How Priceory Matches Products Across Stores — Honest Methodology

Why this page exists

Price comparison only helps shoppers if the same product is compared. A site that pairs a single can with a carton, or “any Milo-ish drink” with a different pack, will show exciting savings and still be wrong.

Wrong matches destroy trust faster than missing a store. This guide explains how Priceory links products, what we refuse to link, and how you should read every compare screen.

Priceory is independent. We are not the official channel of any supermarket. Prices come from public catalogues, flyers, product pages, and approved marketplace feeds we process with quality gates.

Matching quality beats raw product count

Barcode → name+pack → brand+pack → reject bad ratios

Thousands of SKUs mean nothing if half the “deals” are pack mismatches. We optimise for:

Precision first — fewer false multi-store pairs
Readable names — not OCR garbage or test fixtures
Honest labels — real shelf vs carton vs derived vs online marketplace
Reversible hygiene — soft-disable junk instead of silent wrong prices

If you ever see a suspect pair, check pack size first, then brand, then store.

Our matching layers (simplified)

Why carton vs single should not be matched as same product — Single vs carton mismatches destroy trust — we block many

1. Barcode / GTIN (strongest)

When both merchants expose a reliable barcode, we prefer exact code identity. This is the gold standard for same-SKU confidence (common on some web catalogues).

2. Name + pack structured keys

We extract pack tokens (e.g. 1L, 5kg, 6×200ml, 40s) and normalise titles. Different packs stay different products — 1L milk is not 3×1L multipack.

3. Brand + pack keys

Brand-led keys help when titles are noisy (abbreviations, ALL CAPS catalogue names). Mydin-style abbreviations are cleaned for display; matching still looks at structure.

4. Fuzzy name similarity (last resort)

Fuzzy matching is capped and supervised. It is never the only reason a homepage “Save RMXX” card is allowed to run without pack/plausibility checks.

5. Price plausibility

Extreme ratios (e.g. one store 10× another on a “same” item) are treated as likely pack or category errors. Those groups are blocked from deal boards even if a fuzzy link exists.

What we intentionally do not force together

Situation	Why
99 carton vs single retail tin	Unit economics differ; many 99 rows stay unlinked on purpose
Private label vs national brand	House brands are not forced into same-SKU matches
GWP / “Not for sale” stubs	Soft-disabled — often absurd placeholder prices
Test / validity fixtures	Soft-disabled
OCR garbage titles	Soft-disabled when unreadable
Marketplace multi-seller clones	Within Shopee, one best offer per logical SKU (price / sold / rating)
Unreadable marketplace spam titles	Soft-disabled — promo/emoji/junk noise; UI is EN + BM for family shoppers

Related reading: 99 carton vs single · bulk-buy mistakes.

Price source labels (how to read the UI)

Not every number is a live single-unit shelf sticker:

Label (concept)	Meaning for you
Real shelf price	Best-effort live / catalogue unit shelf style price
Carton / bulk	Full multipack price — do not treat as one piece
Derived from bulk	Estimated from multipack maths — verify before buying singles
Catalogue fill	Coverage fill for matching — prefer real shelf peers when comparing
Online marketplace	Shopee-style online listing — stock and seller may change

Shopee is a merchant on Priceory (online marketplace), not a sponsored strip of random affiliate banners. Outbound Shop at {store} uses the original datafeed merchant URL when available (tracked where configured). Online may earn commission and is not counted in the green TIP physical trip on Plan this week.

What you see on the homepage

Homepage multi-store deals quality gated savings photo

Multi-store deals require 2+ merchants and positive savings after filters.
Savings badges prefer trustworthy pack-aware comparisons.
If savings look “too good,” the UI may hide the save burst and show a short note instead of inventing miracles.
Freshness windows and scrapers rotate; always treat the site as decision support, then confirm in store or on the merchant page.

Online marketplace vs physical supermarket

Physical chains (AEON, 99, Lotus’s, Jaya, Mydin, Giant, etc.) reflect catalogue / shelf-oriented pipelines.

Shopee rows come from an affiliate grocery-focused datafeed subset, then:

Grocery-ish category filters
Multi-seller dedupe (one winner per SKU key)
Spam soft-disable (emoji / promo fluff / unreadable junk titles)
Ranking that still prefers clean supermarket hits for staple queries

That keeps search usable for beras, gula, diaper, milo without drowning in marketplace noise.

How you can help yourself (shopper checklist)

Verified multi-store savings gap example Priceory

Read pack size on every card.
Prefer compare views with multiple merchants (Search 2+ stores only, default OFF).
Use unit price when packs differ (RM x.xx/100g, RM x.xx/L, RM x.xx each).
Treat outlier savings with scepticism — then verify (catalogue ≠ till).
Use Plan this week for multi-item green TIP totals, not single-item FOMO.
For policy context (not matching): Harga Terkawal · MyKasih.

Why we soft-disable instead of hard-delete

Soft-disable (is_available = false) keeps history and lets us re-enable if a row was over-filtered. It also keeps junk out of search, basket evaluate, and homepage boards without destroying audit trails.

Platform maintenance re-applies hygiene rules on a schedule so quality does not silently rot.

Limits (honest)

Matching is probabilistic when barcodes are missing.
Flyer windows expire; a cheap price may be gone by Saturday night.
Store-level stock and geo pricing can differ from national catalogue feeds.
We continuously improve parsers and gates — coverage is uneven by merchant design, not by “random favourites.”