How to Identify Manual Workflow Bottlenecks
Picture a Tuesday afternoon in a small distribution office in Bayamón. Marisol has been typing since eight. Her hands are tired, her inbox is full, and she is certain she is the bottleneck. Yet when the owner asks why a customer's order took nine days to ship, the answer has nothing to do with Marisol's typing. The order sat in a shared folder for two days waiting for someone to notice it. Then it sat on a manager's desk for another three waiting for a credit approval that took four minutes to give.
This is the most common surprise in process work. The people who look busiest are rarely the constraint, and the step everyone complains about is often not the step that costs you the most time. If you own a process and you want to fix it, the first job is not to automate anything. The first job is to see it accurately.
This guide gives you a method you can run with a notebook, a stopwatch, and an afternoon of honest conversation. It is written for the person who owns a process, whether that is order entry, client intake, invoicing, or hiring, and who has to decide where to spend scarce time and money.
Start by watching the work, not asking about it
If you ask people how a process works, you get the process as designed. If you watch, you get the process as lived. These are different, and the gap between them is where most delay hides.
Pick one type of work item. An order, an invoice, a new client, a service ticket. Then follow a handful of real items from the moment they arrive to the moment they are finished. Do not follow the happy path only. Pick five items and make sure at least one was recent and slow, and at least one was recent and fast.
For each item, write down every place it stopped and every time someone touched it. You are building a timeline, not a flowchart. Useful sources include email timestamps, status changes in your CRM or accounting system, file modified dates, and the memory of the person who handled it, checked against those records.
A few rules keep the observation honest.
- Record what happened, not what should have happened. If an order went back to the customer twice for missing information, write that down even if the procedure says it should not.
- Note who held the item at each point. Handoffs are where things go quiet.
- Time the touches separately from the gaps. This is the heart of the method and the next section explains why.
- Do not judge anyone. People hide workarounds when they sense blame, and the workarounds are exactly what you need to see.
Separate work time from wait time
Every step in a manual process has two clocks running. Work time is the minutes someone actually spends doing the thing. Wait time is the stretch between the moment the item is ready for that step and the moment someone starts.
In most small-business processes, work time is a small fraction of elapsed time. The ratio of work time to total elapsed time is sometimes called flow efficiency. It is common for it to land in the single digits or low teens, which means that for most of its life your order is not being worked on at all. It is waiting.
This matters because the two kinds of time have different cures. If work time is long, you speed up the task, simplify it, or remove it. If wait time is long, speeding up the task does nothing. You have to change how work arrives, who is notified, who is allowed to proceed, and how many things are in line ahead of it.
An illustrative example
The numbers below are hypothetical, built to show the method. Suppose an order for a small distributor goes through six steps over nine calendar days.
| Step | Work time | Wait before the step |
|---|---|---|
| Order arrives by email, gets read and logged | 6 minutes | 1 day |
| Entered into accounting system | 12 minutes | 2 days |
| Credit check and approval | 4 minutes | 3 days |
| Pick and pack | 35 minutes | 1 day |
| Invoice created and sent | 8 minutes | 1 day |
| Shipment booked with carrier | 10 minutes | 1 day |
Total work time is 75 minutes. Total elapsed time is about nine days. Even if you count only business hours, work is well under one percent of the clock. Now ask where the real gain is. Typing the order faster saves a few minutes. Getting the credit approval to happen the same day saves roughly three days. Making sure new orders are noticed within hours instead of a day saves another day or two.
Notice that the step with the most work time, pick and pack, is not where the delay lives. If you had asked the team where the problem was, many would have pointed there, because that is where the effort is visible.
Understand queues, because they explain most of the waiting
You do not need a mathematics degree to use queueing ideas. Three pieces of intuition do most of the work.
Little's law in one sentence
On average, the number of items in your process at any moment equals the rate at which items arrive multiplied by how long each one stays. If you receive 20 orders a day and each takes 5 days to complete, you are carrying about 100 open orders. This works in both directions. If you can count the open items and you know the arrival rate, you can estimate lead time without timing anything. And if open items are growing week over week, lead time is growing too, even if nobody has complained yet.
Busy people create long lines
When someone is busy 95 percent of the time, a small wobble in arrivals produces a long line. A rough rule from the simplest queueing model says that average waiting time scales with utilization divided by one minus utilization. Here is what that looks like if one task takes an hour of work.
| Utilization | Approximate average wait, in multiples of the task time |
|---|---|
| 50 percent | 1 times |
| 70 percent | 2.3 times |
| 80 percent | 4 times |
| 90 percent | 9 times |
| 95 percent | 19 times |
The real world is messier than the model, but the shape is right. Waiting does not grow in a straight line. It explodes as you approach full utilization. That is why asking your most loaded person to be even more efficient is often the wrong move. What helps is giving that person slack, or taking some of the arriving work away.
Batching hides delay
Many small teams batch without noticing. Invoices are sent on Fridays. Bank deposits are matched on the first of the month. Approvals happen when the owner returns from a site visit. Batching can be reasonable, but each batch adds an average wait of half the batch interval. A weekly batch adds about two and a half business days of waiting per item before any work happens. Look for batches that exist only because of habit.
