Merge and aliases

One physical product, one sku_id. 23 open candidates across 2,159 SKUs · 214 live aliases · scanner last ran 29-07 03:15 Dubai

SKUs 15

Style groups 5

Aliases only 3

Blocked 4

History 41

All 15

Barcode match 6

Prefix variant 4

Retired code 3

Cross brand 1

Composite pair 1

Blocked 4

Dismissed 9

Weights version v1 · a historical score always explains itself with the weights that produced it

Suspected duplicate SKUs

Ranked by composite similarity. Selecting a row loads it into the canvas below.

6 at 90+5 at 80 to 894 below 80

	Pair	Similarity	Matched signals	Units A	Units B	Ledger rows	State
*	MSR31-07 Sereen Rose / Tan · CIC vs MSR-31-7 Sereen Rose Tan · inv1 excel import	91	barcode supplier part cost	175	4	1,204	in canvas
	LBA01 Leather bracelet black · 12oz vs LBA-01 Leather bracelet blk · shopify legacy	96	barcode image hash title	318	11	2,880	open
	ALM02 Almas pendant · active vs ALM02-OLD Almas pendant · retired 03-2025	99	variant id barcode cost	96	0	744	alias only, zero balance
	FIS12-B Fisha set bracelet · component vs FIS12B Fisha bracelet · pc map import	97	barcode recipe leaf cost	44	2	508	member of 4 recipes
	GIFTBOX_MSR31_JD composite · 2 components vs GB_MSR31_JD composite · 2 components · same BOM	94	identical BOM price title	0	0	312	both composite, balance 0
	STRAP-FST01 Strap tan 20mm vs STRAP_FST01 Strap tan 20mm · underscore variant	95	barcode supplier part recipe leaf	210	7	1,096	open
	CDW01 Classic dial watch vs CDW1 Classic dial watch · ops2 typo	93	title cost category	142	1	1,822	open
	JMS01 Joory set · jewellery_set parent vs JMS-01 Joory set · csd string	92	identical BOM title price	0	0	402	open
	MHS17 Hamdan watch · product_type Watch vs MHS-17 Hamdan · grn string	90	barcode supplier part	61	3	690	open
	AQB01 Aqua bracelet · Minimalist vs AQB01 Aqua bangle · Lulwa	88	code exact different brand	54	39	1,010	do not merge · brand collision
	MSR31-11 Sereen Gold / Mesh vs MSR31-11X Sereen Gold Mesh · duplicate variant	86	title cost image hash	0	18	96	blocked · open transfer TR-0938
	PP02 Perfume Prive 50ml vs PP02-S Perfume Prive tester · SAMPLE_POOL	74	title supplier part	560	22	3,110	distinct product · sample SKU
	DQR13 Dana ring vs DQR13P Dana ring pearl	72	title category	88	64	1,244	blocked · 3 units GIT_INBOUND
	JRY12 Joory pendant vs JRY12P Joory pendant pearl	68	title supplier	120	77	1,918	distinct colourway
	TNS01 Tanya ring vs TNS01-70 Tanya ring · size 70	61	title	32	14	560	size variant, not a duplicate

9 pairs dismissed this quarter stay dismissed until their signals change Scanner runs nightly 03:15 Dubai · n8n · alerts only on a new 90+ pair

Merge pair · MSR31-07 and MSR-31-7

Similarity 91 · resolve each field, pick the survivor, dry-run, then confirm by typing the phrase

dry-run required

Primary

MSR31-07 MSR-31-7

Survives post merge · the other is frozen and becomes an alias

Field

Entity AMSR31-07

Entity BMSR-31-7

SKU code

MSR31-07

MSR-31-7

Title

Sereen · Rose / Tan strap

Sereen Rose Tan

Shopify variant

40218835291

none · never linked

Barcode

6291033987421

Supplier part

CIC-SR31-RT

Aliases preserved after the merge · 5

MSR31-07-OLDx CIC-SR31-RTx MSR-31-7x MSR 31 7x MSR31_7x

Every retired string stays searchable. Historical POs, ledger rows, RFID encodes and webhook payloads keep their original reference, and lookups resolve forward through the alias chain.

Downstream impact

Ledger rows

1,204

stay on sku_id 2604

Serials

146

already keyed on sku_id

PO lines

2 open, repointed

Open cases

CSD 1 · drift 1 · RTS 1 · repair 1

Recipes

3 active, 1 inferred

Value at stake

36K

AED · above the 100K dual gate? no

POST /api/merge/dry-run winner=2604 (MSR31-07) loser=3311 (MSR-31-7) dry_run_hash = 8f3c41d0a97b taken 29-07 14:44 Dubai valid for 15 min ok balance check loser has 4 units at 1 location x 1 stock type ok in-flight check 0 IN_TRANSIT_* 0 GIT_INBOUND 0 IN_PRODUCTION 0 open transfer lines ok cost layers loser has none, nothing to merge into the FIFO stack ok alias collision 3 new alias strings, none already live elsewhere ok recipe membership loser is a leaf in 0 active recipes !! open cases 4 cases reference the loser string, they will be repointed to 2604 compensating rows to be posted (2): ledger sku=3311 loc=HQ_325 type=STOCK qty=-4 idem=merge:MRG-2026-0144:HQ_325:STOCK:out ledger sku=2604 loc=HQ_325 type=STOCK qty=+4 idem=merge:MRG-2026-0144:HQ_325:STOCK:in post state: balance(3311, *, *) = 0 balance(2604) = 175 + 4 = 179 zero rows updated · zero rows deleted · hash chain unaffected

Dual approval not required · AED 36K below the 100K threshold

Confirm irreversible merge

Type MERGE MSR31-07 MSR-31-7 exactly. The commit is rejected if the dry-run hash 8f3c41d0a97b no longer matches the world, which happens if anything moved in the last 15 minutes. Undo is not a button: it is a new merge in the opposite direction, logged as its own transaction.

