Little's Law says that the average number of people in a system equals the average rate at which they arrive multiplied by the average time each one stays. Write it as L = λW. If 30 guests arrive per hour and each spends 0.5 hours in your restaurant, you have 15 guests inside at any moment, on average. The same three-way link tells you how long a line will feel, how many chairs you need and what happens to waits when you add or remove capacity.
This guide explains what the law means, walks through worked examples for service businesses, and shows where people misuse it.
TL;DR
- Little's Law: L = λW. Average number in the system equals arrival rate times average time in the system.
- Rearranged, W = L / λ. You can estimate a wait from the line length and the arrival rate without timing anyone.
- It holds regardless of arrival pattern, but you must define the system boundary and measure over a long enough window.
- You can't change one variable in isolation. Cutting the wait with the same arrivals means fewer people in the system.
- Visible, honest queue status makes waits easier to tolerate, even before you shorten them.
Table of Contents
- What Little's Law says, in plain words
- Why it works for messy, real businesses
- Worked example 1: a barbershop
- Worked example 2: a restaurant with a waitlist
- Worked example 3: a clinic front desk
- The mistake that ruins most calculations: the system boundary
- Measure long enough
- Variability: what the law does and doesn't fix
- Levers: what you can actually change
- Perception matters as much as the average
- Our perspective: use the law as a diagnostic, then make the line visible
- Putting it to work with Ezseat
- FAQ
- Sources
What Little's Law says, in plain words
The law links three averages:
- L: the average number of items (customers, patients, tickets) in the system.
- λ (lambda): the average arrival rate, for example customers per hour.
- W: the average time one item spends in the system, from arriving to leaving.
The formula L = λW was proven by MIT professor John D.C. Little in 1961, and it is still a foundation of queueing theory. Wolfram MathWorld states it for steady-state conditions: the average number in a queuing system equals the arrival rate multiplied by the average time spent there.
Operations teams often use different labels for the same idea: work in progress (WIP) equals throughput times lead time. Everhour's guide shows how rearranging it into lead time = WIP / throughput lets you calculate the effect of shrinking the number of people or jobs in progress.
Three forms you will actually use:
| You know | You want | Formula |
|---|---|---|
| Arrival rate and time in system | People in system | L = λ × W |
| People in system and arrival rate | Average time in system | W = L ÷ λ |
| People in system and time in system | Arrival (throughput) rate | λ = L ÷ W |
Keep the units consistent. If λ is per hour, W must be in hours.
Why it works for messy, real businesses
Many people assume the law only applies to calm systems with perfectly random (Poisson) arrivals, so they dismiss it for a Friday-night dinner rush. That is a misreading. Operations researchers describe it as a rare tautology that is true for any realization of a system, not just stable, predictable ones. It doesn't care whether your guests arrive in clumps or whether service times vary wildly.
What it does require is that the system is not growing without limit over the period you measure. Columbia University's course notes describe steady state as a system that isn't growing in an unbounded way. If the line gets longer every hour and never recovers, the averages you compute are not meaningful.
The practical point is the one experts stress: you cannot independently pick arrival rate, wait time and queue size. Improving one almost always trades off against another.
Worked example 1: a barbershop
A two-chair barbershop on a Saturday. Over four busy hours you count:
- 10 walk-ins arrive per hour (λ = 10 per hour).
- Each customer spends 45 minutes from walking in to leaving the chair, which includes waiting and the haircut (W = 0.75 hours).
L = 10 × 0.75 = 7.5 people in the shop on average.
Now split that. If the haircut itself takes 30 minutes, the time spent waiting is 15 minutes. Applying the law to just the waiting area: L = 10 × 0.25 = 2.5 people on the bench on average, and 5 people in the chairs. (Two chairs can hold only two at a time, so with 30-minute cuts and 10 arrivals per hour, capacity would be 4 customers per hour per chair... wait: 2 chairs × 2 cuts per hour = 4 per hour.) That matters. If arrivals really are 10 per hour and capacity is only 4 per hour, the line grows without limit and the steady-state assumption fails. The honest conclusion is that the numbers were inconsistent: either arrivals are lower than 10 on average, or the shop has more chairs.
