Operational Inefficiencies in Business: How to Find and Fix Them
The hours recorded against a task often leave out the longest part of the job: waiting. An update that takes an afternoon can still take days to reach a customer when it passes through approvals, queues and several different systems. People compensate with reminders, private messages and duplicated records until the workaround begins to look like the normal process. Nobody planned it that way, but everybody has learned to live with it.
That is why operational inefficiency is easy to misdiagnose. The visible symptom might be an overflowing task board, a founder who is always being asked to approve things or a customer who keeps requesting updates. The cause might be somewhere else entirely. Before buying another tool or asking everyone to work faster, find out where the work spends its time and what makes it stop.
The work takes an hour. Why does the customer wait three days?
The work of updating a website may take an hour. Getting it approved and published can take several days. A time sheet captures the hour; the customer experiences the entire wait. If your reports only count work completed, those days of inactivity are easy to miss.
Track both the time someone spent on the request and the time from the customer’s first message to delivery. Shortening that second stretch may help more than asking anyone to work faster.
APQC’s definition of cycle time includes the elapsed time from the beginning to the end of a process. That is a useful way to investigate work that moves through several people or systems. Note when a request arrived, when somebody began working on it, when it stopped and what finally allowed it to continue. Often the delays are unremarkable on their own. Added together, they dominate the schedule.
Follow a real request through the company
A process diagram shows how work is supposed to move. To find out what happens in practice, choose a few requests that were delayed and follow their records from start to finish. The original messages may show an approval sitting unanswered or a missing detail that caused the work to start again. Those records are more reliable than asking everyone to describe the process from memory.
Start with the thing the customer asked for, then follow the work until it reached them. Where did the brief change? Which fields were copied into another system? Who could approve the next step, and did they know it was waiting? If the same information was typed into a form, a spreadsheet and a CRM record, find out which copy people trust when the details disagree.
Don’t assume every handoff is wasteful. A security review may prevent a serious error. A designer reviewing a page before launch may catch something a developer reasonably missed. The question is what each handoff contributes. When nobody can explain why a particular approval exists, it deserves a closer look.
Find the point everybody is waiting on
Producing drafts faster won’t help if every finished draft still waits for the same person to approve it. The queue simply grows. Find out whether that final approval is protecting the work from a real risk or continuing out of habit.
This is the practical value of thinking in constraints. The Lean Enterprise Institute’s overview of operations describes the cost of waiting and unfinished work. You don’t need to adopt a complete management method to borrow the observation. Find the queue that keeps filling up, then ask why it cannot empty.
The answer could be capacity. Perhaps there is genuinely too much work for one specialist. It could also be a rule nobody has questioned. A manager might approve every routine change because a past mistake made the team nervous. If only unusual or expensive changes really need that approval, documenting the boundary could give the manager time back while keeping the important control.
Look at a handful of delayed jobs before changing the process. If most stopped for the same missing decision, start there. If they stalled for unrelated reasons, one new rule probably won’t fix everything.
Sometimes the team really does need more capacity
Businesses can go too far in blaming process. If twelve requests arrive every week and the team can finish eight, the backlog will grow by roughly four a week. Clearer handoffs may help a little, but they cannot make that arithmetic disappear. The company has to reduce incoming work, increase its capacity, change what it promises or address the constraint that limits output.
Track the number of requests arriving, the number completed and how long the oldest open item has been waiting. That last measure matters because a team can look productive while quietly abandoning difficult jobs at the bottom of a queue. If new work always jumps ahead, the total completion count may look healthy while individual customers have a terrible experience.
Equally, more staff won’t cure a process where nobody is authorized to make a routine decision. Hiring people into a broken approval chain can simply give the company more people who are waiting.
Approval is supposed to settle a question
Some businesses accumulate reviewers the way browsers accumulate tabs. An additional person gets copied into an email because they once asked to be kept informed. Soon they are treated as an approver, even though nobody has agreed what they are supposed to decide. The request goes back and forth because the team cannot tell which comments are advice and which are instructions.
It’s worth separating those roles explicitly. Who owns the decision? Who can contribute specialist advice? Who implements it once it has been made? McKinsey’s discussion of decision rights uses the DARE framework to make a similar distinction. You don’t need the terminology on every task, but you do need the authority to be clear.
Take a content approval process. Legal might need to review a regulated claim, while an ordinary sentence edit can be approved by the editor. Treating both as equally risky wastes time and may cause reviewers to pay less attention to the decisions that genuinely need them. Sensible speed comes from distinguishing the two, not from eliminating reviews altogether.
When automation makes the mess travel faster
Connecting two tools feels productive. A form submission creates a contact, a CRM record opens a project and a notification tells the delivery team to begin. It works beautifully in a demonstration where every field is complete and every customer is new. Actual businesses have customers who use a second email address, incomplete requests and projects that aren’t approved yet.
If the process has no answer for those cases, automation creates problems at machine speed. A duplicate customer record becomes a duplicate project. An empty delivery requirement becomes a task somebody has to chase. A failed API request might even succeed remotely and then get repeated because the sender never received the confirmation.
Write down the conditions that must be true before the next step can happen. Check whether the customer already exists. Establish which system owns the current status. Decide what an exception should do and which person reviews it. Then connect the tools. Our guide to AI automation across marketing, sales and operations goes further into those handoffs and the limits of model-based decisions.
There are good opportunities for automation here. Stable intake rules, reminders, status synchronization and copying approved data between systems can remove boring work. But a bot can’t reliably guess a missing policy, and adding AI to a workflow doesn’t give it permission to make a financial commitment or change a customer’s records.
The useful audit fits into one afternoon
You can learn quite a lot without a consulting project or a forty-column dashboard. Pick one process that frustrates customers or staff, pull five to ten recent examples and ask the people involved to help reconstruct what happened. Keep the review small enough to finish. You are looking for a specific change worth trying, not trying to document the entire company.
- Count the full elapsed time. Separate active work from the hours or days spent waiting.
- Mark the stops. Note missing information, approvals, rework and transfers between tools.
- Check whether the queue is growing. Compare new requests with completed ones over the same period.
- Identify a repeatable cause. Look for the same approval, missing field or handoff appearing more than once.
- Change one thing and observe it. Give the revised process a few weeks of ordinary work before deciding it helped.
A simple scorecard usually does the job. Total completion time tells you whether customers are waiting less. Reopened tasks or corrections tell you whether faster work is causing more mistakes. And a quick look at the oldest pending item tells you whether the team is only getting better at handling easy requests.
Be skeptical of improvements that merely move the delay. Speeding up the intake form is pointless if all the new work sits in the same unstaffed queue. A dashboard that reports more completed tasks is not an improvement if the customers now receive more errors or have to call twice.
Make the change survive an ordinary week
Many processes work during the week everyone is paying attention to the rollout. The test comes later, when someone is on vacation, a difficult customer calls and a project needs an exception. Can another person find the current instructions? Do they know who is allowed to decide? Can they see the status without asking three colleagues for an update?
Give people enough authority to finish routine work without waiting for an unnecessary approval. Keep the checks that prevent expensive mistakes. Sometimes clarifying that boundary is enough; if the systems still need to communicate more reliably, CMX’s AI automation and integration services cover that technical work.
Start with the request that has been stuck the longest and find out why. It might tell you more about your operations than a month of reports.
