Key takeaways
- FMCG orders come in several formats at once, and the same customer may use email one week and EDI the next, so validation has to work on the content of the order, not its channel.
- The checks that matter most are product and pack codes, price against the agreed list, trade terms, promotion eligibility and credit position.
- Automation can read the order, run the checks, propose a correction and draft the query to the customer, so the sales operations owner reviews a prepared exception rather than rekeying the order.
- Credit holds, price overrides, promotion exceptions and substitutions stay with a named person with authority to decide.
- Start with a handful of high-volume customers whose orders you already understand well, and measure orders entered without touch and time from receipt to confirmation.
An FMCG sales operations team often starts the day with a queue of orders in five forms: a spreadsheet from a regional distributor, a PDF from a retailer's buyer, a few EDI messages, a portal export and an email that says "same as last week plus two pallets of the new flavor." Each one has to be matched to the right customer, the right product codes, the right price and the right promotion before anyone can enter it. Mistakes made here do not stay here. They become short shipments, invoice disputes and chargebacks weeks later.
What checks should an FMCG order pass before it is entered in the ERP?
Order validation is the set of checks an order must pass before it is entered in the ERP and released to the warehouse. It answers one question: is this order something we can accept and fulfill as written?
| Check | What is tested | Typical failure |
|---|---|---|
| Customer and ship-to | Account, delivery address and delivery window match a known record | Order from a new store with no ship-to set up |
| Product and pack | Code or description maps to a current item and pack size in the product master | Discontinued SKU, case versus unit confusion |
| Quantity rules | Minimum order, case or pallet multiples, available stock | 17 units ordered where only full cases ship |
| Price | Price matches the agreed price list or contract for that customer and date | Old price after a list change |
| Trade terms | Payment terms, freight terms, delivery lead time, minimums | Free delivery requested below the agreed threshold |
| Promotions | The promotion is active, the customer is eligible, volume and date conditions are met | Promotional price on an order placed after the window closed |
| Credit | Order value plus open balance within the customer's limit, account not on hold | Order pushes the account over its limit |
| Duplicates | Same order not already received through another channel | Email order followed by the same order in the portal |
A human who knows the accounts does many of these in their head. The trouble is that it only works while the same few people handle the same customers, and while volume stays manageable.
How do you handle FMCG orders that arrive by email, PDF, EDI and portal?
Orders reach FMCG sales operations in every format, often from the same customer. A structured EDI 850 is the standard electronic purchase order, sent from buyer to seller and typically carrying a purchase order number, item codes and quantities, requested dates and delivery locations, with pricing sometimes included (SPS Commerce). Many retailers and large distributors send these. Smaller customers use whatever is convenient: a spreadsheet template, a PDF purchase order, a message in a chat or an email.
Validation should check the content of an order, not the channel it arrived by. The first step is to read each order into the same set of fields: customer, ship-to, requested delivery date, line items with codes or descriptions, quantities and any stated price or promotion. For structured messages that is a mapping exercise. For PDFs and emails, it is document reading, the same document reading used for invoices and contracts. Once the order is in a standard shape, every check runs the same way.
Which systems hold the data to check against?
Validation is only as good as the reference data behind it, and in FMCG that data lives in several places.
- ERP (for example SAP, NetSuite or Microsoft Dynamics): customer master, item master, price lists, open orders, credit exposure and payment terms.
- Trade promotion or revenue growth management tool: promotion definitions, eligible customers, dates and funding.
- CRM: account ownership, agreed terms and notes from the key account manager.
- Warehouse or supply planning system: stock availability, allocations and lead times.
- Contract or pricing spreadsheets: often the real source for customer-specific deals, and a common reason checks need a data cleanup first.
The workflow reads from these through their APIs or exports and writes back only what the ERP should receive. This is the integration work covered in Connecting AI to your ERP, CRM and data warehouse. Be explicit about which system wins when two disagree. A price in a spreadsheet that contradicts the ERP list is a finding, not a tie-break.
How can AI automate order entry for incoming FMCG orders?
AI can do the reading, matching and checking of an incoming order, while the decisions that carry financial risk stay with people.
A workflow can detect a new order in a shared mailbox, portal or EDI feed, read it, identify the customer and match every line to the product master, including customer-specific item codes and descriptions that do not match your catalog exactly. It then runs the checks above and sorts the result.
Orders that pass every check are prepared for entry in the ERP, either created through an integration or presented as a ready-to-confirm draft, depending on how much autonomy you have agreed for that customer. Orders with a fixable problem, such as a code that maps to a renamed item or a pack size expressed in units rather than cases, get a proposed correction. Orders with a real conflict, such as a price below the list or a promotion that has expired, become exceptions with a short explanation and the supporting data.
It can also draft the reply. Many retailers expect a purchase order acknowledgment, the EDI 855, to confirm whether each line is accepted, adjusted or backordered, and some require it within 24 to 48 hours of the order (SPS Commerce). A workflow that already knows the result of each check can prepare that response, or the plain-language equivalent for customers who do not use EDI.
Who should receive each FMCG order exception?
An exception sent to a general queue waits for whoever notices it first. Route each type to a named role with the authority to resolve it.
- Sales operations owner (customer service or order desk): unmapped products, quantity rule breaches, missing delivery details and duplicate orders. They contact the customer and correct the order.
- Key account manager: requests that depend on a commercial agreement, such as a price or term the order assumes but the system does not record.
- Credit controller or finance: orders that exceed a credit limit or involve an account on hold.
