The Problem Shows Up Before People Do
By the time someone complains about a broken process, that process has usually been broken for months. People adapt. They build workarounds. They come in early on Fridays. They stop noticing the extra steps because the extra steps became the job.
I have walked into operations where a team of twelve was doing work that should take two people, and nobody had filed a complaint in two years. The work got done. The pain was invisible. What was visible, if you knew where to look, was friction. Five specific kinds of it.
Here is how to find each one in a week, before you spend a dollar on a solution.
Signal One Re-Keying the Same Data Twice
A logistics coordinator in distribution receives a carrier confirmation by email. She reads the tracking number, opens the warehouse management system, types the tracking number in. That is re-keying. It takes thirty seconds. Multiply by two hundred shipments a day and you have lost nearly two hours of attention to a task that adds zero information to the world.
Re-keying is not just slow. It is where errors enter. A transposed digit in a part number on a manufacturing floor can hold a production order for a day while someone figures out why the pick list does not match inventory.
How to measure it in a week: ask one person in each function to put a tick mark on a sticky note every time they type something they already received in another system or document. Do not explain why. Just count at the end of the week. In a mid-size distribution operation I worked with, one role logged over three hundred tick marks in five days. Nobody had ever counted before.
Signal Two Waiting on a Reply to Move Forward
This one hides inside the word approval. A field operations team cannot close a work order until a supervisor signs off. The supervisor is on site. The sign-off happens by email. The email sits for four hours. The work order ages. The technician moves to the next job but the first job is still technically open, which means it shows up in the backlog, which means someone has to manually reconcile it later.
In finance, I have seen invoice processing workflows where a single line item question to a vendor could pause payment for a week. Not because anyone decided that was acceptable. Because the routing logic assumed someone would follow up and nobody owned the follow-up.
How to measure it in a week: pull the timestamps from whatever system records task state changes. Look for the gap between when a task reaches a waiting state and when it moves again. If you do not have timestamps, have someone note the time when they send a request and the time they get a response for every handoff over five days. Average wait time above two hours in a same-building team is a signal worth taking seriously.
Signal Three Copy-Paste Between Systems
This is re-keying's faster cousin and people are proud of it. They will tell you they have a system. They open the source, they highlight, they switch windows, they paste. It feels efficient. It is not.
A credit analyst in financial services told me she copies customer data from the origination portal into a risk scoring tool into a summary document before every credit committee meeting. Three systems, same data, every time. She had built a keyboard shortcut sequence to do it faster. She had optimized a process that should not exist.
Copy-paste is also a compliance risk in regulated industries. When data moves through a human clipboard, there is no audit trail. You cannot prove what version of a number went into a decision.
How to measure it in a week: screen recording is the cleanest method if your team will accept it for a short observation period. If not, a simple self-report log works. Ask people to note every time they move information from one application to another manually. In manufacturing, I have seen quality teams log forty or more manual transfers per shift across inspection, ERP and reporting tools.
Signal Four Spreadsheets Acting as Queues
When a spreadsheet is the source of truth for who does what next, you have a queue problem. Spreadsheets are fine for analysis. They are bad at concurrency, bad at state management and bad at notifying anyone of anything without a human refreshing them.
I have seen this in field operations where a shared sheet tracked open service tickets. Three dispatchers edited it simultaneously. Rows got overwritten. Assignments disappeared. A technician drove forty minutes to a job that had already been reassigned. The dispatcher who made the error did not know it had happened until the technician called.
In distribution, I have seen inbound receiving queues managed in spreadsheets where the formula that calculated available dock doors was wrong for six months before anyone noticed. The team had adapted by eyeballing it.
How to measure it in a week: count the number of spreadsheets your team opens every day that contain a list of work to be done or a status of work in progress. If the number is above three per function, you have spreadsheet queues. Then ask how many people edit each one. More than one editor on a live queue is a risk that compounds daily.
Signal Five The Friday Backlog
Every operation has a rhythm. Work arrives, work gets processed, work closes. When that rhythm has a leak, the leak drains into Friday afternoon. The Friday backlog is not a scheduling problem. It is a friction accumulation problem. Every re-keying event, every approval wait, every copy-paste and every spreadsheet queue contributes a small delay that compounds across the week.
The Friday backlog is also where quality drops. People rush. Approvals get rubber-stamped. Exceptions get deferred to Monday, which means Monday starts behind.
In a manufacturing finance team I worked with, month-end close had a predictable twelve-hour crunch on the last Friday. When we mapped the week backward, we found that thirty percent of the crunch was caused by three re-keying steps that could have been eliminated. The crunch felt like a capacity problem. It was a friction problem.
How to measure it in a week: compare task completion volume by day of week over the last month. If Friday volume is more than twenty-five percent higher than Tuesday volume and the work arriving each day is roughly equal, you have a backlog accumulation pattern. Then look at error rates and exception flags by day. If they spike on Friday, the rush is real and the friction is the cause.
What to Do With What You Find
Run this audit before you buy anything or build anything. One week of honest observation will tell you more than a consultant's process map.
Rank the five signals by frequency and by the cost of the errors they produce, not just the time they consume. Re-keying that causes a transposed part number is more expensive than copy-paste that causes a formatting inconsistency. Prioritize where friction meets consequence.
Then ask a simple question for each signal: if a system could do this step reliably and leave a record, what would the person do with that time? The answer to that question is where you find the real case for change. Not headcount reduction. Not cost cutting. Speed gained on the work that actually requires a human, and manual work removed from the work that does not.
AI agents can close most of these gaps in production environments today. But the agents that work are the ones built against a clear map of where the friction actually lives. This audit is that map. Do it first.