Forty suppliers, three thousand SKUs, one deadline
Wholesale is a business of small percentages at large volumes, which makes it unusually sensitive to exactly the errors nobody has time to look for. A two percent costing error is survivable in a business with forty percent margins. In distribution it can be most of the profit.
The business described here is a mid-sized distributor: forty active suppliers, around three thousand SKUs, three hundred purchase lines a month, a warehouse with two people on receiving and a finance team of two. It is a shape that recurs constantly, and it sits exactly at the point where the informal methods stop working — past the size where one person can hold the price list in their head, below the size where anyone has built a system for it.
What follows is the monthly routine that holds at that size, and honest numbers about what it costs and returns.
The shape of the problem
Work in stock control splits into two kinds with very different economics. There is work that scales with document volume — reading invoices, keying quantities, filing delivery notes. And there is work that scales with the number of exceptions — investigating a shortfall, deciding whether a price change should be passed on, chasing a credit.
The second kind is the job. The first kind is preparation, and in most distributors it consumes so much of the month that the second kind happens only when something goes visibly wrong. That is why price rises get discovered at year end and credits get discovered never.
Everything below is organised around moving the first kind out of the way so the second kind can happen at all. It is a sequencing change more than a technology one, though it needs the document work to be fast enough to stop being the constraint.
Who does what
Three roles, and the boundaries matter more than the headcount. The most common structural failure in wholesale is splitting price and value between purchasing and finance, so that nobody owns the question of whether the two agree.
| Role | Owns | Does not own |
|---|---|---|
| Receiving | Counting at the door; the GRN is a real number | Deciding what a difference means |
| Purchasing | Price changes, supplier conversations, claims | Stock valuation |
| Finance | Reconciliation, accruals, margin reporting | Chasing individual deliveries |
The first row carries the whole structure. If receiving records a signature rather than a count, everything below it becomes guesswork — and no amount of finance sophistication can recover a number that was never written down.
The monthly cycle
Throughout the month — extract as documents arrive
Invoices and delivery notes converted the week they land. Credits filed against the invoice they correct, not into a folder.
Day 1 — completeness
Every supplier used, every invoice present, every delivery with a note. The absences are the expensive part and they are found first.
Day 1–2 — reconcile by value
Invoiced versus received by product, working down from the largest. Materiality threshold applied and written down.
Day 2 — price review
Unit price by product against last month. Changes go to purchasing the same day, while selling prices can still move.
Day 3 — close and hand over
Accrue received-not-invoiced, record unresolved differences with amounts, note the freight treatment applied.
The first line is the design. Everything below it assumes the data exists — and where it does not, each subsequent day absorbs the document work as well as its own, which is exactly how a three-day routine becomes a fortnight and then becomes optional.
Where the hours actually go
Measured over a month for this business — forty suppliers, roughly three hundred purchase lines, a hundred and twenty deliveries. The absolute numbers vary; the ratio between the columns is what matters.
| Activity | Keyed by hand | Extracted |
|---|---|---|
| Purchase lines into data | 25–35 hrs | 2–3 hrs |
| Delivery notes into data | 8–10 hrs | 1–2 hrs |
| Invoiced vs received comparison | 6–8 hrs | Under 1 hr |
| Price change review | Rarely done | 30 min |
| Credits identified and claimed | Usually missed | Part of the reconciliation |
| Answering a margin question | Hours, unpredictable | A filter |
Two rows have no left-hand number, and those are the interesting ones. Price review and credit recovery are not slow when done manually — they are simply not done, because the first row consumed the month. That is a capability difference rather than a time saving, and it is where the money is.
The ROI, honestly
Roughly 30 to 45 hours a month of document handling. At a loaded internal cost of £25 to £35 an hour that is £800 to £1,500 monthly, and the software cost is a rounding error against it.
But the honest version includes two things vendors leave out. First, the saving is not the whole amount: investigation, supplier conversations and judgement remain, and they are perhaps a quarter of the total. Second, and much larger, is what the freed capacity finds — recovered credits, price rises caught inside the month, cost errors corrected before they compound. Those are irregular, which makes them impossible to budget and easy to leave out of a business case where they belong.
