Automating revised POs and change orders

Why a revised PO is harder than a new one

A new PO is work, but it's honest work. You read it, you enter it, you confirm it. A revised PO looks almost exactly like the order you already entered. Same customer, same PO number, same 40 lines. Somewhere in there, one thing is different.

So someone on your order desk prints both versions, or puts them side by side on two screens, and goes line by line. Is line 12 still 400 pieces, or is it 450 now? Did the ship date on line 30 move? Is that a new line at the bottom, or was it there before? The customer's buyer rarely tells you. Sometimes the PDF says "Rev 2" in small print. Sometimes it doesn't say anything at all.

The risks are real and they're not small:

• The change gets missed. The old quantity ships, and you find out when the customer short-pays or sends it back.
• The revision gets entered as a new order. Now you have two orders for the same thing, and the plant builds both.
• The change lands after the work started. The date got pulled in on a job that's already on the floor, and nobody told production.

None of this is a skill problem. It's an attention problem. Comparing two nearly identical documents is exactly the kind of work people are bad at and software is good at.

What the order research says about changes

There isn't much public research on revised POs alone. The best numbers come from APQC, a nonprofit benchmarking group, in two reports on sales order processing. Both were sponsored by Esker, which sells order automation software, so read them with that in mind. Both are also several years old. The patterns still match what order desks describe today.

In its 2016 report, APQC says the most significant cost driver in sales order processing is labor. It tells controllers to count not just the time to enter an order, but the time and cost when part of an order is lost, data entry errors are made, or "a change to an order arrives." In the same report, the cost to manage a sales order ranged from $5.11 per order for top performers to $40.87 for bottom performers (N=608), almost eight times more.

APQC's 2018 report looked at why orders need a person to step in. Across 724 respondents, here's the median share of manual sales order interventions by cause:

Why sales orders need a person to step in

APQC, 2018. Median share of manual sales order interventions by cause, N=724.

1. Invoicing issues25%
2. Pricing issues20%
3. Contract issues18%
4. Wrong customer master data15%
5. Other causes20%

Source: APQC, Transform Customer Service and Operations Through Order Automation (2018, Esker-sponsored), Figure 3.

Pricing, contract terms and customer data are exactly what a revised PO tends to touch. A buyer changes a price, moves to a new contract, or sends the order to a different ship-to. Those are three of the top four causes in APQC's chart.

EDI doesn't make the problem go away. APQC notes that up to a third of EDI orders present some sort of discrepancy. That figure comes from an Esker paper the report cites, not from APQC's own benchmark, so treat it as a vendor estimate.

And the payoff for getting changes right shows up in the results. APQC found that companies in the top quartile for orders that need no human intervention to create, modify or fulfill scored 96 on perfect order performance, against 90 for the bottom quartile. Perfect order means complete, on time, undamaged and with the right paperwork. That's a correlation, not proof that automation alone caused it. But the direction is clear.

Three ways a change arrives

How hard a change is to handle depends mostly on how it shows up.

EDI has its own messages for this. The X12 standard defines a Purchase Order Change Request - Buyer Initiated (860) for the customer's change, and a Purchase Order Change Acknowledgment/Request - Seller Initiated (865) for your answer. If your big customers send 860s, your EDI software may already apply some changes for you. The trouble is that most mid-size manufacturers and distributors get far more orders by email and PDF than by EDI. We cover that side in automating email and PDF orders without EDI.

The revised PDF is the one that eats the most time, because the document doesn't tell you what changed. You have to find out.

What software can do for you

Here's what an automated change-order step looks like when it's done well. It's the same reading that handles new POs, plus one more step: compare.

1. Spot that it's a revision. The PO number already exists as an open order for that customer. That alone is enough to stop it from being entered twice.
2. Read the new version the same way it reads any PO: customer, ship-to, lines, part numbers, quantities, prices, dates.
3. Compare it to the open order, line by line. Not to the old PDF, but to what's actually in your ERP. That catches the case where someone already changed the order by phone.
4. Show only the differences. "Line 12: quantity 400 to 450. Line 30: date Nov. 20 to Nov. 13. Line 41: new." Everything that matches stays out of the way.
5. Check each change against what's already happened. Has the line been picked? Is the job released to the floor? Did it ship? Does the new price match the contract?
6. Draft the change in the ERP and the reply to the customer, and wait for a person to approve both.

Your ERP can help keep the history straight. Microsoft Dynamics 365 Business Central, for example, can archive a sales order several times, saving a different version each time, and can archive automatically when someone changes a document's status or sends it by email. Other ERPs have their own version or change-log features. Whatever you use, the goal is the same: when a customer asks "what did we change?", you can answer in a minute.

The reply matters too. A change isn't done until the customer knows you accepted it, or that you couldn't. That's the same step as any order confirmation, which we walk through in automating order confirmations and acknowledgments.

Let software find what changed. Let a person decide what you'll accept.

What stays with a person

Finding a change is mechanical. Accepting it is a business decision. Keep these with a person, every time:

• Changes to work already in progress. A date pull-in on a job that's on the floor may mean overtime, a bumped order, or saying no.
• Price changes. If the buyer's new price doesn't match your contract or quote, someone who owns the account should call, not a system.
• Cancellations of made-to-order or special items. Your terms may allow a charge. That's a conversation.
• Anything already shipped or invoiced. That's a return, a credit or a new order, not a change.
• Ship-to changes on orders that are close to shipping. Confirm them with a contact you already know.

The software's job is to put these in front of the right person fast, with the facts already pulled together. The decision stays where it belongs.

Where the other side has a point

To be fair, not every company needs this.

• If changes are rare, a short checklist may be enough. If you get three revised POs a month, a careful person and a highlighter will do.
• If your changes come by EDI, your EDI provider or ERP may already handle 860s. Ask them before you buy anything new.
• The research is vendor-sponsored and dated. APQC's numbers come from reports Esker paid for, from 2016 and 2018. They're useful for the pattern, not as a promise of what you'll save.
• Bad reference data breaks the compare. If one customer exists under three account numbers, or customer part numbers aren't cross-referenced to yours, the software will match the wrong order or flag every line. Clean those lists for the customers you start with. Our piece on clean data before AI covers which lists matter.

How to start

You don't need a big project to find out if this is worth automating. Try this for 30 days:

1. Count your revisions. Tag every revised PO, change email and change call.
2. Log what changed. Quantity, date, price, ship-to, added line, dropped line, cancel. Keep it to one line per change.
3. Log what it cost. Time to find the change, and any change that got missed or caught late.
4. Look at who sends them. Check whether a few customers send most of the revisions.
5. Start there. Automate the compare for those customers, with a person approving every change, and measure the time per change before and after. Our guide to measuring office automation ROI shows how.

If revisions turn out to be a big share of your order desk's week, that's a strong sign the desk needs help overall. We list the other signs in signs your order desk needs automation.