PO ↔ invoice 3-way match

Live Finance Correlation key: po_number · 94.2%

Live Automated since Jan 2026 — 4.5 h/week reclaimed across everyone who runs it, 125 h to date. 0 h/week of hands-on work is left, and that is the figure quoted everywhere on this page.

Case footprint: median 5.4 h elapsed, 18m 22s of it hands-on, across 1 work session and 0.2 calendar days. 12 of the 12 happy-path waits exceed the 150 s idle-gap threshold, but a gap only ends a work session when it outlasts one (59 min, the fleet's mean). None here do: the case is put down and picked up inside the same sitting, while the person works other cases.

Cases: 132
Activities: 15
Variants: 31
Median throughput: 5.4 h
p95 throughput: 31.6 h
Working time / Waiting time: 18m 22s / 5.1 h
Rework: 19%
Fitness: 0.86
Precision: 0.7

No filters — showing every case

132 of 132 cases (100%)

Directly-follows map

Discovered from the event log. The editable automation design is the last tab.

showing 15 of 15 activities · 100% of activity instances covered

Geometry is computed at build time, so nothing reflows when you filter. 47 of 87 directly-follows relations sit below the 0.20 noise threshold and are hidden; together they carry 3.9% of transitions.

Node fill 0–20%20–40%40–60%60–80%80–100% Happy path Loop back Self-loop

Five discrete steps. Edge thickness is a square-root scale on frequency, so the busiest edge is 8 px and the quietest 1 px.

Where the hours go

Edges ranked by total recoverable waiting time, not mean.

9.8% working 90.2% waiting · mean 37m 59s · p95 2.2 h · 129 cases affected · 11.5% of total cycle time
5.9% working 94.1% waiting · mean 36m 56s · p95 2.1 h · 106 cases affected · 9.5% of total cycle time
4.3% working 95.7% waiting · mean 31m 45s · p95 1.8 h · 122 cases affected · 9.1% of total cycle time
10.1% working 89.9% waiting · mean 24m 18s · p95 82m 37s · 130 cases affected · 7.4% of total cycle time
11.2% working 88.8% waiting · mean 26m 07s · p95 88m 48s · 120 cases affected · 7.4% of total cycle time
10.6% working 89.4% waiting · mean 21m 57s · p95 74m 38s · 109 cases affected · 6.7% of total cycle time
9.1% working 90.9% waiting · mean 22m 04s · p95 75m 02s · 125 cases affected · 6.5% of total cycle time

Variant explorer

One row per distinct end-to-end activity sequence, ranked by case count. Row 1 is the happy path and every other row is column-aligned to it, so deviations read as vertical discontinuities.

Top 12 variants cover 84.1% of cases.

Select exactly two variants to compare.

#	Variant	Sequence	Cases	% of cases	Cumulative	Median	p95	Steps	Δ vs happy path	Conformance	Compare
1	V-01 Happy path	OBLGMPMQVACCAPFDOSRMHE	36	27.3%	27.3%	4 h	16.5 h	11	—	Conforming	Compare V-01
2	V-02	OBLG^×2MPMQVACCAPFDOSRMHE	13	9.8%	37.1%	4.8 h	39.6 h	12	+1 · 45m 04s	Conforming	Compare V-02
3	V-03	OBRQLGMPMQVACCAPSAFDOS^×2RMHE	11	8.3%	45.4%	5.3 h	34.2 h	13	+2 · 74m 20s	Deviating	Compare V-03
4	V-04	OBLGMPRQMQVAAP^×2CC^×3APFDOSRMHE	10	7.6%	53%	7.5 h	2 d	16	+5 · 3.5 h	Deviating	Compare V-04
5	V-05	OBLGMPSAMQVACCAPFDOSRMHE	8	6.1%	59.1%	4.2 h	30.4 h	12	+1 · 8m 20s	Deviating	Compare V-05
6	V-06	SWOBLGMP^×4MQVACCAPFDOSRMHE	7	5.3%	64.4%	6.2 h	40.6 h	14	+3 · 2.1 h	Deviating	Compare V-06
7	V-07	OBLGMPMQVAAPCCFDAPFDOSRMHE	6	4.5%	68.9%	4.4 h	20.2 h	12	+1 · 21m 47s	Deviating	Compare V-07
8	V-08	OBLGMPMQVACC^×3APFDOSRMHE^×4	5	3.8%	72.7%	7 h	30.9 h	15	+4 · 3 h	Deviating	Compare V-08
9	V-09	OBLGMPMQVASW^×2CCAPFDOSRMHE	5	3.8%	76.5%	5.9 h	31.9 h	13	+2 · 1.8 h	Deviating	Compare V-09
10	V-10	OBLGSWMPMQVACCAPFDRQOSRMHE	4	3%	79.5%	5.1 h	38.9 h	13	+2 · 60m 52s	Deviating	Compare V-10
11	V-11	OBLGMPMQVAMQCCVAAPFDOSRMHE	3	2.3%	81.8%	4.2 h	24.7 h	12	+1 · 10m 51s	Deviating	Compare V-11
12	V-12	OBLGMPMQVACCAPSAFDOSRMHE	3	2.3%	84.1%	4.7 h	39.1 h	10	-1 · 38m 26s	Deviating	Compare V-12
—	Other variants (7) · 5.3% of cases	Activity sequence —	7	5.3%	100%	—	—	—	—	—

