What to charge as a freelancer: your floor first, the model second

There is no rate table you can look up, and that is worth saying before anything else, because most of the pages promising one have invented it.

The closest thing to a published figure is an average hourly rate of €103 in 2026 from freelancermap's annual study: 5,400 freelancers, DACH region only, skewed heavily toward IT and engineering contractors. If you write copy in Manchester, it tells you nothing. IPSE's UK day rates do have a documented method (1,500 respondents, four quarters of 2024), but the figures are members-only, so you cannot read them. Neither Upwork nor Fiverr publishes a rate-by-discipline table with a stated methodology. Every "average rate by skill" chart ranking for this question is unsourced.

So the number is computed, not looked up. What follows is the arithmetic that produces your own floor, and then a rule for choosing which pricing model to put that floor inside.

Your floor is arithmetic, and it moves more than you think

Start with the number below which work is a hobby. Here is the calculation, using illustrative figures rather than observed data, in dollars for a US sole trader.

Start with the profit you need the business to make: say $75,000 before tax. Self-employment tax is 15.3% on 92.35% of net profit, so $75,000 × 0.9235 × 0.153 is $10,597, and that is before a cent of federal income tax against a 2026 standard deduction of $16,100 for a single filer. Add business costs of $9,000 a year for software, insurance, accounting and hardware. Revenue required: $84,000.

Now the part that decides everything. Working 46 weeks at 40 hours gives 1,840 hours. If 55% of those reach a client, you have 1,012 billable hours, and $84,000 ÷ 1,012 is $83 an hour. If your billable share is actually 45%, which is common for anyone doing their own sales, admin and chasing, you have 828 hours, and the same $84,000 needs $101.45 an hour. Your floor moved 22% on a variable most freelancers have never measured.

That is also why the day rate arithmetic misleads. At the 55% figure, $83 × 8 is a $664 day, but a day rate assumes eight billable hours inside a working day, and the whole point of the 55% number is that they do not exist. The rate calculator runs this from your own inputs (target income, billable hours a week, weeks off, monthly expenses, tax set-aside) and returns both figures.

UK readers should run the same structure with different constants: Class 4 National Insurance at 6% on profits between £12,570 and £50,270 and 2% above, income tax at 20% and then 40% from £50,271, and one threshold that changes your pricing rather than just your tax bill — VAT registration at £90,000 of rolling twelve-month taxable turnover. If your clients are businesses that reclaim VAT, crossing it costs you administration. If they are consumers or exempt bodies, crossing it makes you 20% more expensive overnight, and the price you set at £80,000 of turnover should already anticipate that.

That gives you a minimum. It does not tell you what to sell: an hour, a day, a deliverable, or a share of an outcome. Pricing advice usually answers that with a ladder: beginners bill hourly, professionals quote fixed, the enlightened charge on value. The ladder is mostly fiction. In a study of more than 1.8 million job posts on Freelancer.com between January 2022 and April 2024, software-integration tasks carried 3.82 times the budgets of content-generation tasks, on the same platform in the same period. No pricing model closes a gap that size, and none creates one. What the four models actually differ on is which risk you agree to hold.

Hourly hands the estimation risk to the client and taxes your speed

Hourly is the honest model. The client pays for what the work took, so a job that runs long costs them more, and neither party has to guess correctly in advance.

Its failure mode is structural and well known: it prices your slowness. Get 30% faster at the thing you do most and your income falls 30%, unless you raise the rate to compensate, which means every efficiency gain has to be re-negotiated. It also caps you at the number of hours a human can sell, and it invites the client to audit the meter rather than the outcome, which turns every long day into a conversation.

The less obvious failure is that hourly only works if the record behind it is trustworthy. Hours you forget to log are hours you have donated, and the ones most often missed are the ones nobody starts a timer for. Comparing time trackers on stopwatch features misses this entirely; the question is whether a stopped timer becomes a line on an invoice, which is the axis the Clockify alternative comparison is built around.

Hourly is right when the scope genuinely cannot be fixed: discovery, ongoing advisory, incident work, anything where the shape of the job is what you are being hired to find out.

The day rate is an hourly rate with the overrun hidden inside it

A day rate looks like a simplification of hourly and behaves like a fixed price for a unit of one day. That difference is where it goes wrong.

The billed day and the worked day are not the same length. If your day rate assumes eight hours and you routinely work nine and a half on client days, you have taken a silent 16% discount, because 8 divided by 9.5 is 0.84. Nobody notices, because the invoice line still says one day. Multiply that across 120 client days a year and you have given away roughly 180 hours without a single conversation about price.