That is the first lesson. Little's Law is a consistency check on your own data. When the three numbers you measured don't fit together, one of them is wrong or the system isn't stable.
A corrected version: 3 customers arrive per hour, each stays 40 minutes in total (about 0.67 hours). L = 3 × 0.67 = 2 people in the shop on average. Fits two chairs with a short wait.
Worked example 2: a restaurant with a waitlist
You run a 60-seat restaurant. Between 6 and 8 p.m. parties arrive at 24 per hour on average, and 8 parties on average are waiting for a table at any moment (L = 8 for the waiting line only).
W = L ÷ λ = 8 ÷ 24 = 0.33 hours, or 20 minutes average wait.
You didn't need a stopwatch. You counted the line a few times and counted arrivals. Now test a change. If you speed up table turns so the average line falls from 8 parties to 5, with the same 24 arrivals per hour:
W = 5 ÷ 24 ≈ 0.21 hours, about 12.5 minutes.
The reverse also works. If a host tells guests "about 15 minutes" but the line holds 8 parties with 24 arrivals per hour, the quote is optimistic by about five minutes. Quoted waits should come from the arithmetic, not hope. Our guide to wait time estimation methods and tools goes deeper on how to set those quotes.
A 2025 study of a restaurant queueing model describes how managers can use mean arrival and service rates to compute a utilization factor (ρ = λ/μ). If arrivals approach or exceed what the kitchen and tables can serve, the line stops being stable and prospective customers walk away.
Worked example 3: a clinic front desk
A small clinic sees patients from 8 a.m. to noon. Over those four hours, 48 patients arrive (12 per hour). On average 9 patients are in the building, whether in the waiting room or with a provider.
W = L ÷ λ = 9 ÷ 12 = 0.75 hours, so a patient spends 45 minutes in the clinic on average.
Now the useful move: split the system. If 3 of those 9 are with providers and 6 are in the waiting room, the average wait before being seen is 6 ÷ 12 = 0.5 hours, or 30 minutes, and the average time with a provider is 3 ÷ 12 = 15 minutes. One law, three different boundaries, three different answers. That leads to the most common error.
The mistake that ruins most calculations: the system boundary
The law is fractal. It applies to a whole restaurant, to one line inside it, and to a single prep station. According to Columbia's lecture notes on analysis, you can pick any system boundary you like, but you have to count arrivals, time and people for the same boundary.
Before you measure, write down:
- Where the clock starts. The door, the moment someone joins the waitlist, or the moment they take a ticket?
- Where it stops. Seated, served, paid, or out the door?
- Who counts. Do parties of four count as one arrival or four? Pick one and stay with it for λ and L alike.
Mixing boundaries, for instance counting people waiting outside but timing only from the moment they check in, produces numbers that never reconcile.
Measure long enough
Short observation windows distort results. If you count at 6:00 p.m. when the line is at its peak and stop at 6:30, more people may have arrived than left, so the averages don't reflect a balanced system. A paper on the law's practical use points to end effects: over a long enough period, cumulative arrivals and departures should match, so the system effectively clears.
A simple routine for a small business:
- Choose one daypart, such as Friday dinner, and measure the same daypart across several weeks.
- Start and end the window when the line is empty, or close to it.
- Log arrivals per hour, the line count every 10 to 15 minutes, and a sample of actual times.
- Compare W from the stopwatch with W from L ÷ λ. A big gap means a boundary problem or an unstable window.
Variability: what the law does and doesn't fix
Little's Law doesn't make a rush smaller. Arrival rates and service times in restaurants, hotels and clinics are rarely constant. Research on four-star hotels built on the Process Management Triangle, which rests on Little's Law, shows how inventory, variability and capacity utilization are tied together: change one and the others move.
In practice that means the law gives you the framework for sizing a buffer. If you know your peak arrival rate and your target wait, you can calculate how many people will be in the system and plan staff or space for that number. We cover the staffing side in peak hour capacity management without extra hires.