18 variants (58.1% of all variants) occur exactly once, together 13.6% of cases.

Variant distribution

Rank on a log scale against cases on a log scale, with cumulative coverage on the right axis.

Cases (log scale) Cumulative coverage (%) Variant rank (log scale) 11 variants cover 80% of cases

Variant comparison

Two variants on the same alignment columns. Colour says which side; hatching repeats it, so the diff survives greyscale.

Variant A

Variant B

A · V-01

Merged activity set

B · V-02

Merged activity set

Every alignment column, coloured by which side uses it.

In both Only in A Only in B In neither

List of changes

Side by side
Side by side	A	B	Δ

Attributes associated with each branch

These are associations measured on the log, not causes. Every rule prints the number of cases it was measured on.

Supplier country `supplier_country`

131 of 132 cases · 1% missing

ES34 · 25.8%
DE27 · 20.5%
FR21 · 15.9%
NL17 · 12.9%
IT13 · 9.8%
PT11 · 8.3%
Other8 · 6.1%

Rule statements

Ranked by lift × support. Each rule is a sentence you can read aloud.

When gr_present = false → the case takes “Obtain a second approval” in 49.1% of cases (baseline 21.2%, lift 2.32×, n = 45).
49.1% confidence 21.2% baseline 2.32× lift n = 45
When line_items > 33 → the case takes “Contact the vendor” in 8.3% of cases (baseline 2.3%, lift 3.61×, n = 27).
8.3% confidence 2.3% baseline 3.61× lift n = 27
When amount_band = > €50k → the case takes “Request the missing purchase order” in 68.2% of cases (baseline 21.2%, lift 3.22×, n = 23).
68.2% confidence 21.2% baseline 3.22× lift n = 23
When has_po = false AND line_items > 33 → the case takes “Contact the vendor” in 8.3% of cases (baseline 2.3%, lift 3.61×, n = 10).
8.3% confidence 2.3% baseline 3.61× lift n = 10
When supplier_country = FR AND gr_present = false → the case takes “Obtain a second approval” in 80.4% of cases (baseline 21.2%, lift 3.79×, n = 8).
80.4% confidence 21.2% baseline 3.79× lift n = 8
When cost_centre = CC-2400 AND amount_band = > €50k → the case takes “Request the missing purchase order” in 62.9% of cases (baseline 21.2%, lift 2.97×, n = 8).
62.9% confidence 21.2% baseline 2.97× lift n = 8

Decision point

Pick an exclusive choice in the map to see its branch shares and the attributes that discriminate between them.

Decision point

Branch

Match quantity and price102 · 79.1%
Request the missing purchase order13 · 10.1%
Obtain a second approval8 · 6.2%
Check the VAT code6 · 4.7%

Pick an exclusive choice in the map to see its branch shares and the attributes that discriminate between them.

Conformance against the reference model

Reference model v3, 2 Jul 2026

Fitness0.86

Precision0.7

Fitness and precision trade off; we report both rather than a single model-quality score.

Conforming cases — 49 (37.1%) Non-conforming — 83 (62.9%)

Deviations by type

A deviation is not necessarily an error — it is behaviour the reference model does not allow.

					Breaks conformance
Activity retry	54	40.9%	4m 42s	4.2 h	—
Invalid activity	61	46.2%	1m 32s	1.6 h	Breaks conformance
Loop back	32	24.2%	2m 48s	89m 36s	—
Skip	48	36.4%	62s	49m 36s	Breaks conformance
Exclusive-choice violation	12	9.1%	75s	15m	Breaks conformance
Invalid start/end	7	5.3%	0s	0s	Breaks conformance

Conforming cases over the window

Conforming cases (%) · 16 Jun – 13 Jul 2026

Conforming cases (%)16 Jun – 13 Jul 2026

Rework and loops

Self-loops and loop-backs, with loop inflow, outflow and net gain.

