The Work: one customer account, worked through

the account

The order cleared. The rule failed.

Priced for 22% in 2019, earning 14.3% across eighteen plants, with an ERP cutover coming and the price on nine of them still corrected by hand.

A distributor was preparing for an ERP cutover. Every customer’s terms had to be settled before the integrator loaded them.

Plant 14 placed an order on a Tuesday: six line items, standard SKUs, the kind the plant had been placing for three years.

The order priced from the price file, six percent above the contract price. The CSR who usually corrected the plant-level price by hand before release was on PTO. The order released through standard workflow, shipped, and was invoiced on Friday. The customer’s AP system matched the invoice to Plant 14’s purchase order, found the unit prices six percent above it, and blocked payment.

The plant’s buyer called the rep. The rep asked the order desk lead for a credit before the end of the day. The credit clerk would not issue an unreferenced credit that size without the regional sales director’s approval, and the director was in customer reviews until Thursday. By the time the approval came, the customer’s category manager had frozen the distributor’s account: every open invoice on hold until the recurring price variance was explained.

That was when it reached the CFO and the CRO: not as a pricing question, but as a payment hold across eighteen plants.

The CRO opened the account: plants 1 to 9 mapped to the 2019 contract; plants 10 to 18 not. The dispute log showed eleven earlier disputes from this customer in two years, each resolved the same way: a credit memo for the price difference. This one, the twelfth, was settled with another credit memo, and the hold came off.

The controller, asked how an account that size kept generating credits, set its last two years against the 2019 account model. The contract had been priced for a 22% margin after product, freight, kitting, and account-service costs. On that basis, the account now earned 14.3%.

The model assumed monthly consolidated buys. The plants now ordered weekly, each order about a quarter the old size, and every one of them shipped and invoiced on its own. Freight to the southeast plants had moved to a 3PL the model never priced.

The model costed kitting as standard master-carton packs. Over three years, the plants’ maintenance planners had asked for pre-sorted job kits, bagged and tagged, so their crews would not sort parts on the floor. Customer service agreed to protect the relationship, and the warehouse added two bench steps to every pick for eleven SKUs. Nobody added a charge.

Service touches per order had doubled.

The contract was still the contract on file, but the account no longer behaved like the account priced in 2019.

Against the original account model, the shortfall came to $1.1M a year, and it was widening.

The CRO put thirty minutes on the calendar for Thursday with the pricing lead and the master data team. The meeting moved twice. The same people were loading customer files into the test environment and working through the integrator’s defect list, and the cutover date did not move for anyone.

On cutover weekend, the integrator would load the customer file as it stood: nine plants on the contract, nine on the price file, and no trace of the corrections made by hand in between. In the new system, nobody at order entry could change a price without an approval route, and the route had not been built.

Six other accounts on the books had the same shape. Plant 14 was only the second-largest.

the pattern

Settled twelve times. Solved none.

Each credit cleared one held invoice and left the terms for the next order where they had always been, outside the system.

The 2019 contract priced an operating profile: nine plants, monthly consolidated orders, standard master cartons, one freight lane. Nine more plants came on at the contract price, by practice, never by amendment. All eighteen kept changing how they bought. The contract stayed in the file.

Everything that no longer fit it was handled order by order. The CSR corrected the plant-level price by hand before release. The warehouse built the kits. When a correction was missed, the customer’s AP held the invoice and AR credited the difference. Each settlement taught both sides something: the customer’s AP, that a held invoice comes back credited; the order desk, that a correction by hand is faster than a change to the contract. Every settlement closed a dispute. None wrote a rule.

The first time, it is a favor. The second time, it is expected. By the third, the customer enforces it on the invoice.

None of it was a mistake. Those accommodations helped win the account and keep its plants running, and as the ordering pattern changed, they became the baseline the business honored.

None of it was free. Four orders where there had been one meant four pick tickets, four packing slips, four invoices, and minimum freight charges on shipments that used to move consolidated. The kits took bench time the warehouse budget never carried. The carrier bills sat with every other customer’s freight, the kitting hours in the warehouse line. All of that came out of the account’s margin, and none of it showed up against the account. The credits cost something else: each one brought the price back to the contract, but only after a held invoice, a dispute, and weeks of cash waiting.

