Manufacturing and Distribution
Orders arrive as documents and become orders by being typed in. We put the reading and the checking into a system that runs the same five checks every time, and send the commercial exceptions to whoever has to defend them.
What it costs today
Every one of these is a check somebody would have made if they had time, on a document somebody had to read anyway. Six patterns turn up in almost every distribution business we map.
Email, portal downloads and attachments all end up as somebody reading a document on one screen and keying it into another. The customer sees the ones that get keyed wrong.
Corrections, re-ships and a credit note nobody budgeted for.Contract pricing was uplifted in April, the customer is still ordering at last year's number, and nobody notices until the invoice is queried or the margin report comes in low.
Margin given away one line at a time.Allocated, on hand and inbound are three different numbers, and the one a salesperson quotes from is whichever screen they had open.
A promise made on stock that is already spoken for.Different delivery points, a name that changed, and an account somebody opened rather than searching. Each copy carries its own price agreement and its own credit position.
Credit exposure that is real and invisible.The purchase order slipped, the sales order kept its promised date, and the two facts live in systems that have never been introduced.
A service reputation you are paying for and not earning.Margin by product, by customer and by territory is assembled by hand out of two systems, so it arrives late enough to be history rather than a decision.
The reporting week is its own job.The operating pattern
Controlled delegation is the whole design, and everything below this band is an application of it. Reading and checking go to the system. Anything that costs a customer money or changes what they were promised goes to a person.
Purchase orders, remittances, supplier confirmations, delivery notes and enquiries. The same shapes over and over, arriving faster than a person can read them, and every one handled as though it were the first.
Matching the customer, checking the price against the agreement, checking the quantity against stock, checking the date against replenishment and checking the value against credit. Five checks, in the same order, on every order.
A price above what the customer wrote down, an allocation between two accounts, a substitution, a credit call. These go to whoever will have to defend them on the phone, with the options already costed.
See it run
The order gets read, matched to the customer, checked against stock, contracted pricing and credit, and split so the clean lines move while the exceptions wait for a person.
Connected systems
Purchase order PO-48927 Trentwood Facilities Ltd
Result
Completed automatically
Requires a person
Releasing the order to the warehouse.
Sales desk
Connected systems
Purchase order PO-51330 Baymoor Industrial
Result
Completed automatically
Requires a person
Whether the customer takes a split delivery.
Sales desk
Connected systems
Purchase order PO-50874 Ellerby Maintenance Group
Result
Completed automatically
Requires a person
Honour the old price or correct it, and who tells the customer.
Account manager
This demo uses fictional data. The workflow gets configured around your systems. Build a workflow like this
The work
Order entry is what businesses come for. The customer and pricing record is what makes it safe. The two usually get built together, with the cleaning slightly ahead.
Every line checked before it becomes an order
Reads the purchase order whatever form it arrived in, matches the customer, and checks every line against the price agreement, live stock, the replenishment schedule and the credit position. Clean orders release under a rule you write. Anything that fails a check goes to a person with the options already worked out.
Data: orders, customer master, pricing, stock and credit. Human: a salesperson decides anything commercial.
The unglamorous work that everything else needs
One customer master without duplicates, price agreements that are current, and a stock position that means what it says. This is the least interesting workflow on the page and it is the one that decides whether the other three are possible, because a check is only as good as the record it checks against.
Data: your customer, pricing and inventory masters. Human: commercial owns what a merge does to a price.
Which accounts are moving, and which stopped
Buying patterns by account and by territory read together, so a customer whose order frequency has halved surfaces while there is still a relationship to save. The account manager gets the account, the history and a reason to call, and decides whether to make the call.
Data: your own order history by account. Human: the account manager decides who gets called.
The questions that currently take a week to answer
Margin by product, customer and territory after settlement rather than at the quote, service performance against what was promised, and demand patterns that inform replenishment. This is also where predictive work belongs, once the record underneath it is good enough to trust, and not before.
Data: your own order, delivery and financial records. Human: analysts validate before anything is acted on.
