Operator's reference · Valet throughput

How many cars can one valet attendant actually handle per hour?

Short answer: about five complete guest visits, or nine to twelve individual car movements, and both of those collapse fast as the lot gets further away. Here is where those numbers come from, why most published figures quietly answer a different question, and how to measure your own crew with a stopwatch in twenty minutes.

Published 17 August 2026 · Operator's reference · 13 min read

The question has three answers, and mixing them up is why nobody agrees

Ask five operators how many cars an attendant handles per hour and you will get five numbers between four and forty. They are not contradicting each other. They are answering three different questions without saying which one.

The third one is the one people quote when they mean the first one, and it is the reason a crew can be correctly sized on paper and still have thirty people standing at the podium. We covered the staffing ratios themselves as part of how to build a valet quote. This page is about the rate.

What we could and could not verify. There is no US industry rate or productivity survey for valet that we could find. The National Parking Association offers a Valet Operations Certificate and a Certified Valet Attendant credential, but we found no published throughput benchmark from any parking association. Every figure below traces to a named operator writing about their own operation, to a public engineering standard, or to arithmetic we show our working for. Where a variable is real but unmeasured in public, we say so rather than filling the gap with a plausible number.

The one published cycle time methodology we found

The most useful public source on this question is not a benchmark at all. It is a worked staffing example in Richard Raskin's write up on managing a valet operation, which gives three inputs and then does the arithmetic in front of you:

InputValue
Time to park a vehicle and return to the porte cochere5.5 minutes
Time to retrieve a vehicle and deliver it to the guest4.25 minutes
Productive time per attendant50 minutes of each 60 minute hour

His example is a hotel morning with 15 arrivals and 50 departures in the same hour. That is 82.5 minutes of parking plus 212.5 minutes of retrieving, just under 300 minutes of work, which at 50 productive minutes each requires six attendants on the floor.

Two things about that are worth more than the numbers themselves. First, it is a method, so you can substitute your own measured times and it still works. Second, the 50 minute assumption is doing real work: it says that one sixth of the shift is not a cycle at all. It is greeting, waiting for the porte cochere to clear, taking payment, hunting for a key, handing off to a manager, and being human. Any model that assumes 60 productive minutes is wrong by 20 percent before it starts.

What that method produces

Run those inputs out and you get the three answers to the question at the top of the page:

MeasureArithmeticPer attendant per hour
Parking only, arrival rush50 ÷ 5.59.1 cars
Retrieving only, departure rush50 ÷ 4.2511.8 cars
Balanced mix of both50 ÷ 4.87510.3 movements
Complete guest visits, park and retrieve50 ÷ 9.755.1 cars
≈ 5

Complete guest visits per attendant per hour, at a venue where the lot is a short walk from the door. Nine to twelve if you count individual car movements instead.

Notice that retrieving is faster than parking in these figures, by more than a minute. That is not a mistake. On the way out the attendant walks to the car and drives back; on the way in they drive out and walk back, but they also greet the guest, write or scan the ticket, adjust the seat, and deal with whatever the guest left in the driver's door. Retrieval front-loads none of that. It just feels slower because the guest is standing there watching a clock, which is a service problem rather than a throughput problem.

Distance to the lot is the number, and everything else is a rounding error

Raskin does not say how far his lot was. You can back it out, and the exercise is the most useful thing on this page.

The return leg of a park cycle is walked, and walking is slow. US traffic engineering practice sizes the pedestrian clearance interval at a signal on a walking speed of 3.5 feet per second, about 210 feet a minute. That is a deliberately conservative design figure meant to cover slow walkers, and a fit attendant on clear pavement will beat it, but it is a published number rather than a guess, and it is close enough for planning. Driving inside a lot at a sane 5 mph is 440 feet a minute.

So a one way distance of d feet costs about d/440 minutes driving out plus d/210 minutes walking back, roughly 0.007 minutes per foot for the pair. If you assume 3 minutes of fixed handling per park (greet, ticket, get in, manoeuvre into the stall, lock up), Raskin's 5.5 minute cycle implies a lot about 350 feet from the door. Change the fixed handling assumption to 2 minutes and the implied distance jumps to about 500 feet; change it to 3.5 and it drops to about 285. That sensitivity is the point: two operators can both be honest and land two hundred feet apart.