The most defensible framing internally is not hourly savings. It is that a monthly reconciliation becomes possible at all, and with it a gross margin figure that survives being questioned. That argument holds up in a way an hourly-rate calculation sometimes does not.
The price-watch habit
If a distributor adopts exactly one habit from this page, it should be this: once a month, look at unit price by product against the previous month, and send the changes to purchasing the same day.
Supplier price rises are correctly invoiced and correctly paid. Nothing in an approval process catches them, because there is nothing wrong with the transaction. What breaks is everything downstream that assumed the old cost — the selling price, the standard cost, the margin in a spreadsheet somebody built two years ago.
The value of catching it within the month is not the accounting. It is that a selling price can still be moved. A cost increase found in month one is a pricing decision; the same increase found at year end is eleven months of margin that was given away, and no reconciliation can recover it.
Which suppliers actually need watching
Not all forty equally. Attention should follow risk, and in wholesale the risk concentrates in predictable places.
High value per line. Obvious, and the one everybody already covers.
Frequent small deliveries. Less obvious and more dangerous: many small shortfalls each below any sensible threshold, adding to something material across a year.
Suppliers who changed invoicing software. New layouts are where unit conventions quietly change — the same product suddenly priced per case rather than per unit.
Suppliers whose deliveries arrive at awkward times. The eight-in-the-morning delivery gets signed rather than counted, and that is a data quality problem with a staffing cause.
Once a few months of extracted data exists, this stops being intuition. The differences cluster, and the clusters name the suppliers.
Seasonal peaks, where it matters most and happens least
Most wholesalers have a peak, and the peak has a predictable pattern: volume triples, receiving gets rushed, counting slips, and the monthly reconciliation is skipped because there is no time. Which means the months carrying the most value are the least controlled.
It is also when errors are most likely. Rushed receiving produces uncounted deliveries. New temporary staff make unit mistakes. Suppliers under pressure send partial shipments and invoice for the whole. Every failure mode on the stock value page becomes more likely at once.
Automating the extraction is what makes the peak survivable, because the part that scales with volume stops scaling. The reconciliation still takes judgement, but judgement on a peak month is a bounded task rather than an impossible one — and skipping it during peak is precisely the decision that costs most.
Cycle counting instead of the annual shutdown
The full annual count exists in many distributors because nothing else was trusted. It is disruptive, expensive, and produces a difference at exactly the moment nothing can be done about any of its causes.
Once the purchase side is reconciled monthly, the case for it weakens considerably. Cycle counting — a portion of SKUs each month, weighted toward high value and high movement — gives more current information at a fraction of the disruption, and the differences it finds are recent enough to investigate.
It also changes what a count means. Against unreconciled purchases, a count difference is an unexplained total. Against a settled purchase side, it is a physical question about a specific set of products — which is the only version anyone can act on.
When margin becomes a usable number
The outcome that matters most is hard to put in a business case: gross margin becomes something people quote. In distributors where stock costing is unreliable, margin quietly stops being used — nobody says so, but decisions start being made on revenue and cash instead.
That is a worse problem than it sounds. Pricing decisions made without trustworthy margin are guesses. Supplier negotiations without accurate landed cost are conducted from the weaker side. Product range decisions get made on volume because volume is the only number anyone believes.
Reconciling purchases does not fix valuation policy or physical control. What it does is remove the documentary noise, so the margin reported reflects decisions made rather than documents unread — and once that is true for a few months, the number starts being used again.
What it changes for the buying side
Finance benefits are the easy part of this case. The larger prize sits with purchasing, and it is rarely argued for because the data has never existed to argue with.
A buyer with a year of line-level purchase history can answer questions that were previously matters of impression. What did we actually pay for this product across all suppliers? Which supplier is genuinely cheapest once carriage is included? How often has this supplier raised prices, and by how much each time? Is the volume discount we negotiated actually being applied?