Levers: what you can actually change
From W = L ÷ λ and L = λW, the options for shortening the average time in the system are limited:
| Lever | Effect on L | Effect on W | Example |
|---|---|---|---|
| Cut service time | Falls | Falls | Pre-collect party size and name before the host is free |
| Add parallel capacity | Falls | Falls | Open a second register or chair |
| Reduce arrivals (λ) in the peak | Falls | Falls | Shift demand with off-peak offers or booking |
| Cap the line (limit L) | Falls | Falls for those admitted | Close the waitlist at a set length |
| Do nothing | Unchanged | Unchanged | Hope |
The third row is the one many owners overlook. Because arrivals multiply wait, even a modest drop in peak arrivals shortens the line. Everhour makes the same point for project teams: if you want to reduce lead time without changing throughput, you must lower work in progress.
Perception matters as much as the average
The math gives you the average wait. Customers experience something else: uncertainty. Service research from 2025 notes that failing to manage the queue leads to prospective customers departing, and that transparency about the wait lowers the anxiety of not knowing how long it will last.
That is why showing status matters. A guest who can see their position, or who gets a message when their turn is close, can sit in the car, browse next door or order a drink instead of hovering at the host stand. Our pieces on fixing queue customer experience and queue display screens cover the practical side.
Once the line is under control, the next question is how to bring those customers back. Some owners pair a queue with digital loyalty and reward cards so a good first visit turns into a second.
Our perspective: use the law as a diagnostic, then make the line visible
We build a web-based queue system for restaurants, clinics, food trucks, barbershops and events, so we see the same pattern in small businesses again and again. Owners know the line "feels" long but have no numbers. Little's Law is the shortest path from feeling to numbers: count arrivals, count the line, and you have the average wait.
Two things we have learned from how operators use queue tools:
- Collecting the data should cost nothing. When customers join through a browser and the operator manages the line from a phone, every arrival and every call is already logged. No clipboard, no shouting names. That makes λ and L easy to track instead of a one-off exercise.
- Fixing the average is half the job. The other half is making the wait understandable. Customers who don't need to install an app or stand at a counter tolerate a wait far better than customers who feel stuck.
You don't need heavy hardware or per-message fees to do this. A clean flow, a visible number and an honest estimate go a long way.
— Ezseat
Putting it to work with Ezseat
If you want your queue to produce the numbers this article uses, Ezseat lets customers join from a web browser with no app to download, and lets you run the line from a phone or tablet. You can add public screens for ticket numbers, a reception kiosk, SMS or call notifications and custom queue fields.
Plans on ezseat.app: Light is $3.90 per month or $39.00 per year, and Pro is $14.90 per month or $149.00 per year. Every account starts with a free two-month trial before any charge. If you're still comparing options, see our overview of queue software pricing and the best queue management software for SMBs.
FAQ
What is Little's Law in simple terms?
It says the average number of people in a system equals how fast they arrive times how long each one stays. If 20 customers arrive per hour and each stays 15 minutes (0.25 hours), about 5 are inside at any time. It works for a whole shop, a single line or one station.
Does Little's Law only work when arrivals are steady?
No. It does not require Poisson or evenly spaced arrivals. It does require that the system isn't growing without limit during your measurement, and that you count over a window long enough for arrivals and departures to balance.
How do I use Little's Law to estimate a wait time?
Count the average number of people waiting (L) and the average arrival rate (λ), then divide: W = L ÷ λ. Eight parties waiting with 24 arriving per hour gives 20 minutes. Re-check the figure at different times of day, because both inputs change.
Is Little's Law only for restaurants and queues?
No. Operations teams use it for inventory, lead time in software projects and call centers, usually written as WIP = throughput × lead time. Queues just make it easy to see, because you can count the people.
Sources
- Little's Law as Viewed on Its 50th Anniversary
- Little's Law, Wolfram MathWorld
- Little's Law: How to Manage Queues and Boost Efficiency, Everhour
- A Practical Approach for Studying the Impacts of a Queuing Model to Enhance Restaurant Service
- Chapter V: Analysis of Packet Switching Networks, Columbia University
- Chapter 5 Little's Law
- An exploration of process variability and its management: a case study of four star hotels
- Features, Beepify