How many hours a week does a virtual assistant usually work?

There is no usual, which is the problem this calculator exists to solve. Engagements run from a few hours a week to full time. What matters is your own task list: count the recurring work, add the check-in time, divide by a realistic utilization factor, and the number falls out. Most first hires land somewhere between 10 and 25 hours a week, then grow once the owner sees what actually transfers.

Should I start part-time or full-time?

Start at the number your task list produces, not below it. Deliberately underbuying is a common instinct and it usually backfires: a person hired for eight hours to do fifteen hours of work falls behind in week two, and the owner concludes that delegation does not work. If your honest count is near a boundary, hire at the lower band with a written plan to widen the brief at 60 or 90 days.

Why does the calculator add hours I did not enter?

Two adjustments. Check-in and handover time is added because a standup is paid time on the assistant's side even though it produces no task output. Then the total is divided by a utilization factor, because nobody spends every paid minute on task work. Breaks, tool problems, waiting on you for an answer and general friction are real. Buying paid hours as though they were pure task hours is the most common sizing error we see.

My estimates feel like guesses. Does that make the number useless?

No, but it makes it a first draft. Research on the planning fallacy has found for decades that people underestimate their own task durations even for work they have done many times, and that the reliable correction is to check the estimate against what actually happened rather than against how the task ought to go. Track one ordinary week against these rows before you commit to a contract length, and the number gets sharp fast.

The tool says my list is two roles. Do I have to hire two people?

Not immediately, but you should not advertise it as one job. When two skill families each carry eight or more hours a week, you are screening for two specialisms at once, and the pool of people genuinely strong in both is small. The practical route is to sequence: hire for the heavier family first, let that person absorb the procedural parts of the second family, and hire again when the remaining specialist work is clearly its own job.

How much of my own time will supervising cost?

More in month one than you expect and less by month three than you fear. Plan for roughly a fifth of the delegated hours going back into review, correction and answering questions in the first month, settling toward a twentieth once SOPs exist and the assistant has seen the edge cases. The net hours you get back are still substantial, but they are not the full delegated total, and a plan that assumes otherwise is what makes month one feel like a mistake.

What if my volume is seasonal?

Size for a normal week and name the overflow separately. If your December is triple your March, hiring for December means paying for idle capacity nine months a year, and hiring for March means a December that breaks. The usual answer is a steady-state hire plus a documented plan for peak: pre-written SOPs so a second person can be added quickly, and a queue that degrades gracefully rather than silently.

Virtual Assistant Hours Calculator: How Many Hours Do You Need? (Free)

Q: Does 30 hours a week count as full time?

For several practical purposes in the United States, yes. The IRS treats an employee averaging at least 30 hours of service a week, or 130 hours a month, as full time for employer shared responsibility purposes. The Bureau of Labor Statistics uses 35 hours as its own dividing line in the Current Population Survey. Neither definition governs a contractor engagement abroad, but they are useful anchors when you are deciding what to call the role.

The question that comes before the price

Almost every tool built for people hiring an assistant starts in the same place: choose a role, choose part-time or full-time, and here is your monthly cost. Our own cost calculator does exactly that, and so does everyone else's. It is a useful answer to a question most buyers cannot yet answer, because the input it depends on is the thing they were trying to work out. How many hours do I actually need? Nobody tells you. You pick a number that sounds sensible, and then you find out whether it was right by living with it for six months.

The two ways of getting it wrong are not symmetrical. Overbuying is visible and mildly embarrassing: you pay for hours that go unused, you notice within a month or two, and you can usually adjust. Underbuying is worse and much harder to see. An assistant hired for twelve hours to do a job that genuinely takes twenty does not tell you the maths is wrong. They fall behind quietly, prioritise whatever shouted loudest that morning, drop the slow-burning work that has no deadline attached, and start to look like a disappointing hire. The owner concludes that delegation does not work for their business, and goes back to doing everything themselves with a story that now feels evidence-based. The failure was in the sizing, not the person.

This page is the sizing step. It is deliberately unglamorous: a task list, a frequency, a duration, and three corrections that most estimates leave out. What comes out the other end is a number of paid hours per week, a plain statement of what shape of engagement that is, and a warning if the list you have assembled is quietly two jobs wearing one coat.