Inside the business, no one could settle it alone. Sales would not reopen a contract and risk an account that size over a freight surcharge. The warehouse could not refuse a kit a plant was waiting on. Finance saw the cost months later, blended into freight and labor variances it could not trace to one customer. Master data could not load a rule nobody had decided.

The cutover ended the option of not deciding. The new system would price every order from whatever terms it was given. Load the price file, and nine plants are billed above the contract from the first order. Map all eighteen plants to the contract, and the account goes on earning 14.3% where it was priced for 22%. Neither has to be the last word. The business can keep today’s prices through cutover on purpose, with an owner and a date to revisit them.

What it cannot do is let the integrator load the file before anyone decides.

the evidence

The dispute is only where the rule surfaced.

The cutover needs a decision about what the account actually gets, and the grounds for it sit in records nobody has read side by side.

The easy answer is to put plants 10 to 18 on the 2019 contract and move on. It would stop the disputes, and with nobody owning the terms, it would lock in the shortfall. I do not take it to management until four questions have answers.

If a nineteenth plant comes online, does it inherit the contract automatically? If weekly orders keep replacing monthly consolidated buys, does the account still carry the same freight logic? If kitting and service touches are now part of normal service, does the price still mean what it meant in 2019? If the next invoice prices from the price file before anyone catches it, does AR defend the invoice, credit the difference, or send the issue back for a decision?

Answering them starts with the whole record, not a sample. A review usually pulls a few months of one account’s invoices. I work from all of it, on the account and on every account like it. Code joins what the systems never joined: ship-to sites against the plants a contract covers, customer part numbers against the item master, a carrier invoice that names a job site instead of an order number. The contract then sits beside what was billed, order by order. AI tools read the text no join can use: the order notes, the credit memo reasons, the contract clauses. What was never kept, or will not connect, I list as unmatched rather than guess at.

On Plant 14, that comparison holds every order from plants 10 to 18: the ones corrected by hand before invoicing and the twelve that were not. Across the book, it turns up the six other accounts with the same shape, most of them smaller and further down the list than reviews usually reach.

Every discrepancy comes with its source attached: the order line, the note, the clause. I take those to the people who handled the account. The CSR, the buyer, and the warehouse lead can say why each arrangement exists and what it protects.

What nobody wrote down, they can tell me; the tools cannot.

The code and the AI tools run on the company’s data inside its confidentiality agreement, kept out of model training and apart from any other company’s. At handover, the findings and the evidence behind them go into the company’s own systems, and my working copy is deleted.

The answer is not always a price. Weekly orders may be what the plants actually need, and a stocking program near the southeast plants may cost less than a freight surcharge would recover. The kits may be worth keeping at a charge. The rep’s plan may pay on volume and never see the freight. I test each explanation against the records before the terms change.

For Plant 14, three things have to be settled: which terms commercial can change or price without damaging the relationship, which fulfillment rules operations can change to lower the cost to serve, and how master data will maintain what was agreed. I put the choices and their customer consequences before management, settle the decisions within the agreed authority, and give AR an instruction it can use the next time an invoice is disputed. Any new freight term or kit charge gets an effective date, a conversation with the customer, and, where the contract requires it, the customer’s agreement before billing applies it.

The records can show where the margin went. Only management can decide whether to take it back.

The logic has to live where the business runs: in the customer record, in the pricing configuration, in the order-release workflow, in the report the controller closes against. The record owners and the integrator put it there; I check that what they load is what was decided.

the override

Whatever bills under exception is the real rule.

The routine order proves nothing. The one nobody planned for does, and only if it reaches someone who can see what approving it costs.

The work is tested by the order that does not fit. On Plant 14, it is easy to name: a nineteenth plant’s first order arrives before anyone has set its terms, or a kit order comes up one component short on a Friday afternoon with the plant’s crew waiting.

At that moment, someone acts. The CSR releases the order or holds it. The pricing desk lets the price stand or changes it. A manager approves an override from an email so the truck can leave. Billing sends the invoice as it stands. What they do is what the business actually charges that account.