Day rates also encourage the client to buy in whole units and then use them badly. A half-day of actual work rounds up to a day on your invoice and rounds down to nothing in the client's mind, so they book you for days they have not prepared for. That is fine on your side of the ledger, until it becomes the reason they think your day rate is expensive.

Use it where your presence is the product: embedded work, on-site delivery, reserved capacity, workshops. If you want to see what a given day rate implies as an hourly figure or an annual salary equivalent, the day-rate-to-salary converter moves between the three, which is also the fastest way to sanity-check a rate a client has proposed to you.

Fixed price sells the client an insurance policy you priced without an actuary

A fixed price is a promise that the work will cost them the same whether it takes you 60 hours or 160. That promise has a value, and clients pay for it, which is why fixed-price work can carry a premium over the same job billed hourly.

You are the insurer. Every hour of overrun comes out of your margin, and unlike an insurer you are writing one policy at a time, so you cannot spread the variance across a book of business. One badly estimated project is not offset by nine good ones; it just costs you a project.

The failure mode is not the overrun itself, it is quoting a fixed price for work whose estimate you cannot bound. If you have done this exact job several times and your estimates land within 15% of actual, fixed price is a good trade and the premium is real profit. If you are estimating something novel, you are selling insurance on a risk you have not priced, and the honest version of that quote includes a contingency the client will not want to see.

Fixed price also requires the boundary to be written down. A fixed fee with an unbounded revision count is not a fixed price; it is an hourly engagement where you agreed not to send the second invoice. The revision cap, the approver list and the exclusions belong in the proposal document alongside the number, because the number is meaningless without them.

Value pricing needs an attribution you usually cannot get

Value pricing means charging a share of the economic outcome rather than the cost of production. Done properly it is the highest-margin model available, and it is genuinely correct in a narrow set of cases.

Those cases have a specific shape: a single, measurable financial line that your work moves, agreed with the buyer in advance, where nobody else's contribution can plausibly claim the credit. A landing page rebuild for a business with one traffic source and one conversion event qualifies. The same rebuild for a company simultaneously running a rebrand, a paid campaign and a new sales hire does not, because when revenue moves you cannot show it was you, and when it doesn't the client will be certain it wasn't.

The second problem is that value pricing requires the client to disclose the value. Most will not tell you what the project is worth to them, and the ones who will are often the ones for whom the number is small. Asking for the figure that sets your own price puts you in a negotiation where the other side controls the only input.

Treat value pricing as a modifier rather than a model. Anchoring a fixed fee against a stated business outcome is available to almost everyone; charging a percentage of that outcome almost never is.

Matching the model to the project

One table, and it is the only one you need. Read the left column against the project in front of you.

That last row deserves more than a row, because the four common retainer structures behave very differently and two of them do not smooth cash flow at all. Which retainer structures actually stabilize income works through the mechanics.

A note on presenting the number, since the tooling shapes it. Proposal suites price at $15 to $59 per user per month at Bonsai's published rates, and they will happily print any figure you type. What they do not do is tell you whether that figure cleared the floor you calculated above, which is the gap the Bonsai alternative comparison is about. Presentation does matter. Two or three options beat a single number, because the conversation shifts from whether to buy to which one. But presentation applied to a price below your floor only sells the loss faster.

The test that tells you when fixed price is the wrong answer

Estimate the job twice: your likely case and your pessimistic case, the one where the client's approvals arrive late and the second round of feedback contradicts the first.

Then do one division. Take the fee you are about to quote and divide it by your floor rate. That gives the hour count at which the project stops paying. Quote $8,250 against a floor of $83 and the break-even is 99 hours, because 8,250 ÷ 83 is 99.4. If your pessimistic estimate is 105 hours, that fixed price returns $78.57 an hour in the bad case, which is below your floor, and you have written an insurance policy at a loss-making premium.

The rule: quote fixed price only when your pessimistic estimate still clears your floor. When it doesn't, you have three honest options — raise the fee until it does, narrow the scope until the pessimistic case shrinks, or bill time and hand the estimation risk back to the client. Picking the model you wish you were mature enough for is not among them.

Whichever you choose, the number that tells you whether it worked is not the one you quoted. It is the effective hourly rate the project actually returned, measured at close, across every hour the job consumed.

Worklyn's CFO Mode includes Rate Check, so the floor you worked out once gets compared against what you are actually quoting rather than filed and forgotten.