Data harmonization

Map every system into the one your teams report against

Your item master, your retailer feeds, your syndicated file and your planning system each describe the same product differently. Harmonization maps all of them into one schema, yours, so the data joins instead of being reconciled by hand every quarter.

Nine top shelves for one bag of chips

One product, fourteen systemsData harmonization
Top-level categories it sits under
9

The same bag of tortilla chips. Nine different top shelves, and no two systems file it the same way. Nothing joins on a category when the category is nine things.

Live catalog paths · one UPC · category field only
How deep each system files itI · III
Path depth · systems filing at that depth
IThere is no standard depth0 to 5
No path2
1 level2
2 levels2
3 levels2
4 levels2
5 levels4

Two systems carry no category for it at all. Four carry five levels. Rolling the deep ones up is a call somebody has to make, and every team makes it differently.

IIOne system has it as a grilling product1 of 14

Food, then seasonal grocery, then grilling. A tortilla chip filed next to charcoal. It is not a mistake in that system, it is a merchandising decision, and it is still wrong in yours.

Right there, wrong here
IIITwelve paths, twelve answers0 matches

Of the twelve systems that file it somewhere, not one path matches another. There is no majority to take and no shortcut to a mapping.

No consensus to borrow

Harmonization is not picking the best taxonomy. It is deciding, once, where each of these belongs in yours.

Category paths taken from the live catalog for a single UPC in tortilla chips, across the fourteen systems holding a record for it. System names are generalized to their type.

How a mapping stays trustworthy

A crosswalk is easy to build once and hard to keep honest.

Mapped into your taxonomy
not into a standard of ours

The target is the taxonomy your teams already plan and report against. Asking a business to adopt someone else's category tree is how harmonization projects turn into reorganization projects and then stop.

Ambiguity is surfaced
not quietly resolved

Some paths genuinely have two defensible homes in your tree. Those come back as a decision for you rather than a coin flip made quietly inside a model, because a mapping nobody chose is a mapping nobody will defend in a review.

Held as the sources move
taxonomies change without notice

Retailers reshuffle their category trees on their own schedule and tell nobody. The crosswalk is re-checked on a cadence, so a path that moved is caught as a change to review rather than found six weeks later as a hole in a report.

Every path this product arrives under

Thirteen systems, thirteen category paths, and the one row your teams actually report against.

UPC 851769007003 Tortilla chips 14 systems
Distributor feedShopPantrySnacks, Chips & DipsChips, Crisps, PretzelsTortilla
MarketplaceGrocery & Gourmet FoodSnacks & SweetsSnack FoodsChips & CrispsTortilla
Retailer siteGlobal Primary CategorySnacks & CandyChipsCorn & Tortilla ChipsTraditional Tortilla Chips
Retailer APIGrocerySnacksChipsTortilla Chips
Retailer APIFoodSeasonal GroceryGrilling FoodsHispanic grilling
Pharmacy retailGrocerySnacksChips & Pretzels
Retailer siteNatural & OrganicGrocerySnacks
Retailer siteGroceryChips, Snacks & Cookies
Grocery platformPantryChips & Pretzels
Label databaseChips, Pretzels & Snacks
Retailer APIFood
Retailer siteno category
Retailer APIno category
Your schema Food & BeverageSalty SnacksTortilla & Corn Chips

The target is whatever your taxonomy already is. Harmonization does not ask you to adopt a new one, it maps every incoming path into the one your teams already report against.

Thirteen of the fourteen systems holding a record for this UPC, with the category path each one files it under, taken from the live catalog. System names are generalized to their type and the target schema shown is an example. The one row in red files a tortilla chip under grilling, which is a perfectly sensible merchandising call in that system and a broken join in yours.

Harmonya Helps You Make Faster, More Confident Decisions

Mapping a system in is wasted if the values inside it are wrong. The two jobs are usually the same project.

Data cleansing

One field, 146 distinct values, and what each of the wrong ones should have said.

See the corrections

Request a Demo

Thirty minutes on your categories. How the record gets built, what it holds, and the questions your team could put to it.