Stress-testing virtual assistant agency prices before clients do

What to take away

No rate, price or margin is stated on this page. The letters below stand for your own figures.

What the tests borrow from your records

Every test below reads the same four figures, taken from a real month. L is the fully loaded monthly cost of one delivery person, H is paid hours, D is delivered hours on client work, and C is L divided by D.

This page does not rebuild those figures. The price list and profit model shows how they become rates. Here they only feed the tests, so a guess in any of them makes a test pass when it should fail.

H is what you buy and D is what the client receives. Training, internal meetings, absence, gaps between assignments and waiting on a client's answer all sit between them. Utilization is D divided by H, and a real month beats a plan.

Test one: the bad month

Recompute the price with utilization down and one client paused. If it still covers L, it holds. If it only works at full utilization, you have priced a perfect month.

Run this on every price you publish, not only the headline one. A line that survives the bad month on paper but breaks on the page of assumptions is telling you which input you guessed, and that input is the next thing to measure.

A worked failure shows how this reads. An agency sells reserved hours to three clients. One pauses, and the assistant's paid hours stay the same while delivered hours fall. C rises, the reserved-hours price no longer covers L, and the line that broke first is idle reserved capacity, so that line is the one to reprice.

Where each pricing model breaks first

The bad month hits each way of selling differently, because each converts the floor into a month of product in its own way. Write down the assumption you made, since it is the one to revisit when the test fails.

Test two: the short client

Model a client who signs, is onboarded, and leaves after a short relationship. The first month of a new assignment consumes far more supervision and clarification than the tenth, so that client carries the heaviest cost for the shortest time.

If the client loses money, one of three things is wrong: your notice period, your setup step or your loading. Fix one of the three rather than hoping for long tenures.

Either price a separate setup step, or load the recurring price enough that a short relationship does not lose money. The onboarding workflow shows which phases consume that supervision.

Both tests run on the whole price, so the costs above C must be in it first: sales time, administration, bad debt, idle reserved capacity, absence cover and profit. Turn each into a percentage of C, and the margin layers show where each lands.

Idle reserved capacity and profit you set deliberately, rather than discovering them at year end. If you sell availability, some of it goes unused and you have paid for all of it, so that cost belongs in the price of the availability product.

Test three: the discount request

Keep two numbers: the list price, and the floor, which is C plus the loading you consider non-negotiable. The floor is written down and known to whoever is allowed to discount.

One person approves exceptions, and every exception is recorded with a reason. Discounts granted by whoever happened to be on the call stay invisible until the quarter closes, and by then they are precedent.

A buyer who asks for a discount in a live conversation meets the person holding the floor. That conversation is the subject of the sales and proposal guide.

A discount that stays above the floor is a decision. One that drops below it is a loss you have agreed to in advance, so the written reason should say what the agency gets in return.

What to repair when a test fails

Each test points at a different part of the price, so the repair should match the failure. Changing the headline rate after every failed test hides the real cause and trains buyers to wait for the next cut.

Re-run the failed test after the repair, with the same inputs. A repair that only passes with new, friendlier inputs has not fixed anything.

Keep the records that make the test repeatable

Pay is hard to estimate at the start. The Bureau of Labor Statistics runs a program that produces wage estimates for about 830 occupations, so treat it as context for L, not as an answer.

Utilization has no published source. Estimate it low, measure from month one, and correct it. The IRS says good records help you monitor the progress of your business, so keep hours by client and by line alongside the basics.

When a quote leaves the building, the quote template lists the fields it needs, so the price you tested is the price the client sees.

Common questions

What if the price I get is higher than what competitors advertise?

Then either their cost base differs from yours, or they are advertising an entry tier against your full one, or they are losing money. Check what their number includes before you move yours.

Which check should come first?

Start with the bad-month check, because it needs only figures you already hold. The onboarding check needs a view of how much supervision a new client takes, which comes from your own records of a real start.

Should I raise prices for new clients only?

It is the least disruptive way to start, but it creates two populations and a conversation you will eventually have anyway. Plan the second step at the same time as the first.

How often should I redo the tests?

Whenever pay changes, whenever utilization moves for two consecutive months, and once a year regardless.

Can I price by the value the client receives?

You can try, but you need evidence of that value from the client's own records, and few will share it. The floor still applies whatever the story above it.