First, the conclusion: in order entry mapping, the five fields most likely to break an EDI 850 go-live are ship-to versus bill-to addresses, units of measure, date qualifiers, line-item splits, and price and discount fields. Get the checklist for those five right before you talk about go-live. Get any one of them wrong, and the rest of the mapping will not save you. I have seen this pitfall three times, each time wearing a different disguise: goods shipped to the wrong address, quantities off by an order of magnitude, a warehouse preparing stock for the wrong date, the same line picked twice, and finance unable to reconcile an invoice. The root cause was always the same: someone mapped a field by what it looked like, not by what it meant.

Why these five? Because they all look familiar. Everyone recognises an address, a unit, a date, an order line, and a price. That familiarity is exactly what makes people skip verification. In EDI, a field with the same name can carry a different business meaning, and a field with the same meaning can arrive in a different code. Until you have confirmed the source, the meaning, and the edge cases for each field, mapping is guessing with a spreadsheet. For the introductory concepts behind this checklist, see EDI 850 purchase order mapping guide; this article covers what goes wrong after that introduction, in real projects.

Order Entry Mapping: Ship-To and Bill-To Addresses

The most expensive version of this pitfall I have seen happened in the first week of a 3PL go-live in eastern China last year. The bill-to address was mapped as the ship-to address. Nothing errored. The order was valid, the system was happy, and the goods went to the billing address, were refused there, and travelled back before anyone understood what had happened. When we reviewed the mapping document, it said only “address.” It never said which address, or which qualifier on the trading partner's order identified it.

Here is the step-by-step fix I now insist on. First, list every address role on the partner's order, not just ship-to. Second, confirm the qualifier and value for each role, and write it in the mapping table; “by default” is not an acceptable entry. Third, run real order samples through the system and check that the address landing in the system is the address the business team expects, not merely an address that looks plausible. Fourth, make a missing or conflicting address block the order and alert a person. Never let the system silently substitute one address for another. I have zero tolerance for vague wording in a mapping table: any entry that says “generally,” “usually,” or “should” goes back for rewriting.

Order Entry Mapping: Units of Measure and Date Qualifiers

The second pitfall is units of measure. The partner sends cases, the system books eaches, or the reverse, and the quantity is numerically correct while meaning something entirely different. In a southern China warehouse go-live I supported, order quantities looked normal until picking quantities were off by an order of magnitude, and the trail led back to a unit conversion that had never been built. The subtler version is that the same product uses different unit conventions for different customers. If mapping is done by product alone, without customer plus product, a collision is only a matter of time.

The third pitfall is date qualifiers. An order often carries several dates at once: order date, requested delivery date, earliest ship date, and latest ship date. If the system takes the first date field it finds, or treats a requested delivery date as a ship date, warehouse preparation falls apart. In a northern China project, the warehouse planned waves around the first date on the order, prepared the stock, and then learned the customer wanted it on a different day, leaving locations occupied for days for no reason. The fix is equally detailed: give every date field its own row, record its qualifier, its business meaning, its target field, and what happens when it is missing. There is no “approximately” in a date mapping.

Order Entry Mapping: Line Splits, Price, and Discount Fields

The fourth pitfall is line-item splitting. When the same product arrives on two lines, are those two separate requirements, or has one requirement been split in two? Without a rule, the system may merge them or pick them twice. The reverse problem is a single line that mixes batches or delivery requirements, which then breaks downstream splitting and shipping. Test with the partner's real split samples, covering merge, split, and partial cancellation, not just one clean standard order. A mapping that has only ever seen a perfect order has not been tested.

The fifth pitfall is price and discount fields. Unit price, discount, freight, and tax are manageable when each sits in its own place. Trouble starts when a partner hides a discount inside the price, or mixes freight into the unit price, and mapping copies the values literally. Finance then cannot reconcile by even a small amount, and the investigation turns into an argument across systems. Be explicit: which field is tax-inclusive, which is tax-exclusive, whether a discount is line-level or order-level, and whether a missing value means zero or an error. Any price convention not covered by the partner's current published policy must be written down in the mapping table, not left for an implementer to guess on the night before go-live.

Finally, here is the go-live night checklist. Every item must be checked; a single unchecked item means no go-live:

  • Ship-to, bill-to, and every other address role verified with real samples
  • Units of measure confirmed by customer plus product, with conversions in the mapping table
  • Every date qualifier on its own row, with missing-value handling decided
  • Line merge, split, and partial-cancellation samples all run successfully
  • Tax treatment and discount level confirmed for price, discount, freight, and tax fields
  • Exception orders blocked with a clear alert to a person, never silently released

Order entry mapping is not frightening because it has many fields. It is frightening because the fields all look like ones you already understand. Ask each of these five fields what it means until it cannot be asked any further, test it until it cannot be tested any further, and go-live night becomes uneventful. Skip that work, and the first real batch of orders will call you in the middle of the night, about the row in your mapping table that was written most vaguely.

By Lin Xiao, WMS Implementation Consultant