Funding File 1; record 546; Representative distinct sample
@ rate 0%; digit ratio 0.1%; alpha ratio 88.7%; average length 19.612740358792568; temporal rate 0%.
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
STOCKPORT MET BOROUGH COUNCIL-POLAR File 1; record 1,601; Representative distinct sample
@ rate 0%; digit ratio 29.4%; alpha ratio 57.1%; average length 7.430443656676335; temporal rate 0%.
Contradictions / guards: Specific non-contact name evidence
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
NP10 8XG File 1; record 604; Representative distinct sample
LE67 1UE File 7; record 1,085; Representative distinct sample
HR4 9SG File 1; record 559; Representative distinct sample
1074 File 2; record 803; Representative distinct sample
EC2P 2SR File 5; record 1,159; Representative distinct sample
High-frequency candidates
Frequently observed retained values used as supporting context. Counts may be lower bounds where retention is bounded.
Profile scope above applies. These are canonical category spellings, not retained original raw values.
tf3 3bd — at least 1,598 retained occurrences
ec3m 3bd — at least 900 retained occurrences
ec2m 4ag — at least 594 retained occurrences
wa3 6gd — at least 549 retained occurrences
al1 5he — at least 487 retained occurrences
Existing structural examples
Supplemental minority evidence; these examples do not describe the whole field and do not veto the profile-time role.
@ rate 0%; digit ratio 100%; alpha ratio 0%; average length 10; temporal rate 0%.
Contradictions / guards: Identifier name evidence
Inspect bounded values and provenance
Decision record: No value-driven contact role promoted; existing role rules retained.
Representative distinct values
Deterministically selected distinct examples used to inspect value shape and semantic compatibility; not a statistical sample.
bottom_5_xxhash64_raw_distinct_v1; valid non-null values up to 1024 bytes; earliest retained source/record per raw spelling. Oversized observed values omitted: 0.
5100002865 — Phone-compatible File 1; record 250; Representative distinct sample
5100058645 — Phone-compatible File 3; record 704; Representative distinct sample
5100023128 — Phone-compatible File 4; record 1,077; Representative distinct sample
1900120276 — Phone-compatible File 7; record 908; Representative distinct sample
1900075885 — Phone-compatible File 6; record 652; Representative distinct sample
High-frequency candidates
Frequently observed retained values used as supporting context. Counts may be lower bounds where retention is bounded.
Profile scope above applies. These are canonical category spellings, not retained original raw values.
5100012634 — at least 18 retained occurrences; Phone-compatible
5100067641 — at least 18 retained occurrences; Phone-compatible
5100070190 — at least 18 retained occurrences; Phone-compatible
5100075987 — at least 18 retained occurrences; Phone-compatible
5100018853 — at least 9 retained occurrences; Phone-compatible
Identity-role evidence
Profile evidence supporting fields that may identify a person, organisation or record.
Field
Role
Populated
Distinct
Scope
Transaction number
Identifier
9,309 / 9,309 · 100%
8,167 distinct · 87.7% · exact
Full
Supplier Postcode
Postal code
9,309 / 9,309 · 100%
399 distinct · 4.3% · exact
Full
Comparison-field decisions
Why each field was or was not allowed to contribute to matching.
Field
Included
Reason
Amount
Yes
accepted blocking field
Retained discrimination weight: 1.00 · STRONG
Contract Number
No
no populated comparison evidence
Date
Yes
accepted blocking field
Retained discrimination weight: 0.20 · WEAK
Department family
No
constant field
Description
Yes
accepted blocking field
Retained discrimination weight: 0.20 · WEAK
Entity
No
constant field
Expense area
Yes
accepted blocking field
Retained discrimination weight: 0.20 · WEAK
Expense type
Yes
accepted blocking field
Retained discrimination weight: 0.20 · WEAK
Project Code
No
no populated comparison evidence
Supplier
Yes
accepted blocking field
Retained discrimination weight: 0.20 · WEAK
Supplier Postcode
Yes
accepted blocking field
Retained discrimination weight: 0.20 · WEAK
Supplier Type
No
no populated comparison evidence
Transaction number
Yes
accepted blocking field
Retained discrimination weight: 1.00 · STRONG
Retained token-comparison counts
Observed token-comparison activity retained during matching; not a full pair-level score explanation.
Field
Token comparisons
Field-score increases
Threshold-crossing participation
Description
252,766
13,344
0
Expense area
252,766
78
0
Expense type
252,766
2,688
0
Supplier
252,766
3,534
1
Token evidence
Whether richer token-based comparison evidence was available for this field. Retention and eligibility do not by themselves establish that token comparison was used for a pair.
Amount — rich evidence: not collected; comparison eligible: no · Rich token evidence is unavailable.
Contract Number — rich evidence: not collected; comparison eligible: no
Date — rich evidence: not collected; comparison eligible: no · Rich token evidence is unavailable.
Department family — rich evidence: not collected; comparison eligible: no · Rich token evidence is unavailable.
Blocking portfolio: 5 behaviourally viable blocking candidates were classified as refinements and did not create separate matching work. Operational PairKey reach was preserved.
Blocking safety: 74 oversized route bins were observed across the assessed routes. Excluded opportunities were not enumerated; this is a search boundary, not a matching conclusion. Route-specific record membership remains in each route's evidence.
Edge admission: All 1,591 threshold-meeting pairs had sufficient independent comparison evidence to form relationship edges.
Categorical: Description retained categories account for 81.4% of eligible records.
Categorical: Supplier retained categories account for 78.2% of eligible records.
Temporal: Date spans 2026-01-02 to 2026-07-31.