Hold fixed handling at 3 minutes and vary only the distance, and the throughput curve looks like this:

One way distance to the lotCycle timeMovements per hourComplete cars per hour
150 ft, adjacent lot4.1 min12.36.2
350 ft, across the parking area5.5 min9.24.6
600 ft, around the block7.2 min6.93.5
1,000 ft, one fifth of a mile10.0 min5.02.5
1,500 ft13.6 min3.71.8
2,640 ft, half a mile21.6 min2.31.2

A lot 1,000 feet away costs you 60 percent of your throughput compared with one at 150 feet, with the same people and the same effort. That is why two operators quoting the same headcount for the same guest count can produce completely different nights, and it is why "how far is the parking" belongs on your intake form above almost every other question.

The published operator experience matches the shape of that curve. One account of resort valet describes retrieval dropping from 8 minutes to under 3 after adding satellite stations closer to where the cars actually were. That is a distance fix, not a staffing fix, and it roughly doubles what each person can do.

The threshold where you stop hiring and start driving. Look at the bottom rows. Past roughly a quarter mile, an attendant spends more of the cycle walking than doing anything a customer values, and adding people multiplies a bad cycle instead of fixing it. That is the point where a runner shuttle pays for itself: one driver ferrying attendants back to the stand removes the walk from every cycle in the operation. Published examples exist, including golf cart shuttles bridging a five block gap between overflow lots and the gate at a Tampa arena. We found no published data on the exact distance at which a shuttle beats headcount, because it depends on your labour cost and the cart, but the arithmetic above lets you work out your own crossover in about ten minutes.

Burst rate is not sustained rate, and this is where the wild claims come from

If you have heard thirty or forty cars an hour per attendant, that figure is real and it is also not repeatable. A published account of amphitheater valet states that a well coordinated team of eight can process 40 to 50 vehicles in the first ten minutes after a show ends. Divide it out: 5 to 6.25 cars per person in ten minutes, an hourly rate of roughly 30 to 37 movements.

Three things make that possible, and all three expire:

Burst numbers are genuinely useful for one thing: sizing a departure surge. If you know the show ends at 22:30 and you have 400 cars, the burst rate tells you how many people you need for those ten minutes. It tells you nothing about how to staff a Tuesday.

ParkingPro's shift report shows how many vehicles each attendant actually handled on a real shift, which is the only version of this number that is genuinely yours. US$19/month, 14 days free, no sales call.

Start the free trial →

What else moves the number, and how confident you can be about each

Running versus walking

A jog at 8 feet per second instead of a 3.5 foot per second walk cuts the return leg by more than half. At a 600 foot lot that takes the cycle from 7.2 minutes to about 5.6, roughly a 28 percent throughput gain, and the gain grows with distance. It is the single largest controllable variable after distance itself.

It is also the one nobody will put in writing. We could not find a published policy from any major US operator, or any association guidance, either requiring or prohibiting running in the lot. What is published is generic and cautious: safety guidance for valet operations tends to cover seat belts, observing posted speed limits and avoiding sudden stops and starts, and says nothing about the attendant on foot. Draw your own conclusions from the silence, but understand the two real constraints. A sprained ankle on wet pavement is a workers' compensation claim, and no human sustains a jog for the fifth hour of a shift. Whatever you decide, decide it explicitly and put it in the handbook, because an unwritten rule here becomes an argument after an injury.

Vertical movement

A multi level garage behaves differently from a flat lot of the same distance, and worse. A ramp is walked at less than 3.5 feet per second going up. An elevator is not a distance at all, it is a queue, and its wait time gets longer exactly when you need it shortest. If your parking is three levels below grade, measure your cycle rather than estimating it from a floor plan, because the floor plan will lie to you.

Tandem stacking

Stacking is what makes valet worth having: parking tighter than self park allows is commonly cited as fitting 40 to 60 percent more vehicles into the same space. It buys capacity and it sells throughput. A blocked-in car does not cost one movement, it costs three: move the blocker, get the target out, put the blocker back. If one retrieval in five is blocked, your effective retrieve cycle goes up by roughly 40 percent of a single move, and your departure rush is the moment it hits.

The mitigation is not more people, it is better ordering. Parking by expected departure time, keeping short dwell vehicles out of the deep stacks, and running first in first out within a stack all attack the same cost. So does knowing which car is where without walking the lot to find out.