Look for rework, the quiet second process
Rework is any time an item goes backward. A form returned for a missing signature. An invoice corrected because the purchase order number was wrong. A customer record fixed after someone discovered a duplicate. Each loop adds work time, and it also adds a fresh round of waiting, because a returned item goes back to the end of a line.
You can measure rework with two simple questions for each step. What fraction of items arrive here with something missing or wrong, and where did the problem originate? Tally the answers on your observed items. You will almost always find that a large share of downstream pain traces back to one or two upstream causes, such as free-text fields, unclear intake forms, or a handoff by email that drops attachments.
Fixing a defect at the source is cheaper than catching it at the end. This is also where the right kind of automation often belongs. A form that refuses to submit without a valid account number removes a whole category of rework, and no one has to type faster.
Find the real constraint before you fix anything
In any chain, one step limits how much the whole system can finish. Improving any other step does not raise output. It only moves the pile somewhere else.
Here is a practical way to find it.
- Walk the timeline and mark where items pile up. Piles are visible in inboxes, in a folder of unprocessed files, in a spreadsheet tab with rows waiting for a status change.
- The constraint usually sits just downstream of the largest pile. Work accumulates in front of the step that cannot keep up.
- Check whether the pile is caused by capacity or by dependency. Is the person overloaded, or are they waiting on information, a signature, or a system that is down? Those are different problems.
- Test your guess by asking what would happen if that step doubled in speed. If total lead time would barely change, you have the wrong step.
Three ways teams fix the wrong step
Optimizing the visible work. The step with the most typing, clicking, or physical effort gets attention because it looks like the problem. See the example above.
Automating the loudest complaint. The person who speaks up most is not always the person who holds the constraint. Take every complaint seriously and then verify it against the timeline.
Adding people to a queue caused by a rule. If every order over a threshold needs one particular approver, a second data entry clerk does nothing. The rule is the bottleneck, and the fix may be a delegation policy.
Decide what kind of fix fits
Once you know where time is lost, match the cure to the cause.
| What you found | Likely cure |
|---|---|
| Long wait before a step, worker is idle part of the day | Notification or routing so the item is noticed at once |
| Long wait before a step, worker is overloaded | Remove work, add capacity, or smooth arrivals |
| Long wait because of an approval | Delegation limits, pre-approval for low-risk cases, a deadline with a default |
| Long work time on a repetitive task | Templates, validation, integration between systems |
| High rework from one source | Fix the intake form or the source system |
| Batching without a reason | Move to continuous flow or shorten the interval |
| Handoffs through email | Shared queue with ownership and status |
Notice how few of these require software that costs real money. Many are policy and design changes. Technology comes in when the cure is routing, validation, or connecting two systems that currently depend on a person copying data between them.
A Puerto Rico note on resilience
Bottlenecks have a seasonal personality in places where power and connectivity are not guaranteed. If a single person holds a step and that person is out for a week after a storm, the whole process stops. When you map who holds each step, mark the ones with no backup. These are single points of failure, and they are worth fixing even if they are not today's slowest step. A process that depends on one laptop, one inbox, or one person's memory can look efficient right up until the moment it fails entirely.
Small teams also tend to carry bilingual work, where an item moves between English and Spanish speakers. Watch for waiting that happens only because the one person comfortable in the other language is the only one who can proceed.
Measure before and after
Whatever you change, decide in advance what you will measure. Three numbers are enough for most small processes.
- Lead time, from arrival to completion, as a median and as a slow-end figure such as the slowest one in ten.
- Open items at the start of each week, which tells you whether the queue is growing.
- Rework rate, the share of items that go backward at least once.
Record them for a few weeks before you touch anything. Without a baseline you will rely on impressions, and impressions are shaped by whoever spoke last in the meeting. Averages can hide a lot, so look at the slow end. The customer who waited nine days remembers nine days, not the average.
What to do Monday morning
You can start this week with very little.
- Choose one process and one type of work item.
- Pull five recent examples, including a slow one and a fast one.
- For each, write the timeline with timestamps for arrival, each touch, each wait, and completion.
- Split every step into work time and wait time and add them up.
- Mark the biggest piles and ask what is making the work wait there.
- Count rework loops and trace each to its origin.
- Write one sentence naming the suspected constraint and one sentence describing what would prove you wrong.
- Pick the smallest change that addresses the biggest wait, make it, and measure again in two weeks.
Resist the urge to build something large before this is done. The point of the exercise is to avoid spending months automating a step that was never the problem.
When it helps to bring in another set of eyes
Process owners are often too close to see their own queues. A colleague from another department can ask naive questions that expose habits you stopped noticing. If the findings point to work that really does need integration between systems, such as orders retyped from email into accounting software, that is where Ozcorp InfraOps can help, and we are happy to talk through a timeline like the one above with you. If your fix turns out to be a delegation rule and a shared inbox, even better. You will have saved the money for something else.
Discuss your infrastructure requirements
Review the current service offer or schedule a conversation about the scope you need.
Stay in touch
Follow Ozcorp
Leave your details for occasional notes about new articles and Ozcorp updates.