Katana to EDGEBIC: The Data Mapping Reference

Mapping Katana data into EDGEBIC is a column-mapping exercise rather than a development project: you export products, work centers, operations, open manufacturing orders, and optionally logged labor to CSV or Excel, then drag each source column heading onto the matching EDGEBIC field once and save it as a reusable import mask. This reference lists what maps to what, which columns are mandatory, how CSV quirks and unit conversion behave, and what every row outcome means when a run finishes.

EDGEBIC by User Solutions integrates with Katana through those masks rather than an API integration, so nothing is installed in the account and no integration credential exists to maintain. If you have not read the end-to-end version yet, start with the complete Katana integration guide. This post is the lookup table you keep open while building masks.

The load order

Import in dependency order. Later files reference records that earlier files created.

Two entity types have no Katana origin and come from a spreadsheet you write: Shift (one row per shift with a start and end pair per weekday, a blank pair meaning the day is off) and PlantHoliday (closures by name and date). Together they define the calendar every scheduled date is computed against, which makes them the ten minutes with the widest effect in the whole load.

Auto-create is on by default, so a routing file naming a work center that has not been imported yet creates it on the fly. Good safety net, poor strategy: an auto-created work center carries a name and nothing else, meaning one machine, no shift assignment, and no efficiency.

Reading the file itself

Cloud exports have their own habits, and all of them are handled on the mask rather than by editing files.

If a worksheet is renamed later, the reader falls back to the first sheet rather than failing, which is forgiving but worth knowing: re-pick the file on the mask when a workbook's structure changes so the choice stays explicit.

Products to Product

Only the identifier is required. Everything else can arrive later through a narrower refresh file, because blank cells preserve existing values on an update run. Keep the identifier rendered identically across every export: matching ignores case, so BRK-200 finds brk-200, but it does not reconcile a genuine spelling difference or a stray suffix.

Resources to Workcenter

This is where scheduling accuracy is won or lost, because this file carries capacity.

Number_Of_Instances is the single highest-value column here and the one cloud MRP exports rarely carry, because a resource is usually modeled as one entry with a capacity figure. A cell of four identical machines imported as one instance produces a plan roughly four times too long, and it fails quietly: nothing errors, the dates are simply wrong. Fill it deliberately for every multi-machine center; the arithmetic is in how EDGEBIC calculates work center capacity.

Machine pools have no export column

Where machines are genuinely interchangeable, EDGEBIC models a work center group: a named pool whose membership is re-evaluated on every reschedule. A routing step bound to the group picks by strategy (earliest completion, primary first, or earliest start) and applies a per-member efficiency factor, so a slower machine is chosen only when it still finishes first. Operations already started keep their machine, because recorded work is never moved.

None of that exists as an export column. Import the members as ordinary work centers, then build the group inside EDGEBIC: a one-time configuration per pool that changes the schedule more than most imported columns do. The setup path is how to create work center groups.

Operations to BOR

Op_Flag is the field most often set wrong on a first load. True means the row is an operation running on a work center. False means the row is a component consumed at that point in the routing, which is exactly how a subassembly is attached to its parent. Getting the flag backwards turns components into work centers, and the graphical designer shows it immediately.

Why the import takes two passes

A routing cannot be written row by row, because step 10 must point at step 20 and step 20 does not exist yet. The import therefore runs in two passes. Pass one validates each row, resolves or creates the products and work centers it names, and buffers the row. Pass two groups the buffer by end product, sorts by sequence number, writes the steps, and wires the chain: 10 to 20, 20 to 30, and the terminal step to the finished product so the routing renders connected rather than floating free.

Two rules follow. All of one product's steps must travel in one file, because a run wipes and recreates that product's steps on first reference, so splitting one product across two runs means the second wipe deletes the first run's work. And re-importing is safe for orders in progress, because every scheduled job carries a frozen snapshot of the routing it was planned with.

Manufacturing orders to SalesOrder

Two mask options handle the case where an export carries only one date: one treats the job date as the due date, the other derives the job date backward from the due date. Set whichever matches your export once, on the mask. Backward scheduling from a due date is covered in forward versus backward scheduling.

Labor to Actuals

Job number, work center id, and actual date are mandatory. Hours and pieces are each optional, and either can be derived from the other using the operation's rate. A Complete column marks the operation finished, and an explicit start column overrides the default of using the earliest imported date.

The import always overwrites the days a file carries, so re-running a corrected extract fixes the numbers rather than doubling them. Days a file does not mention are preserved unless you ask for them to be cleared. Labor goes last because every row has to find an operation on a job that is already scheduled.

Conversions, blanks, and zeros

Reading the outcomes

Every row resolves to Created, Updated, Reused, or Failed. Reused is the default for a record that already exists, so a weekly product file of 3,000 rows reporting Created 9, Reused 2,991 is the desired result rather than a problem. When counts look wrong, the per-run log file has one line per row with the exact reason for each failure.

What import cannot bring, and why it matters

The settings that most differentiate a schedule have no column in any MRP export: the sequence-dependent setup matrix, operator skills and rosters, work center group strategies, bottleneck anchoring, and lot streaming transfer batches. They are configured once in EDGEBIC and then apply to every imported manufacturing order afterwards. The EDGEBIC product overview maps the engine, and the ERP integration architecture explains why one mask design serves every platform.

For the questions that come up before a project starts, see the Katana integration FAQ. The MRPeasy reference shows how little changes from one cloud MRP to the next, and the ERP scheduling add-on page frames the category.