A checking system is only as good as the record it checks against. Run order entry against a customer master with three copies of the same account in it and the workflow will price against the wrong agreement, check credit against the wrong balance, and do both faster and more consistently than a person ever managed. That is the argument for the second workflow, and it is why the cleaning sits in the sequence below rather than in the assumptions.
The decisions
These get agreed with commercial, customer service and whoever owns the ERP before a build starts. Between them they decide what the workflow is allowed to release on its own.
What counts as working. Accepted output against your own baseline is the proof. The baseline gets taken before the build exists, and afterwards the same work is measured against it with the review burden counted openly, because a system that processes plenty and has to be corrected twice has saved nothing. How we measure AI work sets out the method.
How it starts
The first build runs on one order channel or one customer segment, so the business can judge it on real orders without betting the order book on it.
Proof
The approach on this page comes out of these engagements. Two are businesses that move physical goods, and two are the design patterns the workflows above are built on.
Inventory and operations mapped across every department for a specialty food manufacturer, with the plan assessed and accepted by a funding body. The work of finding out what the operation actually does before deciding what should run it.
Argent supplies furnishing and construction materials and was expanding into new markets with the operation held together by people. Brand strategy, growth marketing and process improvement delivered together, because the demand would have broken the operation on its own.
A safety products business whose reps researched their territories by hand now has that research arriving every morning without anyone triggering it. The pattern behind the coverage workflow above.
Fixed rules produce the classification so the same inputs always give the same answer, and every output waits in an approval queue. The same two design decisions the order workflow above is built on.
Begine Fusion is a Zoho Authorized Partner working across Canada, the United States, the United Kingdom and Nigeria, with a team on the ground in Nigeria since 2013. Every engagement is published in full, including what was measured and what was a design target. The full set is on our work.
Fit
The work pays back when order volume is high enough that consistency matters more than speed on any single order. If that is not where you are, the right column says what to do instead.
Where to start
The same five offers every other client buys, scoped to a business where the same document arrives four hundred times a month. Find the line that sounds like yours.
No price is quoted on this page. A sector is not an offer and does not have a price of its own. What a business here pays follows which of those six lines is the real one and what state the customer and pricing records are in. Working out which line is yours is what the first call is for.
Questions
The seven that decide whether a build is worth starting, answered before the proposal rather than inside it.
Because it needs sensors, a maintenance history and a failure record that most of this market does not have, so it sells a capability the buyer cannot act on for eighteen months. Order handling uses data you already hold, in systems you already run, and can be live in a quarter. Predictive work is a real thing and it belongs after the record is good enough to support it, which is the honest order rather than the impressive one.
Usually not, and we would rather you did not while the customer master still has duplicates in it. The common finding is that the ERP is capable and the records inside it are not, which is a different and much cheaper problem. If the mapping finds a system that genuinely cannot carry the work, we will say so with the reasoning.
No, and that is most of the point. The workflow reads the order in whatever form the customer chose to send it, which means the burden of changing sits with you rather than with them. A distributor who can accept a PDF as cleanly as an EDI message has an advantage over one who keeps asking customers to buy differently.
The auto-release rule, which you write and which is deliberately narrow to begin with: a value ceiling, and every check passing. Every release records which rule allowed it, so after a few weeks you can see exactly what would have happened at a wider setting before you widen it. Most businesses start with clean orders under a low ceiling and move it once.
All of it, which is why the second workflow on this page exists and why the cleaning stage is in the sequence rather than in the assumptions. A checking system built on a customer master with duplicates in it will produce confident wrong answers, and it will produce them faster than a person would have.
A baseline is taken before the build exists: current time per order, the error rate that reaches a customer, the credits raised for order errors and the days spent on the monthly report. Afterwards the same work is measured against it, with the review burden counted openly.
Order entry on one channel is usually four to six weeks once the records underneath it are sound. Getting them sound is the variable, and it runs from a few weeks to a few months depending on what is in there. The mapping at the front is two to three weeks and tells you which it is.
Thirty minutes with someone from commercial, someone from customer service and whoever owns the ERP is enough to name the first workflow, the records it needs underneath it, and the baseline it gets measured against.