Phrase incomplete · 20 of 26 characters

Alias registry

214 live aliases. Append only: an alias is never deleted, only superseded by a later merge.

Alias	Canonical	Kind	Source	Origin	Live from
MSR31-07-OLD	MSR31-07	rename	shopify	variant sku change	14-03-2025
CIC-SR31-RT	MSR31-07	supplier part	grn	PO-2025-0598 manifest	02-11-2025
LBA-01	LBA01	import	inv1	excel_sale backfill	31-12-2024
ALM02-OLD	ALM02	merge	manual	MRG-2026-0139	11-07-2026
JMS-01	JMS01	typo	csd	case line item string	18-05-2026
GB_MSR31_JD	GIFTBOX_MSR31_JD	merge	manual	MRG-2026-0141	19-07-2026
STRAP_FST01	STRAP-FST01	import	sku_intel	parent_child_map	04-02-2026
CDW1	CDW01	typo	ops2	dispatch scan string	27-06-2026
MSK01-VCR	VCR01	legacy_excel	inv1	strapchange transform	31-12-2024
AQB01	AQB01 · minimalist	ambiguous	manual	brand collision with Lulwa	21-07-2026
FIS12B	FIS12-B	import	sku_intel	parent_child_map	04-02-2026
MHN17	unresolved	phantom	csd	rejected expansion of MHS17	17-07-2026

Depth after compaction: 1 · when B merges into A and later A into Z, every alias of B is re-inserted pointing at Z 1 phantom string quarantined, never given a canonical target

Merge history

Last 12 across all entity types. A reversal is a new forward transaction, not an edit.

Merge	When	Loser to winner	Rows	Units moved	Actor	State	Reverse
MRG-2026-0143	28-07 16:02	DQR13-OLD to DQR13	1,244	6	AY	committed
MRG-2026-0142	26-07 11:40	TNS1 to TNS01	560	2	AY	committed
MRG-2026-0141	19-07 09:18	GB_MSR31_JD to GIFTBOX_MSR31_JD	312	0	AY	committed
MRG-2026-0140	15-07 14:55	PP02-50 to PP02	3,110	18	SA	committed
MRG-2026-0139	11-07 10:21	ALM02-OLD to ALM02	744	0	AY	committed
MRG-2026-0138	08-07 17:33	LMW01-18 to LMW01	208	4	FB	reversed	by 0138R
MRG-2026-0138R	09-07 08:02	LMW01 to LMW01-18	208	4	AY	reversal
MRG-2026-0137	02-07 12:09	JFS01 to JFS01P	1,102	9	AY	committed
MRG-2026-0136	27-06 15:44	CDW-1 to CDW01	1,822	1	SA	committed
MRG-2026-0135	21-06 09:55	FIS12E to FIS12-E	488	3	AY	committed
MRG-2026-0134	14-06 13:12	ANQ04 to ANQ04P	640	11	FB	committed
MRG-2026-0133	02-06 10:47	SHOPIFY-VAR-88214 to MSR31-03	96	22	AY	committed

41 merges since phase 0 · 1 reversed · 0 ledger rows updated · 0 deleted The last row cleaned up an i2 placeholder SKU. There were 61 of them at import.

Safety rules

Irreversible

Undo is a separate forward merge in the opposite direction, with its own confirmation. There is no click-to-undo.

Never rewrites the ledger

Historical rows keep the original sku_id. Balance moves by a compensating pair. This is why the hash chain still verifies afterwards.

Blocked while stock is in flight

Any unit in IN_TRANSIT_CUSTOMER, IN_TRANSIT_RETURN, GIT_INBOUND or IN_PRODUCTION, or any open transfer line or GRN session, blocks the merge. Resolve the physical flow first.

Dual approval above AED 100K

Measured on the on-hand value at stake across both SKUs, not on supplier spend. Routes to the approvals queue.

Dry-run staleness

A commit carries the hash of its preview. If any counted object changed in the last 15 minutes the commit is rejected and a fresh preview is required.

Brand collisions are not duplicates

Minimalist and Lulwa reuse short codes. A cross-brand pair is marked ambiguous and resolved by brand at lookup time, never merged.

Detection signals · v1

Barcode / GTIN exact

0.30

Variant title trigram

0.20

Supplier part number

0.15

Cost within 2%

0.10

Image perceptual hash

0.10

Category + product_type

0.08

Co-location pattern

0.07

Fuzzy suffix guessing (append P / G / S, strip a hyphen) is allowed here only, as a scoring hint a human confirms. It is banned on every write path, because that is exactly how the SKU Intel splitter silently attributes value to a different product.

Candidate distribution

23 open pairs by similarity band

merge candidate review likely a real variant

Alias registry growth

cumulative live aliases · 12 months

Aug 25Nov 25Feb 26Jul 26

44 to 214 · the jump in March is the INV1 backfill, not a data quality problem

This month

Merges

5 committed, 1 reversed

Aliases added

14 from merges

Units moved

by compensating pairs

Chain breaks

verified nightly 02:00

The scanner runs nightly at 03:15 Dubai from n8n and posts to #n8n-updates only when a new pair scores 90 or above, deduped by pair id. A silent night is the normal night.

Connected to SKU detail Cost + price trends SKUs (Live) Classic table Audit log Drift Suppliers Locations Rules Settings RFID gates Shopify events