Why the clock underestimates the work

Start with the thing that makes every hand-built estimate too low. When you count the work you want to hand over, you count it as a clean list of durations: fifteen minutes for the inbox, ten for scheduling, forty-five for the report. What you are actually handing over is not that list. It is that list interleaved with everything else, arriving at unpredictable times, in pieces.

Microsoft's 2025 Work Trend Index measured this directly across its own telemetry and found that employees using Microsoft 365 are interrupted roughly every two minutes during the working day by a meeting, an email or a notification, which works out to about 275 interruptions a day. The same report found the average worker receiving 117 emails a day and 153 Teams messages, with 57% of meetings happening with no calendar invite at all. Nearly half of employees described their work as chaotic and fragmented, and so did more than half of leaders.

The cost of that fragmentation is not neutral. Gloria Mark and colleagues at UC Irvine ran a controlled study of interrupted work and found something counterintuitive: people finish interrupted tasks in the same amount of time or faster, because they compensate by working more quickly. They pay for it in stress, frustration, time pressure and effort. In other words, a fragmented hour is not simply a slower hour. It is an hour that costs more to produce and cannot be sustained across a whole week at that intensity.

For sizing purposes the practical consequence is simple. Twenty separate five-minute tasks are not one hundred minutes of work, and any model that treats them as such will under-buy. Two habits fix it. Batch small tasks into a queue that gets worked in one sitting, and count the batch rather than the items. Where a task genuinely cannot be batched, because it is a response with a service-level expectation attached, add an allowance to it rather than counting the raw duration.

Your estimate is low, and it is not carelessness

The second correction is uncomfortable, because it applies to you personally and no amount of experience removes it. In 1994 Buehler, Griffin and Ross published a set of studies in the Journal of Personality and Social Psychology establishing what they named the planning fallacy: people systematically underestimate how long their own tasks will take, while estimating other people's tasks reasonably well. In their honours thesis study, students took an average of 55 days against a predicted 33, and fewer than a third finished by the date they had confidently named.

The mechanism they identified matters more than the numbers. People predict by building a plan-based scenario, imagining the task going the way it ought to go, and they discount their own past experience of the same task by attributing every previous overrun to something specific and non-recurring. Last time the invoice run took two hours instead of one, but that was because of the software update, which will not happen again. It will happen again. Something always does.

The same paper found the correction. The optimistic bias was largely eliminated for participants who were prompted to connect their prediction to relevant past experiences. Applied here, that means the defaults in the calculator above are a starting draft and not an answer. The highest-value thirty minutes you can spend before hiring is to open the tool, keep a rough tally for one ordinary week, and correct the rows against what actually happened. Not what should have happened.

Paid hours are not task hours

The third correction is the one that turns a decent estimate into a usable one. A person paid for forty hours does not deliver forty hours of task output, and no serious operator plans as though they will. Breaks happen. Tools break. The assistant sits blocked waiting for an answer from you. A task turns out to need a decision nobody has made yet. None of this is slack in the pejorative sense; it is the ordinary friction of work, and it is present in every role everywhere.

Professional services firms have priced this for decades under the name utilization, and the honest numbers are lower than most people assume. The calculator offers three settings, and the default sits at 85%, which is a reasonable planning assumption for a settled assistant working clear queues. Drop to 75% when the role is genuinely fragmented, when the person is covering live channels with unpredictable arrival, or when they depend heavily on you for unblocking. Raise it toward 92% only when the work is repetitive, batched, and free of interruption, which describes a single high-volume queue and not much else.

The arithmetic is worth seeing plainly, because the effect is larger than intuition suggests. Twenty-five hours of genuine task work at 85% utilization is not a 25-hour hire. It is a 29.4-hour hire. At 75% it is a 33.3-hour hire, which has crossed from part-time into full-time territory without a single task being added to the list. Owners who skip this step do not discover the gap as a number. They discover it as an assistant who is always slightly behind, and they usually blame the assistant.

