Five specific places spreadsheet purchasing loses money: duplicate orders, approval queues, committed spend, undocumented process, and lost approvals.
Three departments. Ops keeps a purchase log in Google Sheets. The studio keeps one in Excel on the shared drive. The workshop keeps one that is mostly accurate and partly in the workshop manager's head. Every one of those records is honest, maintained by someone competent, and reconciles to its own receipts.
Now ask the question that matters on the 14th of the month: what have we committed so far? The answer takes three phone calls, arrives the following morning, and is out of date by the time it lands, because two more orders went out while everyone was checking.
That is the wall. It is not a headcount number. Purchasing does not break at fifty employees or at a revenue threshold; it breaks the moment the number of people who can independently commit company money exceeds the number of people who can see all the commitments. For most companies that happens at the second or third department, often well under thirty people. Before that, one person knows everything because one person does everything. After that, nobody does.
What follows are the five places the money actually leaves through that gap. They are not carelessness. They are what a recording tool does when you ask it to do a preventing job.
Leak one: the order that gets placed twice
A project needs 40 boxes of a special-order flooring at $86 a box - $3,440. On Monday the site lead emails the supplier and places it. On Thursday the project manager, who has heard nothing and has a delivery date to hit, emails the same supplier and places it again.
Both people did the right thing by the information they had. The supplier did the right thing too: two orders, two confirmations, two deliveries. The company has now committed $6,880 against a $3,440 need. Because the colour is special-order, the return carries a restocking charge - say 10%, $688 - and that is the cheerful version, where the supplier takes it back at all.
The failure here is not attention. It is that a sent email is not a queryable record. The Monday order exists in one person's Sent folder. There is no state on it that anyone else can read, no way for the second person to check before committing, and nothing that fires when the same item goes to the same vendor twice in a week. A shared spreadsheet only helps if both people update it before they order, which is precisely the moment they are least likely to.
And duplicates are usually discovered at invoice time, weeks later - not when the second order is placed, but when the second invoice arrives and someone notices the reference looks familiar. By then the money has moved. (Purchase orders →)
Leak two: the approver who became a queue
A specialist contractor routes every purchase request to the managing director. Forty to sixty a week. Roughly two thirds of them are under $500 - consumables, small tools, fuel, the parts a job needs to keep moving.
Ask how many were rejected last year and the answer is none. Not few. None.
A rule that has never once produced a rejection is not a control. It is a delay with a signature on the end. And the cost is not only the waiting. It is what the waiting does to the person doing the approving: at fifty decisions a week, none of which has ever needed a second look, the approval stops being a decision and becomes a keystroke. The $40,000 request arrives in the same inbox, in the same format, behind eleven requests for drill bits, and gets the same four seconds of attention.
Then come the workarounds, which are the expensive part. When the official route reliably takes two days, people who have a job to finish stop using it. They buy first and get approval after, or they put it on a card, or they ask a colleague to add it to an order that is already going out. None of that is defiance. It is friction, routed around. The spend is still happening - it has simply left the process, and with it any chance of seeing it before the statement. (Approval workflows → - on the Advanced plan)
Leak three: the budget you have already blown
There are three different numbers hiding behind "what have we spent?", and a spreadsheet-and-accounting stack only holds two of them.
Paid is what has left the bank. Invoiced is what has been billed but may not have cleared. Committed is what you are contractually on the hook for the moment an order is approved, whether or not anyone has invoiced you yet.
Take a $60,000 marketing budget, one third of the way through the year. The accounting system shows $31,000 invoiced. That reads as slightly ahead of schedule and entirely fine. But there is $14,000 of approved work that has not been invoiced yet, and $9,000 sitting in drafts about to be approved. The real remaining figure is $6,000, not $29,000.
Nobody in this scenario is careless. The number they are looking at is accurate. It is just answering a different question from the one being asked, and it is the only number the tooling can produce. Committed spend is the one figure that would have stopped the overspend while it was still stoppable, and it is exactly the figure a spreadsheet cannot maintain, because maintaining it means every draft order in the company updating a central total the second it is raised.
Next week's article takes this one apart properly. (Budgets →)
Leak four: the process your new hire invents in four seconds
A project coordinator starts on a Monday. On Tuesday she needs a software licence. She asks the nearest person how purchasing works here.
The nearest person is busy, means well, and says: "Just put it on the company card and expense it."
That is now the purchasing process. It was invented in four seconds by someone with no authority to invent one, and it will propagate, because the next new starter will ask her.
Every company in this position has a policy document. The policy document is not the problem. The problem is that it does not survive contact with a busy Tuesday, and that an undocumented process is not an absent process - it is a process being continuously reinvented by whoever happens to be standing nearest. The version that wins is not the correct one. It is the one that got said out loud most recently.
Leak five: "who approved this?" as a search problem
The last leak costs the least money and the most credibility.
Six months after the fact, a supplier disputes an order, or an accountant queries a line, or an insurer asks for the file. The question is simple: who approved this, when, and on what basis? The answer is in an email. Possibly in a reply to a forwarded thread. Possibly from someone who has left. Possibly it was a verbal "yeah, go ahead" in a corridor that someone later summarised in an email that no longer exists.
An email approval is a decision with no durable link to the thing it decided. It lives in a different system from the order, it is searchable only by whoever owns the mailbox, and it degrades the moment that person's account is closed. Answering the question means searching inboxes rather than opening a record - and "we think it was probably Dave" is not an answer you want to give twice.
This is the leak that shows up in an audit, in a dispute, and in due diligence. It is also the one people discover last, because until someone asks, nothing appears to be wrong.
Three out of five is a tooling problem
Read back through the five. If you recognise one, that is a bad week. If you recognise three or more, the pattern is not discipline.
It is worth being precise about why, because the instinctive fixes both fail. A better spreadsheet fails because every one of these leaks happens in the gap between a decision being made and that decision becoming visible to anyone else - and a spreadsheet contains only what someone remembered to enter after the fact. A sterner policy fails because none of the five was caused by someone deciding to ignore a policy.
What closes the gap is a system in which the commitment and the record are the same object. Raising the order is what creates the record, so there is no moment where a commitment exists but is invisible. That is the whole idea. Everything else - routing, budget checks, matching, audit trail - is what becomes possible once it is true.
That is what Foxpedite is: purchase orders, budgets, vendors and approvals in one place, with a full change history on every record. It starts at $30 per seat per month on the Regular plan, minimum five users, with a 14-day free trial and no credit card to begin. (Pricing →)
Next week's article covers the approval half in detail: how to set thresholds that route on risk rather than seniority, and why your budget check is probably measuring the wrong number. (Approval workflows that don't end at the owner's desk →)
The figures in this article are worked illustrations, not measured results or customer data.