The Hidden Hours Audit: how to find where your business is losing time

Why can't anyone in your company tell you how long anything takes?

Because the work is fragmented. Nobody sits down and does invoicing for two hours. They do eleven minutes of it, answer a Viber message, do six minutes more, then walk to the bodega to check whether the stock actually went out. Ask a branch manager how long invoicing takes and the number will be wrong by half, in whichever direction makes him look better. Not dishonestly. He has never counted.

Asana's Anatomy of Work index, surveying 9,615 knowledge workers across six countries, found 58% of work time going to what they called work about work: coordination, searching for information, chasing approvals, updating status. Not the job. The scaffolding around the job.

In a business with physical operations that scaffolding is worse, because the information lives in three places that do not agree. The delivery happened in the truck. The record of it is on a carbon-copy receipt. The invoice is in a spreadsheet on one laptop.

So you cannot ask people. You have to look at the artifacts.

Signal one: how many times does the same data get typed?

Take one transaction, start to finish, and follow it physically. Count every time a human types or writes information that already exists somewhere else. The threshold is three: once at the point of capture, once for a physical document if you genuinely need one, once into your system of record. More than three and you are paying people to be a data bus.

Every re-key is also a place where a digit changes. That is why this signal predicts your error rate as well as your labor cost, and why it is the one to count first.

Signal two: how many hours go into making two numbers agree?

Reconciliation is the purest form of hidden work, because it produces nothing. At the end you have the same information you started with, only now you trust it. Count the monthly hours going into bank reconciliation, stock counts and their investigations, matching delivery receipts to invoices, matching supplier invoices to purchase orders, and the branch-versus-head-office argument about what was actually sent.

Include the investigation time, not just the counting time. The count takes two hours. Finding out why it is off by eleven units takes two days.

The threshold is 20 hours a month. Below that it is a cost of doing business. Above it you do not have a reconciliation problem, you have a capture problem, and you are paying for it at the back end every month, forever.

Signal three: how much time goes into chasing information you already have?

Open the Viber or WhatsApp group your operations actually run through and read a week of it. Count the messages that are somebody asking for a number the business already possesses. How much stock is left at the branch. Whether the payment came in. Where the delivery got to. The threshold is 30% of messages.

Each of those is one person interrupting another to retrieve a fact that exists but is not visible. If nearly a third of your operational chat is retrieval rather than decision, the information is there and simply not reaching the people who need it.

This is usually the cheapest thing on the list to fix and the one with the fastest felt improvement, because it gives people their attention back.

Signal four: what share of transactions need a human decision?

Pull last month's transactions and sort them into two piles: the ones that went through under a standard rule, and the ones where somebody had to decide something. The threshold is 15%. If more than one in seven needs a judgment call, you do not have a documented process. You have a person with a process-shaped job.

A special price for a long-standing customer. A payment term that is not on any document. A delivery that goes out before the payment clears because the owner knows the family. None of these are wrong, and they are often why the business has customers at all.

But every one is a rule that lives in a head rather than a system, and every one means that person cannot take a week off. That is fine at 30 staff. At 200 it is a ceiling, and you can usually feel exactly where it is.

Signal five: how much of the process exists only in someone's head?

Pick your most critical operational process. Ask the person who runs it to write down how it works, in enough detail that a competent new hire could run it on Monday. Give them an hour. Then read what they wrote and count the gaps: every step that says "then I check with Rey" or "depends on the customer" or "you just know which ones."

The threshold here is not a percentage. It is one. If a single person leaving would stop you invoicing, or ordering stock, or paying staff, that is the risk to fix before anything else. It costs nothing to start, and it has nothing to do with technology.

How do you run this audit in a week?

You do not need software and you do not need a consultant to do the counting. You need one week and somebody willing to be boring about it. At the end you have five numbers that are counts rather than estimates.

• ·Monday: pick one process end to end. Order to cash is usually the right one. Follow a single transaction through it physically and count the re-keys.
• ·Tuesday: pull last month's reconciliation hours. Ask the people who do it, then add 40%, because they will underestimate.
• ·Wednesday: read one week of operational chat. Tally retrieval messages against decision messages.
• ·Thursday: sort last month's transactions into standard and judgment-call. Get the ratio.
• ·Friday: run the one-hour documentation test on your most critical process. Count the gaps.

What do you do with the five numbers?

Convert them to money before anything else, because that is what tells you whether to act. Take the hours, multiply by the loaded cost of the person doing them, annualize. Then add the error cost: credit notes, short deliveries, duplicate payments, write-offs from stock that walked.

Most businesses find the error cost is larger than the labor cost, and nobody has ever put the two on one page together.

Now the decision is actually a decision. A process costing fifteen thousand a year is one to leave alone. Fix how it is captured and move on. A process costing a hundred and forty thousand is a business case, and you can go and get quotes knowing what a sensible one looks like.

Which of these should you not fix?

Some of the manual work in your business is load-bearing. The handshake pricing that keeps your top ten customers is not inefficiency, it is the relationship. Automate it and you will find out the hard way what it was doing.

This is where most operations content stops being useful. The point of the audit is not to find things to automate. It is to find which problems are big enough to be worth the disruption of changing them, and which ones you should deliberately keep doing by hand.

It is also why the order matters. Find where the work leaks, fix and document the process, then automate what is left. Automating a process nobody has written down does not remove the problem. It makes the same problem run faster, and now it runs somewhere you cannot see it.

Common questions

Sources

• Parseur and QuestionPro, Manual Data Entry Report. Survey of 500 US professionals across operations, finance, admin, IT and customer support. Fielded July 2025, published 19 August 2026.
• Asana, Anatomy of Work Global Index 2023. Fielded by GlobalWebIndex. 9,615 knowledge workers across the US, UK, Australia, France, Germany and Japan. Published 8 March 2023.