Reading the number you get

Once the hours are counted, the shape of the engagement follows. The bands below are practical rather than legal, but two real definitions are worth knowing because they anchor what people mean by full time. The IRS treats an employee averaging at least 30 hours of service a week, or 130 hours a month, as full time for employer shared responsibility purposes. The Bureau of Labor Statistics uses 35 hours as its dividing line in the Current Population Survey. Neither governs a contractor engagement with someone in Cape Town or Manila, but they explain why 30 hours reads as a full-time role to most people, and they matter directly if you are weighing up whether the relationship is a contract or employment. Our guide on contractor or employee status for offshore assistants covers that question in depth.

Paid hours	Engagement shape	What it fits	What to watch
Under 10 hrs/wk	Light part-time	A single narrow queue: inbox triage only, or listings only, or invoicing only.	Too thin to hold someone good for long. Either widen the brief before you hire or accept that you are sharing the person.
10 to 20 hrs/wk	Part-time	Two or three related queues owned end to end, on fixed days and fixed hours.	Scattering the hours across five days in fifteen-minute slivers destroys most of the value. Block them.
20 to 30 hrs/wk	Half-time to three-quarter time	A dedicated person with named queues, their own SOPs and real ownership of outcomes.	This is the band where supervision has to become a system rather than a series of interruptions.
30 to 40 hrs/wk	Full-time	A full role with capacity for the unplanned work that always arrives.	If you sized this from a busy week rather than a normal one, you have bought about 15% more than you need.
Over 40 hrs/wk	More than one person	Nothing. One week does not contain more than one week.	Cut the list, sequence it across two hires, or split it by skill family. Do not hand it to one person and hope.

One band deserves more attention than it gets. Under ten hours a week is a real amount of work and a difficult thing to hire for well. The problem is not the money, it is retention and attention: a strong assistant with eight hours of your work will fill the rest of their week elsewhere, and your queue becomes the one that flexes when something else runs late. If your honest count lands there, the better move is usually to widen the brief until it reaches a band that sustains a proper working relationship, or to accept explicitly that you are buying shared rather than dedicated capacity and to set your expectations accordingly.

When the number says two people

The most expensive sizing mistake is not getting the total wrong. It is getting the total right and then asking one person to deliver it across skills that do not travel together. The calculator flags this when two skill families each carry eight or more hours a week, and the flag is worth taking seriously.

Consider a common founder list: ten hours of bookkeeping, ten hours of social media, and eight hours of inbox and calendar. The total is twenty-eight hours, which reads as a comfortable three-quarter-time hire. It is not. A person who is genuinely good at bank reconciliation and genuinely good at community management exists, but they are rare, and the search for them takes far longer than two separate searches would. What usually happens instead is that you hire someone strong in one family and passable in the other, and the weaker half of the job quietly degrades until you stop asking about it.

There are three honest ways out. Sequence the hire: take the heavier family first, let that person absorb the procedural parts of the second family, and hire again when the remaining specialist work is clearly its own role. Split it deliberately: two part-time engagements rather than one confused full-time one, which is more common than founders expect and works well when the queues are genuinely separate. Or cut the list: decide that one family is not ready to delegate yet and keep it, which is a legitimate answer as long as it is a decision rather than an accident. What does not work is writing a job advert that lists both and hoping the market solves it for you.

Month one is not steady state

A new assistant does not deliver their steady-state output in the first month, and planning as though they will is what makes a perfectly normal ramp feel like a hiring mistake. The calculator assumes roughly 60% of steady-state throughput in month one and gives you a shorter first-wave list to match.

The order of that list is deliberate. It is sorted by how little of your context the task needs, then by how many hours it frees. Tasks that are procedural, where a right answer exists and can be written down, transfer in days. Tasks that need your judgement, your relationships or your knowledge of a specific customer transfer in months, and only after the assistant has watched you do them a few times. Handing over the judgement-heavy work first is the single most reliable way to produce a bad first month, because both of you end up doing the same work twice.

Budget for your own time as well, because that is the cost nobody puts in the plan. Expect to spend roughly a fifth of the delegated hours on review, correction and answering questions in month one, settling toward a twentieth once the SOPs exist and the edge cases have been seen. That is a real return, but it is not the full delegated total, and knowing the shape of it in advance is what stops week two from feeling like evidence against the whole idea. Our onboarding checklist generator and SOP generator exist to make that stretch shorter, and the access plan generator handles the system permissions the first wave will need.