Vehicle difficulty

No public data exists on this that we could find, and it is not for lack of looking. What every operator will tell you anecdotally is that the distribution has a long tail:

Weather

Also unquantified in public. The operationally nasty part is that it moves both sides at once: rain slows every cycle and simultaneously converts people who would have walked into people who arrive by car. Your worst throughput and your highest volume show up in the same thirty minutes.

The arrival curve

Not a throughput variable at all, strictly speaking, but the one that decides whether your throughput is enough. Event guidance describes a peak arrival window of 20 to 40 minutes for most events, and one operator account cites holiday resort weekends seeing 200 or more vehicles arriving inside a 4 hour window. Those are two completely different problems. A hundred cars over four hours needs 25 movements an hour, which is three people. A hundred cars in twenty minutes needs 300 movements an hour, which is thirty people you do not have, so the real answer is to stack at the curb, take keys faster than you park them, and clear the backlog during the meal. Curb stacking is a legitimate technique, not a failure. It buys throughput with curb space, and the only thing that makes it dangerous is running out of curb.

Measuring your own number, which takes about twenty minutes

Every number above is a starting point. The one that matters is yours, and getting it is genuinely quick.

  1. Pick your worst thirty minutes, not a representative half hour. A Tuesday cycle time is useless for sizing a Saturday. Measure the condition you are trying to survive.
  2. Define the cycle as door to availability. Start the stopwatch when the attendant takes the key, stop it when they are back at the stand and free to take the next one. Not "time to park the car". The walk back is half the cycle and it is the half people forget to count.
  3. Time ten parks and ten retrieves separately. They are different numbers and they peak at different times of the night.
  4. Use the median, not the average. One manual transmission, one blocked stack, one guest who cannot find their phone, and a mean of ten samples is garbage. The median tells you what a normal car costs.
  5. Measure your productive minutes instead of assuming them. Shadow one attendant for a full hour and total the time they are not on a cycle. Raskin assumes 50 of 60. If your attendants also take payment at the curb, work the key board, or hold the podium, yours will be lower, and that gap is where the missing capacity went.
  6. Do the arithmetic. Movements per hour equals your productive minutes divided by your median cycle. Complete cars per hour equals your productive minutes divided by the sum of your median park and median retrieve.
  7. Check it against reality. Take a real shift's vehicle count, divide by attendant hours worked, and see whether it lands near your calculated figure. If the real number is well below the calculated one, the difference is not cycle time, it is idle time, waiting, or a bottleneck at the podium. That is a different and usually cheaper problem to fix.

The most common way this measurement goes wrong. Managers time the fast attendant because they are the one who looks impressive with a stopwatch pointed at them. Your crew capacity is set by the median attendant on the median car, and on a busy night it is set by the slowest link in the chain, which is often not an attendant at all but the porte cochere itself. If cars cannot be handed off because there is nowhere to put them, hiring a seventh person changes nothing.

Turning the number into a crew size

Once you have your own cycle times, the crew calculation is one line, and it is Raskin's:

Attendants needed = (expected arrivals × your park cycle + expected departures × your retrieve cycle) ÷ your productive minutes per hour.

Do it for the worst hour, not the average hour, and round up rather than down. Then sanity check the result against the ratios in the valet pricing guide, which is where this number turns into money: crew size times shift hours times your loaded labour cost is the floor your quote has to clear. If your cycle-based crew size and your ratio-based crew size disagree by more than one person, the cycle number is the one to trust, because it is measured and the ratio is borrowed.

It is also worth putting in a proposal. When you are bidding against operators who quote a headcount with no reasoning attached, showing the venue that your crew size came from a measured cycle time at their distance to their lot is a real differentiator, and it costs you nothing to include. There is more on that in the piece on bidding a restaurant valet contract.

Where the software fits, and where it does not

ParkingPro does not calculate staffing. There is no crew size calculator, no cycle time timer, and no scheduling module, and we are not going to imply otherwise.

What it does have is the raw material for step 7 above. The shift report shows how many vehicles each attendant handled during a real shift, with the times attached. Divide that by the hours they worked and you have your actual throughput, from your actual lot, with your actual crew, instead of an average borrowed from somebody else's hotel. Run it for a slow night and a busy night and the gap between the two tells you more about your operation than any published ratio will.

Timestamps on tickets are the other half. Because entries and exits are logged, you can see when the arrival curve actually peaked rather than when you remember it peaking, which is usually a different time and is the input your crew schedule should be built on.