Most overrides are approved. That was never the problem. The problem is the override whose cost nobody sees: the freight absorbed, the kit built at no charge, the price held below the file. An override can be the right call. Its cost should be in front of the person approving it, and it should stay on the account where the next review will find it.

Every override is a pricing decision, whether or not anyone priced it.

For Plant 14, the standard is whether those orders stop at someone authorized to approve them, with the cost in view, instead of going around the new system through a side file. The standard does not change with the form of the work; who runs the test does. In advisory work, management runs it against the decisions it made. In interim and project work, I run it with management through repeated use, including the cases that send people looking for the old files, until routine business carries the decision and a named manager decides what falls outside it.

That manager inherits the decisions, the records that carry them, and the authority to keep them current. The next CSR does not dig through old emails and credit memos to learn what Plant 14 gets. The account record says so.

the result

The same question stops coming back.

The account terms live where the next order can find them, and the people who kept them straight by hand turn their judgment to what is new.

When the work holds, the first change is quiet. The dispute log stops carrying the same question about the same plants. Billing sees the account terms before the invoice leaves. AR has a decision to apply when a payment comes back short, instead of a phone call to make.

The cost of the old way was never only margin. It was the most experienced people on the desk working as a buffer between a stale price file and an irritated plant: the senior CSR correcting prices by hand before every release, and the pricing analyst rebuilding the same account from the contract, the override log, and the last credit memo.

What changes is what their judgment is used for. It goes to the next exception instead of re-settling the last one.

Some exceptions should stay. The kits may be worth building at no charge; a plant may be worth a freight term no other account gets. The difference is between a subsidy nobody priced and a concession someone approved knowing what it costs, recorded where the next order will find it.

Time back on the desk is capacity, not cash. It turns into money when the business stops paying for overtime and outside help, or takes on more orders without adding people; the orders reach earnings first and cash only when they are collected. Fewer credit memos show the invoices are right the first time. Margin shows up as freight and kits billed where they were agreed; cash shows up as payments arriving on terms. Until the ledger shows it, it has not happened.

Every term on the account is one the business chose, and knows the cost of.

the entry

The plan is approved before the operating questions are answered.

A system replacement, a portal, an acquisition, a new site, or an owner’s plan, each arriving with a date that will not wait for the decisions it needs.

A system replacement puts a date on every open question. ERP is being replaced or CPQ rebuilt, the go-live is on the calendar, and the integrator needs pricing hierarchies, configuration logic for built-to-order products, availability promises, and account terms settled before the build applies them to orders.

A portal takes out the person who cleaned up the order. The inside rep caught the unloaded contract price and the wrong part number before release. On the portal, the customer’s order goes straight to the warehouse and the invoice: the wrong part picked, list price billed, and a return and a held payment behind them.

An acquisition puts two order books under one name. The plan calls for the acquired lines to sell through the existing channel, but they arrive with a separate item master and their own customer price files, freight terms, and distributor pricing agreements. The combined business needs shared accounts assigned, stock placed, and terms settled while both order desks keep taking calls.

A new site splits what one building used to ship complete. Unless someone decides which site serves which account and which stock sits where, orders split across two buildings, transfers between them carry freight nobody priced, and the account list, stock assignments, and replenishment rules are still being settled the week it opens.

An owner’s plan asks for two things the order book cannot always give at once: the service levels that won the accounts, and less inventory behind them. The first stock marked for reduction is whatever looks like excess against the item record, and some of it is the safety stock behind the tightest delivery promises. Unless the promise rules change with it, dates start slipping and expedited freight comes out of margin.

None of them waits for the order book to stop.

the first step

The first conversation is about what is not working.

It should leave the business with something useful whether or not an engagement follows, starting with a clearer view of where the margin is going.

Write to consult@davelloandco.com. It reaches the person who would read the records, sit with the desk, and carry the decisions into use.

The reply says plainly whether this is work for me, for the company’s own people, or for someone else.

A worked example: one customer account, from payment hold to new terms.