Six ways the count goes wrong

These are the errors we see most often when a founder brings us a number they have already worked out. None of them are careless; they are all reasonable-looking shortcuts that happen to be expensive.

The miscount	What it costs	The correction
Counting the task, not the switch	Twenty small tasks at five minutes is not one hundred minutes of work	Batch by queue and count the batch, or add a switching allowance to any task under ten minutes
Sizing from a busy week	A 10 to 20% overhire that shows up as idle time in month three	Size from a normal week and handle month end, launches and peak season as named overflow
Forgetting the exceptions	The refund, the angry customer, the mis-shipped order, the invoice nobody can explain	Count exception handling as its own line, because it is the line that eats the week
Pricing check-ins at zero	Two to four paid hours a week that were never in the plan	Put standups, the weekly review and ad hoc explaining into the model as paid time
Assuming 40 paid hours is 40 task hours	A permanent 15% shortfall that reads as a performance problem	Divide productive hours by a utilization factor before you buy hours
Ignoring the ramp	A month one that feels like a failed hire when it is a normal one	Plan for roughly 60% of steady-state output in the first month and hand over the low-context work first

The exception line is worth dwelling on. Ask an e-commerce operator how long order processing takes and you will get an answer about the normal order, which takes four minutes and is boring. The week is not shaped by normal orders. It is shaped by the mis-shipped one, the customer who paid twice, the address that fails validation, the return that arrives without a reference. Those cases take twenty minutes each and there are more of them than anyone remembers. Count them as their own row, with their own frequency, or the estimate will be wrong in the direction that hurts.

Seasonality and the overflow plan

If your volume swings hard across the year, do not average it. An average is the one number that is wrong in both directions: it overbuys for nine months and underbuys for three. Size the hire from a normal week and treat peak as a separate, named plan.

A workable overflow plan has three parts. Write the SOPs before peak rather than during it, so that a second person can be added in days rather than weeks. Decide in advance which queues are allowed to degrade and how, because something will, and choosing beforehand is better than discovering it at the worst moment. And agree the trigger explicitly with your assistant, so that asking for help at 130% of normal volume is the expected behaviour rather than an admission of failure. The alternative, which is the default, is an assistant who absorbs the extra silently until something visible breaks.

What to do with the number

Once you have hours you believe, the rest of the sequence is straightforward. Price it with the cost calculator and check what the same hours cost in different markets with the salary by country comparison. If the work needs live overlap with your day rather than asynchronous handoff, the time zone overlap calculator will tell you which markets can supply it. Turn the task list into a posting with the job description generator, screen against it with the interview questions generator, and set the targets you will judge the hire against with the performance scorecard. If you want to sanity-check whether the return justifies the spend at all, the ROI calculator does that arithmetic, and the contract template covers the paperwork once you have chosen someone.

It is also worth saying what the number is not. It is not a promise about how much better your week gets, because that depends on which hours you free rather than how many. Four hours removed from the fragmented middle of your day, the part that keeps getting interrupted, is worth considerably more than eight hours removed from a Sunday evening you were spending on admin anyway. When you look at the first-wave list, ask not only which tasks are largest but which ones are breaking your concentration most often. Those are the ones worth handing over first even when the hours are modest.

And if the list you have built genuinely does not fit one person, that is useful information rather than a problem with the tool. Browse the roles we source, the industries we support and common use cases to see how other teams split the work, or bring us the output and we will tell you on one call whether it is one hire, two, or a smaller first step than you thought.

Sources

The interruption and communication figures come from Microsoft's 2025 Work Trend Index special report and from the CHI 2008 study by Mark, Gudith and Klocke at UC Irvine. The estimation research is Buehler, Griffin and Ross in the Journal of Personality and Social Psychology. The full-time thresholds come from the IRS and the Bureau of Labor Statistics. The task durations, utilization bands, ramp assumption and split threshold in the calculator are our own planning defaults, drawn from the roles we place, and they are meant to be overwritten with your own measurements.

Virtual Assistant Hours Calculator