Examination boundary
What the run did and did not examine. “Where eligible” means evidence was collected only where existing eligibility rules applied.
Analysis
Status
Scope
Detail
Input parsing
Performed
configured input sources
Parsing conditions, recovered records, excluded malformed values, and skipped records were counted during ingestion.
Bounded Email/Phone hypotheses use name priors, retained profile posture, high-frequency candidates, deterministic distinct representatives and existing whole-value recognisers. Retained structural examples provide supplemental minority evidence. Role inference remains distinct from the bounded PII/identity-bearing assessment.
First pass profiling
Performed
full observed source population
Field profiles describe the retained observed scope; sampled evidence is not extrapolated.
Structure and cardinality
Performed
full observed source population
Populated counts, distinct evidence, and observed structural traits were examined for retained field profiles.
Categorical terrain
Performed where eligible
Exact counts over the full observed source population
Categorical distributions were retained for 4 eligible fields.
Temporal terrain
Performed where eligible
Exact counts over the full observed source population
Temporal distributions were retained for 1 eligible field.
Record population explanation
Performed where eligible
Confirmed populations and explanations retain their own observed scopes; candidate discovery used full observed source population
7 bounded structural candidates were planned and 7 structural record populations were confirmed during engine-managed scans; 35 representative values were retained across 7 structural populations and 5 head/tail categorical dimensions were tested.
Material absence populations
Performed where eligible
no eligible retained material absence candidates
0 material absence populations were identified from retained first-pass evidence. 0 were selected for exact investigation within the one-per-field, four-field budget; 0 were not investigated because of the per-field limit and 0 because of the run budget. Absence analysis was bounded and was not exhaustive; selected populations used the existing indexing pass and required no additional source pass.
Operational marker detection
Performed where eligible
eligible values in the existing indexing pass
The operational_marker_matcher.v1 bounded vocabulary inspected 16 eligible fields and 32 field/family combinations during the existing indexing pass; 0 marker populations were confirmed. Detection was opportunistic and non-exhaustive; numeric magic values were not assessed.
Unexpected value recurrence
Performed where eligible
eligible values in the existing indexing pass
4 eligible fields produced 100 bounded recurrence candidates. 12 candidates were selected and exact-confirmed during the existing indexing pass. 7 recurrence populations met the promotion criteria. No additional source pass was required.
Targeted semantic checks
Performed where eligible
eligible noteworthy populations supported by retained whole-value hints
Retained evidence was checked for eligible Email, IBAN or Phone signals; no population justified deeper semantic enrichment under the current evidence-driven rules.
Blocking
Performed where eligible
operational candidate search space
Exact blocking evidence describes only eligible routes and safety-bounded operational bins.
Matching
Performed
scored operational record pairs
252,766 operational candidate pairs were scored.
Value review signals
Performed where eligible
fields inferred as Date of birth; exact full observed source scope
No eligible Date-of-Birth field was identified.
PII assessment
Performed where eligible
inferred roles and bounded retained field-semantic evidence
Bounded identity/PII signals were assessed from inferred field roles, name/profile posture and retained representative/high-frequency values. Payment-card Luhn and IBAN MOD-97 checks are whole-value structural/checksum compatibility where eligible. This is non-exhaustive, not a legal classification or comprehensive PII/DQ scan; absence of a signal does not imply absence of personal data.
Limitations
Constraints on interpretation that remain after the examination boundary.
Review signals: v1 checks only bounded rule-backed conditions currently implemented by Preflight. Absence of a review signal does not imply that all field values are plausible or suitable for their intended use.
The age threshold is a review heuristic, not a business rule or proof that a value is invalid.
Operational markers: Uses a small bounded marker vocabulary and deterministic matching rules; absence of a marker finding does not imply that no organisation-specific sentinel values exist. Numeric and date magic values were not assessed.
Recurrence: Examines a bounded set of first-pass retained repeated-value candidates in eligible fields; retained candidates are not guaranteed global top values, only selected candidates are exact-confirmed, and absence of a recurrence finding does not imply that no other repeated value exists. Repeated values do not establish duplicate records.
Targeted semantic checks: Targeted semantic checks are evidence-driven and non-exhaustive. Email checks remain limited to eligible Phone populations; Phone checks are limited to eligible Email populations; IBAN checks are triggered only by retained strong whole-value hints in noteworthy populations. Absence of a Finding does not imply that no email, phone, or IBAN-like values exist elsewhere.
Phone semantics: Phone matching is a plausibility check; it does not establish that a number exists or is in use.
IBAN semantics: An IBAN semantic match establishes structure and checksum consistency, not account existence or ownership.
Blocking safety: Theoretical route-pair opportunities across assessed routes can overlap and are not globally unique pair space. Refinement routes may be included; excluded opportunities were not enumerated.
Blocking inference: 33 generated blocking candidates fell outside the provisional cap and were not behaviourally tested.
Blocking: Blocking keys use exact normalized values; records with differing blocking values are not nominated through that route.
Categorical evidence: Description candidates came from bounded first-pass values and are not guaranteed to be the mathematical global top values.
Categorical evidence: Supplier candidates came from bounded first-pass values and are not guaranteed to be the mathematical global top values.
Likely roles are hypotheses supported by bounded observed values, not a claim about every value. Distinct representatives are non-statistical examples, not a whole-field prevalence measure. Role inference remains distinct from the bounded PII/identity-bearing assessment.
PII and identity-bearing signals are bounded, evidence-driven observations, not exhaustive or legal classifications. Absence of a signal does not prove absence of personal data. Luhn compatibility does not prove an active payment card; IBAN structure and MOD-97 do not prove account existence or ownership. Country-specific personal-identifier validation was not performed.