Straight about the limits. The report gives you vehicles per attendant per shift, not a broken-out park cycle and retrieve cycle. For those you still need the stopwatch, at least once. There is no automatic staffing recommendation, no labour scheduling, no payroll, and no licence plate recognition. Tax authority integration exists today only in the Dominican Republic and Mexico; in the US you get standard receipts.

Every cycle time and staffing figure on this page is attributed to a named public source consulted on 17 August 2026, or is arithmetic derived from one with the working shown. There is no industry productivity survey for US valet operations that we could locate, and the operator figures in circulation describe individual operations rather than a benchmark. Walking speed figures are US traffic engineering design values for signal timing, used here as a planning proxy and not as a measurement of any specific person. Treat everything here as a starting point for your own measurement. This article is general operational guidance and not legal, safety or employment advice. ParkingPro Cloud is a product of Abalon LLC.

How many cars can one valet attendant park per hour?

Using the only published cycle time figures we could find, roughly 9 to 12 movements per hour, where a movement is either parking one car or retrieving one car. That comes from an operator estimate of 5.5 minutes to park a vehicle and walk back, 4.25 minutes to retrieve one, and about 50 productive minutes in each 60 minute hour. It assumes a short walk to the lot. Every extra 100 feet of distance takes the number down.

How many complete cars per hour is that, counting both drop off and pickup?

About 5. A guest who arrives and later leaves consumes roughly 9.75 minutes of attendant time in total, 5.5 to park and 4.25 to retrieve. Divide 50 productive minutes by 9.75 and one attendant supports about 5.1 complete guest visits per hour. This is the number to use when you are sizing a crew against expected covers or guest count, and it is roughly half the movements per hour figure.

Is there an official industry standard for valet attendant throughput?

We could not find one. The National Parking Association runs a Valet Operations Certificate and a Certified Valet Attendant credential, but we found no published throughput benchmark from any US parking association, and no rate or productivity survey for the valet sector. The numbers in circulation come from individual operators writing about their own operations. Treat them as informed starting points, not as a standard.

How does distance to the parking lot change the number?

It is the single biggest driver, because the return leg is walked and walking is slow. At a US traffic engineering design speed of 3.5 feet per second, roughly 210 feet a minute, a 600 foot walk back adds nearly 3 minutes to every single cycle. With 3 minutes of fixed handling, a 150 foot lot supports about 12 movements per hour and a 1,000 foot lot supports about 5. Past roughly a quarter mile, driving attendants back in a cart or shuttle beats hiring more of them.

Why do some operators claim 30 or more cars per hour per attendant?

Because they are quoting a burst rate, not a sustained rate. One published account of amphitheater valet describes a team of eight processing 40 to 50 vehicles in the first ten minutes after a show, which works out to roughly 30 to 37 movements per attendant per hour. That rate depends on vehicles pre-staged before the crowd moves and on every person being on the floor at once. It cannot be held for an hour, let alone a shift.

How do I measure the number for my own operation?

Time ten park cycles and ten retrieve cycles with a stopwatch during your busiest 30 minutes, measuring from the moment the attendant takes the key to the moment they are back at the stand and free. Take the median rather than the average, since one manual transmission or one blocked-in car will distort a mean. Then multiply your productive minutes per hour by 60 and divide by the median cycle time.

Does one attendant per 15 vehicles mean 15 cars per hour?

No, and this is the most common confusion. Ratios like one attendant per 15 vehicles at peak or one per 25 to 30 active vehicles at a resort describe how many cars sit under one person's management, which is a stock. Cars per hour is a flow. A crew can look correctly staffed by the ratio and still build a queue if the arrivals all land inside twenty minutes.

Should my attendants run?

Running roughly halves the walk back and can add close to 30 percent to throughput at a 600 foot lot. We found no published policy from a major US operator or any association guidance either requiring or prohibiting it, so this is your call rather than an industry rule. The two constraints are real: injuries on wet pavement are a workers' compensation problem, and nobody sustains a jog into the fifth hour of a shift. Whichever way you go, write it down in the handbook before you need it.

Stop guessing your throughput. Measure it.

ParkingPro's shift report shows how many vehicles each attendant handled on a real shift, with timestamps, so you can calibrate this number for your own lot instead of borrowing an average. Plans start at US$19 and include 14 days free.

Create free account →