- Trade marketing or promotions owner: promotion eligibility questions and disputes about conditions.
- Supply planning: orders that cannot be met from available stock and need an allocation or substitution decision.
Each routed exception should carry the order, the failed check, the data it was compared with and a proposed next step. The person should be able to decide in one read.
Which order validation decisions should stay with a person?
Order validation prepares the decision, and five kinds of decision remain with a named, accountable person. How to keep a named approver in the loop covers the design in more detail.
Credit releases. NetSuite, for example, can be set to block an order that puts a customer at or above its credit limit, or to warn and let the person entering the order choose whether to continue (Oracle NetSuite). Either way, whether to release an order for a customer near its limit is a financial judgment for credit control, not for the workflow.
Price overrides. Honoring a price different from the list changes margin and sets a precedent with the customer.
Promotion exceptions. Accepting an order that misses a promotion's conditions can fund a discount the budget never planned. Approval of promotion terms is a separate workflow, covered in FMCG trade promotion approvals.
Substitutions and short shipments. Offering a different item or splitting delivery affects the customer relationship and often the retailer's own planning.
New customers and new terms. Setting up accounts, ship-to locations and terms follows your normal controls.
Illustrative example: Consider a beverage manufacturer selling through regional distributors and a few national retail chains. On a Monday morning, thirty-odd orders arrive: some by EDI, several as spreadsheets and PDFs, two by email. A workflow reads them all and runs the checks. Most pass and are queued for entry. One distributor's spreadsheet uses an old code for a renamed product, so the workflow maps it, shows the match and asks the order desk to confirm. A chain's order includes a promotional price after the promotion's end date, so it goes to the key account manager with the promotion terms attached. Another order would take a distributor over its credit limit by a small margin, so it goes to the credit controller with the open balance and payment history. The order desk spends the morning on four real exceptions rather than entering thirty orders line by line.
How do you know order entry automation is working?
Four measures show whether validation is helping the business.
Orders entered without a touch is the share of orders that pass every check and reach the ERP without anyone rekeying or correcting them. Track it by customer, since it shows where rules or data need work.
Time from receipt to confirmation measures how long customers wait to hear that their order is accepted, adjusted or queried.
Order errors found after entry counts the short shipments, wrong prices and invoice corrections traced back to the order. Buyers often reconcile invoices against orders, and pricing mismatches are a common cause of disputes and short payments, as SPS Commerce notes in describing the EDI 810 invoice (SPS Commerce).
Exception age shows how long a flagged order sits before someone acts, which tells you whether routing and deadlines are working.
Take a baseline first. Count orders received, errors found after entry and hours spent entering orders over a few representative weeks.
How do you start automating FMCG order entry?
- Choose three to five customers with high volume and predictable orders, covering at least two different channels.
- Write down the checks your order desk applies today, and the data each one uses. Mark which are hard stops.
- Check the reference data first. Compare the price lists and product master with what the team actually uses, and fix the largest gaps before blaming the workflow.
- Run validation in shadow mode on live orders for a few weeks, so you can compare each finding with what the order desk caught by hand.
- Turn on routing and drafted replies, and name the owner and response time for each exception type. Decide how much oversight each step needs before widening automation.
- Add customers and checks gradually, and review the false alarms every week at the start.
How Kastling approaches order validation
Kastling treats order validation as one workflow to understand before anything is built, as part of the AI Integration & Automation service. It starts with a free discovery call. A separately scoped paid audit is common next for FMCG operations, reviewing how orders arrive today, which systems hold customers, products, prices and promotions, and where errors and rework appear.
The validation connects to the ERP and other systems you already run rather than replacing them. Rules are agreed with the order desk and the people who own pricing, promotions and credit. Exceptions go to the person who owns the decision, and credit releases, price overrides and promotion exceptions stay with them. We agree up front how success will be measured and test the workflow on your real orders before it runs on live ones.
Human approval planner
Describe one AI task and see the level of human control it needs, with a control pattern your team can set up.
Questions
Is this the same as EDI order processing?
EDI handles the transport and format of an order from customers who support it. Validation is the layer that checks what the order says against your own data, and it is needed whatever the channel. In practice, EDI orders are often the easiest to validate because the fields are structured, while emailed spreadsheets and PDFs need reading first.
Which customers should we start with?
Pick customers with high order volume and a repeating pattern, where your sales operations team already knows the common errors. Avoid starting with a customer whose terms are negotiated order by order, since the rules are too unstable to check against. Add the more complicated accounts once the first group runs with few corrections.
Can the system accept an order that fails a check?
That is a policy choice, and it should be set per check. A product code that does not exist cannot be accepted. A price a few cents off the list might be corrected automatically and noted. A credit breach goes to the credit owner. Write the rule for each check before building, and make the default for anything uncertain be review.
How do we keep the product master and price lists reliable enough to check against?
The checks will expose every gap in them, which is useful but uncomfortable. Treat the first weeks as a data cleanup exercise: each false alarm usually points to a stale price, a missing pack size or a customer-specific code that was never recorded. An assessment of your data before building shows how much work this is likely to be.
How do you automate order entry from email or PDF purchase orders?
Read each email or PDF into the same fields a structured order has: customer, ship-to, requested date, line items and quantities. Match every line to the product master, run the validation checks, then create the sales order in the ERP or present it as a ready-to-confirm draft. Orders with a conflicting price, promotion or credit position go to a named person instead of straight through.