The Cost of a Process: how to put a yearly number on the work eating your team's week

Why you cannot name the line item

Ask your finance lead what it costs to answer supplier queries about payment dates. Or to re-key orders from email into the order system. Or to chase sign-off on holiday requests. You will not get a number. Not because nobody knows the work exists, but because the work does not exist as itself anywhere in the accounts.

It is spread across people who were hired to do something else. The credit controller does some of it. The office manager does some. A director does a little on Friday afternoons. Each of them books their time to a salary line, and the salary line says “operations” or “finance”. The process itself has no cost centre, no budget and no owner. It appears in nobody’s job description, so nobody defends it and nobody attacks it.

This is why proposals to remove recurring work stall. Someone says “this takes ages” and someone else says “how much is ages?” and the conversation ends. A process with no number cannot be compared with anything, so it cannot be ranked, so it never reaches the top of the list.

The method below fixes that. It needs one person with access to the team, a stopwatch and a spreadsheet. It produces one figure: what the process costs your business each year, with the method shown, so anyone in the room can check it.

The five inputs and where to find each

Monthly volume

How many times does the process run in a month? Count items, not hours: invoices queried, orders re-keyed, requests approved. Find the count in the system where the work lands. An inbox folder, a spreadsheet tab, a queue in your order system. If the work lives in no system, keep a tally for a full month. Do not ask people to estimate volume from memory. Memory rounds busy weeks up and quiet weeks down, and you have no way of knowing which happened.

Minutes per item

Time it. Do not ask. Sit with the person who does the work, or have them time themselves over a run of items, and take the median. Timed figures are usually a long way from guessed ones, in both directions. A person who finds the work tedious will overstate it. A person who has done it for years will understate it, because they no longer notice the parts that take time. Include the whole item: finding the record, doing the work, filing the result, and writing the note that nobody reads.

Loaded hourly cost

Salary alone understates the cost of an hour. Loaded cost adds employer taxes, pension contributions, benefits, and a share of the desk, equipment and software the person uses. Your accountant or payroll provider can give you the loaded figure for each role, and many already calculate it for job costing. Use the cost of the person actually doing the work, not an average across the business. If a director does part of it, that part is priced at a director’s loaded rate.

Downstream delay

Something waits while this work sits in a queue. Cash waits for an invoice query to be resolved. A delivery waits for an order to be keyed. A new starter waits for a laptop while an approval sits in an inbox. Record two things: what waits, and how long it waits on average. The wait is usually visible in the timestamps of the systems the item passes through.

Not every delay converts cleanly into money. Where it does, as with cash, record it in money. Where it does not, record it in days and carry it alongside the total. A figure in days next to a figure in pounds is more honest than a guess in pounds.

Error rate and the cost of fixing an error

Some items are done wrong. Wrong amount keyed, wrong customer emailed, wrong date approved. Find the rate by sampling: take a run of completed items and check each one. Then time what fixing one takes, end to end, including the apology. Errors cost more than they look because fixing one takes longer than doing it right, and the people doing the fixing are usually more senior than the people who made the mistake.

The worksheet

One page. Five inputs, arithmetic down the side, one yearly figure near the bottom, and the delay carried alongside it.

The numbers below are illustrative inputs to show the arithmetic, not a measurement of any business. The process in the example is handling supplier queries about payment status, a common piece of work in the finance team of a mid-sized firm. Replace every figure with your own.

In the illustration, direct time on line E is the biggest single item, but it is not the whole cost. Everything from line F down is what most first attempts leave out. The next section explains each of those.

Two rules for filling in your own version. Every input in the right-hand column gets a note saying how you got it: “timed across a run of completed items”, “sampled from last month’s completed queue”, “payroll loaded rate for the role”. And nothing in that column comes from memory. If you have not timed it or sampled it, leave it blank until you have. A blank is better than a guess because a blank cannot be quoted back at you.

The three undercounts

The manager’s review time

Somebody checks the work. They sign off the payment run, glance over the keyed orders, or approve the approvals. That time is rarely counted because the manager does not think of it as doing the process. They think of it as managing. It is still an hour of a more expensive person, spent because the process exists. Time it the same way you timed the item, and price it at the manager’s loaded rate, not the clerk’s. Lines H to J in the worksheet.

The interruption cost

Recurring work rarely arrives in a tidy block. A query lands, the person switches to it, resolves it, and switches back to whatever they were doing. The switch back is not free. There is a gap between finishing the query and being productive again on the previous task, while they work out where they had got to. Time that gap the same way: from the moment the item is closed to the moment the earlier work resumes properly. Multiply by volume. Lines F and G. The figure surprises most people who measure it, because each gap is small and there are a great many of them.

The customer-facing delay nobody bills for

Line Q. A supplier waiting for a payment answer, a customer waiting for a quote, a new starter waiting for access. Nobody sends an invoice for the wait, so it appears in no ledger. It still has a cost. Cash arrives later. Quotes are lost to someone who answered first. The new starter sits idle on salary. Where you have a trustworthy way to convert it, such as your own cost of financing for delayed cash, convert it and add it to the total. Where you do not, leave it in days and put it next to the total on the same page. A board can weigh a figure in days against a figure in pounds. It cannot weigh something that is not written down.

The ranking rule

Once you have costed more than one process, rank them. The rule is yearly cost divided by effort to remove. Effort means weeks of your own people’s time plus whatever you would pay anyone outside. A process that costs a lot and takes little to remove goes first. That sounds obvious, and it is, but the temptation is to rank by cost alone, and that is wrong often enough to deserve three paragraphs.

Dependencies. The costliest process may sit downstream of another one. If order re-keying exists because quotes are produced in a spreadsheet that does not talk to the order system, removing the re-keying without fixing the quoting moves the problem rather than solving it. Draw the chain and start at the end you can reach.

Data access. Some work runs on information that lives in a system your business can read from cleanly. Other work runs on information in a locked supplier portal, a shared drive of scanned PDFs, or one person’s memory. The second kind costs much more effort to remove regardless of what it costs to keep. Score it honestly in the denominator.

Who owns the process. A process with a named owner who wants it gone will be removed. A process spread across several departments with no owner will be argued about, and the argument will outlast the budget. Before you rank, write a name against each process. If you cannot, that process drops down the list whatever its cost.

A second-ranked process with a willing owner and clean data often pays back before the first-ranked one has cleared its dependencies. Do the one that will actually finish.

What to do with the number

Three uses.

First, take it into the free AI Readiness Assessment. That session scores your readiness and costs one process with you. If you arrive with the worksheet already filled in, the time goes on what to remove and in what order rather than on measurement.

Second, use it to price any claim anyone makes to you. When a vendor says their software will save your team hours, ask which line of the worksheet those hours come from and what happens to lines G, J and N once the software is in place. Does review time go up or down? Does the error rate change, and who fixes the new kind of error? A claim that cannot be placed on a specific line of the worksheet is not a saving. It is a hope. That applies to our claims exactly as much as to anyone else’s, and you should hold us to it.

Third, keep it. Re-run the worksheet a quarter after any change, with volume, minutes and error rate measured again rather than assumed, and you have a before-and-after figure with the method shown. That is the only kind of result worth putting in front of a board.

The number is yours whether or not you hire anyone. That is the point of it.