The person who interrupts you at 9:17 cannot see the person who interrupted you at 9:06. Each arrives with a reasonable request and the innocent confidence of being the only reasonable request in the building. By lunchtime, you have worked continuously and completed almost nothing. Everyone has been promised a quick favor. The favors have formed a committee.

It is tempting to interpret this as a personal failure. Perhaps you need more discipline, a better morning, or a notebook with a more authoritative cover. Sometimes you need a queue. A visible, agreed way for work to enter, wait, and leave can do more than another attempt to become the sort of person who is never interrupted.

A queue sounds bureaucratic because we picture a ticket machine beside a tired plastic plant. But its essential function is considerate: it prevents several people from unknowingly making promises about the same hour of your life.

Full is a condition, not an achievement

Suppose a small service desk receives, on average, six requests an hour. Each takes, on average, ten minutes. On paper, the arithmetic looks beautiful. Sixty minutes of demand meet sixty minutes of capacity. Nothing is wasted. Somebody may even put this on a slide.

There is no margin for variation. One request takes twenty minutes. Two arrive together. The person serving them needs to clarify an address, replace a battery, or drink water. When arrivals and service times vary, operating close to full capacity can produce long waits. Exact waiting times depend on the pattern of work; the basic practical problem is that the plan has left no room for ordinary reality.

This is why an empty slot can be useful capacity rather than evidence of idleness. You would not schedule every road junction on the assumption that no driver would ever hesitate. Human work contains its own hesitations, broken instructions, and unusually complicated left turns.

The first question to ask about a permanently overloaded day is therefore not whether you can concentrate harder. Ask whether the amount arriving can fit through the available capacity at all.

Count the work inside the system

A useful relationship from operations research, Little’s law, connects average work in a system, average throughput, and average time spent there. For a stable, consistently defined system, the first equals the second multiplied by the third. It is an accounting relationship, not a spell for making people faster.

Consider an invented example. A team has an average of twenty requests in progress or waiting, and completes ten a day. That corresponds to an average of two days in the system. Those averages do not promise that a particular request will finish tomorrow. If the average number in the system rose to forty while throughput remained ten a day, the corresponding average time would be four days. The relationship describes the resulting averages; it does not predict the exact delay caused by the next request.

Define the boundaries honestly. If you count waiting work in the backlog, include its waiting time in the elapsed time you measure. If a request regularly leaves your team to wait for approval, decide whether that wait belongs inside the system you are trying to understand. Moving it into an invisible spreadsheet does not make the customer’s calendar move faster.

Make arrival a decision

A workable queue answers four questions. Where does a request go? What information does it need? Who decides its position? What does “finished” mean? You can answer them with a shared list and a short conversation. The sophistication of the tool is rarely the interesting part.

For a household repair list, the request might need the location, the problem, and a photograph. For a work team, it might need the desired outcome and the date that actually matters. “As soon as possible” describes a preference shared by nearly everyone. A real deadline explains what changes if the work waits.

Give urgent work a route, but define the route. A genuine safety problem should not stand politely behind a request for a nicer folder color. Equally, a queue in which everything can acquire an emergency sticker is simply a collection of stickers. Someone must be allowed to ask what makes this request different.

That person also needs authority to say which existing commitment moves. Otherwise urgency becomes a system for adding work while concealing the subtraction.

Finish a little more; begin a little less

Limiting work in progress can feel oddly uncomfortable. Beginning a task provides the social pleasure of saying yes. Finishing it requires the less glamorous business of resolving details. A team can acquire an impressive number of beginnings while its customers collect reasons to ask for updates.

Try a modest limit rather than a revolution. Agree that each person will actively work on one or two substantial items, with exceptions made explicitly. Keep waiting items visible. At a short daily check, examine the oldest blocked item and decide who can remove its obstacle. A list of fresh starts is less useful to the person still waiting for last week’s answer.

The purpose is to reduce avoidable switching and expose obstacles. A limit that prevents sensible collaboration has become another obstacle and needs adjustment. A queue should help people notice the work, not become a moral theory about how many browser tabs they are permitted to have.

Keep a place for surprise

Review the pattern after a week. How much unplanned work arrived? Which requests repeatedly lacked essential information? Which “emergencies” were predictable events that nobody had assigned a date? The answers are material for redesigning the system, rather than evidence for a prosecution of the people inside it.

Reserve some capacity for the interruptions you genuinely need to handle. If that reserve is repeatedly exhausted, revise the commitments. Calling the same overload exceptional every Wednesday is an unusually elaborate way of declining to learn from Wednesdays.

The next person who arrives at 9:17 can still have an urgent problem. Now you can show them the commitments already in the hour and decide, together, which one should move. That is a more useful answer than another yes you will spend the afternoon trying to make true.

Sources and further reading

MIT OpenCourseWare: review of queueing models and Little’s law

More from The Journal →