Integration and Cloud

When your customer demands an EDI connection

In this article

Nori.mx7 min read

ANSI X12830Forecast862Firm schedule850/860PO and change856Ship notice810Invoice997Acknowledgment824Error notice856 · ASNPallet › box › pieceSent when the truck leavesBox labelSame ID

The email arrives without warning and it has a date on it. From that day, that customer’s delivery schedules and shipment notices move over EDI, there is a test to pass first, and if you are not ready the account is at risk.

The first reaction is usually to go find out what an EDI costs. That is the wrong question, because what they are asking for is not a product: it is three different capabilities, and only one of them is software.

What it is, without the brochure

EDI is agreeing on the format of a message so that two systems can read it without anyone typing. It is older than the web, which is why it looks the way it does.

On the border, in manufacturing and automotive, it is almost always the ANSI X12 standard. The messages they will name are few:

  • 830 the forecast, the demand schedule several weeks out.
  • 862 the firm shipping schedule, the short-term one.
  • 850 / 860 the purchase order and its change.
  • 856 the advance ship notice, the ASN. This is the one that matters.
  • 810 the invoice.
  • 997 the acknowledgment: confirms your message arrived and could be read.
  • 824 the notice that the content had a problem.

If your customer is European-owned they may ask for EDIFACT or Odette instead of X12. The names change, the substance does not.

What they are actually asking for

Three capabilities, and it is worth seeing them separately because they fail separately.

Understanding the demand. The 830 and the 862 arrive every week and change every week. The classic trap is treating a schedule as if it were an order: you build to a forecast, the customer cuts it, and the inventory stays with you. You need to know which part of the message is a commitment and which part is visibility, and that lives in the commercial agreement, not in the standard.

Telling them before the truck arrives. The ASN has to go out when the shipment leaves, not when it lands. Its entire value is that the customer’s warehouse already knows what is coming and can receive it against a scanner instead of against a piece of paper. An ASN that arrives after the truck is worth nothing, and it gets charged for.

Labeling so their scanner closes the loop. This is the part almost nobody mentions in the first email, and the part that generates the most chargebacks. The ASN describes containers with a unique identifier; the label stuck on the box has to carry that same identifier as a barcode, in the format the customer requires. If the data and the label do not match, the box gets received by hand and you pay for it.

The hidden cost of doing it by hand

Many suppliers “comply” by logging into the customer portal and typing. It works at low volume and it fails exactly when you grow.

What makes that path expensive is not the wage of whoever types. It is the chargebacks. A late ASN, an ASN that does not match what was on the truck, a wrong label: each has a fee, per shipment, and it is written in the supplier manual they sent along with the email. It is worth reading that table before deciding anything, because it usually answers the budget question by itself.

And there is a cost that never gets invoiced: manual entry happens at the end of the day, in a hurry. It is the single most error-prone moment in the whole operation.

Three paths, cheapest to most complete

1. Customer portal and manual entry. Zero investment, a cost per shipment, and no traceability. It is defensible as a temporary step, on one condition: give it an expiry date the day it starts. Without a date, it stays for three years.

2. An EDI service that translates. The realistic path for most suppliers. A provider handles the envelope, the connection — AS2 or SFTP — the map for each customer, and watching the acknowledgments. What reaches your side is a file you can actually read, and what you send out is the same. You pay a fee, but the map stops being your problem.

3. Direct integration with your own system. This makes sense when volume or the number of customers justifies it. And here is what nobody warns you about: every customer has their own version of the same standard. There is no such thing as “complying with the 856”. You comply with that customer’s 856. Three customers means three maps, three supplier manuals and three ways of numbering a box.

The decision is rarely technical. It is about volume, about how many customers you have, and about what one chargeback costs you.

What decides whether this goes well

It is not the EDI. It is whether your own data is ready. Four questions:

  • Is the customer’s part number tied to yours? If that table lives in one person’s head, the EDI will not work. It is the first thing to build and there is nothing glamorous about it.
  • Do you know what is in each box before you close it? An ASN describes packaging structure: pallet, box, piece. If your system only knows the shipped total, you cannot issue a valid ASN no matter how much EDI you buy.
  • Is the label printed at packing time, or afterwards? If it is printed afterwards, the identifiers and the boxes drift apart. Always. It is the most common cause of an ASN that does not match.
  • Who watches the 997s and the 824s? An EDI nobody watches is worse than no EDI, because it leaves you believing you gave notice. There has to be a person and a screen where a rejected message is seen the same day.

The first three are inventory questions, not messaging ones. If the warehouse does not know where its stock is, EDI only makes the discrepancy travel faster: we covered that in when a WMS is actually needed.

Four questions for the EDI provider

  1. Who maintains the map when the customer changes it? They will change it. If the answer is that each change is quoted separately, get the rate now.
  2. What happens if a message is rejected at two in the morning? Who finds out, through what channel, and how fast.
  3. Can I take my maps and my history with me if I leave? In writing, before you need it.
  4. How is it billed? Per document, per character volume, or a flat fee. Volume-based billing surprises you in the good month, the one where you shipped more.

What EDI does not fix

It does not fix inventory you do not trust, it does not define your packaging, and it does not substitute for the supplier manual nobody has read all the way through yet.

What it does do, properly built, is remove the point where the operation breaks today: one person, in a hurry, at the end of the shift, copying data from one place to another. That is integration work, and it looks more like putting your own house in order than like buying a system.

It also does not decide what you calculate your promised date from. EDI is one of the paths an order reaches you by, and if that promise is not calculated against anything, connecting it only brings forward the moment the customer finds out.

And if EDI is one of several fronts open at once, the order matters: some of them produce the data the others need.

If that email just landed and you have a date on you, the first move is not to get a quote. It is to read the supplier manual and answer the four questions above. Twenty minutes is enough to know which of the three paths is yours.

Related service

Integration and Cloud

Make your ERP, your warehouse and your store speak the same language.

See the service

Articles on this service

IRS calendarNov 1Last testsNov 9TCC changesNov 19, 3 p.m.FIRE shuts downJan 1IRIS onlyYour FIRE file1099 · 2026from your accounting systemIRIS won't take itIRIS has two doorsPortalCSV, up to 100A2AAPI, up to 100 MB

Integration and Cloud

Your January 1099s won't go through FIRE

The IRS shuts down FIRE on November 19 and from January 1099s can only be e-filed through IRIS. Who needs a new code, what changed and which data to check.

6 min read

Maximum lifetimeper yearBefore398 days1Mar 15, 2026200 days2Mar 15, 2027100 days4Mar 15, 202947 days8+Let's Encrypt: 64 days from Feb 10, 2027Where they liveSite and portalCustomer APIVPN and webmailProxyOne, in many places*.yourcompany.com

Integration and Cloud

Your SSL certificate no longer lasts a year

Since March 2026 a certificate lasts 200 days, and in March 2027 that drops to 100. Where yours live, what breaks when they expire and how to stop renewing by hand.

7 min read