That last one is worth dwelling on. Negotiated terms are agreed in a meeting and applied by an invoicing system, and the gap between the two is real. A rebate that stopped being applied in April is invisible in a total and obvious in a price history — and it is the kind of finding that pays for the whole exercise in a single conversation.
It also changes the balance in a negotiation. Arriving with «your prices seem to have gone up» is a position. Arriving with the dates, the amounts and the products is a different meeting, and suppliers respond to it differently.
What changes for the team
Worth addressing directly, because it is the unspoken question whenever data entry is removed. In practice the same team covers a business that has grown, which is how most distributors experience it — the alternative was hiring, not staying still.
The work that remains is more demanding and more visible. Investigating a recurring shortfall, deciding whether a cost increase should be passed on, arguing a claim with a supplier — these need judgement and context, and they are the parts where a finance function is obviously earning its place rather than producing paperwork.
There is a genuine risk worth naming. Someone who keyed every purchase line knew things about the supplier base that a reviewer of a clean file does not — which products always arrive short, which supplier's invoices are unreliable. That knowledge is real and it does fade. The mitigation is the exception list, which pushes attention onto exactly the rows that would previously have been noticed by hand.
Receiving is where the change is felt most positively, and it is usually unexpected. A team whose counts have never visibly mattered starts counting properly quite quickly once a shortfall they recorded turns into a credit note and somebody says thank you.
Objections you will hear internally
"Our suppliers' invoices are all different." True, and it is the case for meaning-based extraction rather than against it. Settle it by converting the three most awkward invoices you have rather than by discussing it.
"We would need to clean up the product master first." The objection that costs most, because it sounds responsible and postpones everything. Reconcile the products you can map today and let the mapping grow — the master gets cleaner as a by-product rather than as a prerequisite.
"The warehouse will never count properly." Often said and usually unfair. Counting collapses when the numbers are visibly never used; it recovers surprisingly fast when a shortfall found at the door turns into a credit note somebody thanks them for.
"We tried an inventory system and it did not stick." The most informative objection. Ask what broke — it is usually that the system needed data nobody had time to enter, which is the problem being removed rather than repeated.
Starting without making it a project
The version that fails begins with all forty suppliers and a product master cleanse. The version that works begins with one supplier next month.
Pick the messiest, not the easiest — the one with handwritten delivery notes and a layout that changed last year. The tidy supplier was never the problem, and proving the method on them proves nothing. Extract a month, reconcile it against the receipts, and compare the result with what the current process produced.
If it agrees, extend to five suppliers the following month, then to the rest. If it disagrees, you have learned something valuable about the current process at the cost of an afternoon. Either outcome is worth having, which is what makes it a safe first step rather than a commitment.
By month three the routine is usually settled, and the remaining suppliers are a formality. Total elapsed effort is a few hours a month, and at no point does the close depend on something untested.
Adding the next supplier — and the next hundred SKUs
The test of any stock process is what happens when the business grows, and for most distributors the honest answer is that it degrades slightly each time without anyone noticing until it is obviously broken.
Because extraction needs no template per supplier, adding one means uploading their documents. The work that grows is the mapping — their product codes to your SKUs, their units to your units — and that is a one-off per product rather than per invoice, so it compounds downward rather than upward.
What genuinely does not scale is investigation, and it should not be automated away. A hundred more SKUs means slightly more exceptions, and exceptions are where the value is. The aim is not to remove that work but to stop it being crowded out by data entry.
Where this stops
It gets supplier documents in — line-level, checked, comparable. It is not an inventory system: quantities, locations, replenishment and valuation stay in your ERP or stock software.
It covers the purchase side of stock. Sales, wastage and physical loss need their own controls, and reconciling purchases narrows how much of a count difference they have to explain rather than explaining it for them.
And it cannot supply a count that was never made. If deliveries are signed without checking, the received record does not exist and no extraction can invent it — see goods received notes.
Processing runs on EU-hosted infrastructure over TLS, the original file is deleted immediately after extraction, and documents are never used to train AI models — the security page has the detail.
Start with your messiest supplier
Extract one real invoice free — no registration — and compare the lines with what your current process produces.