Activity	Kind	Self-loops	Cases affected	Mean repeats	Loop in	Loop out	Net gain	Total waiting
File the document	Loop back	0	15	2×	15	11	+4	16.8 h
Open the supplier record	Self-loop	16	14	2.14×	16	13	+3	54m 05s
Locate the goods receipt	Self-loop	13	13	2×	13	13	0	2.5 h
Match quantity and price	Loop back	0	13	2×	13	13	0	10.4 h
Locate the goods receipt	Loop back	0	11	2×	11	8	+3	14.5 h
Match against the purchase order	Self-loop	26	10	3.6×	26	28	-2	2.2 h
Apply the cost centre	Loop back	0	10	3×	20	16	+4	25.3 h
Match quantity and price	Loop back	0	8	2×	8	7	+1	5.4 h

Rework cost €1,810–2,715 · 25 cases · 91 h Costed at the €25/h loaded rate used everywhere in this dashboard.

Cases

Showing 16 of a 40-case sample drawn from 132 cases, across the 24 variants listed on the Variants tab. The totals above are exact; this table is sampled.

Variant				Roles	Departments		Conformance
V-03	2026-07-09 11:37:04	9.4 h	13	Financial ControllerAP Specialist	financebackoffice	1	Deviating
V-04	2026-07-09 10:16:22	25.3 h	16	Financial ControllerBack-Office Clerk	finance	4	Deviating
V-04	2026-07-09 09:39:52	44m 39s	16	Back-Office ClerkAP Specialist	finance	4	Deviating
V-08	2026-07-09 09:26:15	1.8 h	15	AP Specialist	financebackoffice	5	Deviating
V-01	2026-07-08 16:37:35	3.5 h	11	Financial Controller	financebackoffice	0	Conforming
V-01	2026-07-08 15:17:33	4.2 h	11	Back-Office ClerkAP Specialist	finance	0	Conforming
V-17	2026-07-08 13:26:10	6.7 h	11	AP SpecialistFinancial Controller	finance	0	Deviating
V-01	2026-07-08 13:11:29	2.5 h	11	AP Specialist	financebackoffice	0	Conforming
V-09	2026-07-08 11:34:42	6.9 h	13	Back-Office ClerkFinancial Controller	financebackoffice	1	Deviating
V-13	2026-07-08 10:52:46	4.7 h	12	AP SpecialistFinancial Controller	financebackoffice	0	Deviating
V-01	2026-07-06 08:23:37	2.3 h	11	Back-Office ClerkAP Specialist	financebackoffice	0	Conforming
V-01	2026-07-03 12:35:44	6.3 h	11	AP SpecialistFinancial Controller	finance	0	Conforming
V-13	2026-07-03 11:35:47	1.9 h	12	Financial ControllerAP Specialist	finance	0	Deviating
V-01	2026-07-02 15:14:05	5.9 h	11	Financial ControllerBack-Office Clerk	finance	0	Conforming
V-01	2026-07-02 13:13:31	4.6 h	11	Back-Office ClerkFinancial Controller	finance	0	Conforming
V-04	2026-07-01 16:34:01	1.6 h	16	AP SpecialistBack-Office Clerk	financebackoffice	4	Deviating

16 of 40

Live 90% confidence

Automation design

The Map tab is what the event log shows. This is the automation you are designing from it.

ERP clientExcelSAP (web)

4.5h reclaimed / week across everyone who runs it

Case V-01 Conforming

—

· Process: PO ↔ invoice 3-way match · Correlation key po_number

Total duration

Working Waiting Steps · Rework

Field	Value
`po_number`	PO-2026-63249
`supplier`	Vallcarca Tecnics SL
`amount_eur`	€18,917.3
`cost_centre`	CC-2665
`vat_code`	S20
`iban`	ES•• •••• •••• •••• 1648 · masked at capture (IBAN masked at capture)

Field	Value
`po_number`	PO-2026-75778
`supplier`	Brightline Media Srl
`amount_eur`	€47,167.19
`cost_centre`	CC-4210
`vat_code`	S1
`iban`	ES•• •••• •••• •••• 8881 · masked at capture (IBAN masked at capture)

Field	Value
`po_number`	PO-2026-67519
`supplier`	Atlas Papeleria SA
`amount_eur`	€10,278.32
`cost_centre`	CC-3710
`vat_code`	S14
`iban`	ES•• •••• •••• •••• 3521 · masked at capture (IBAN masked at capture)