Three financial tests to run before you hand in notice
Runway counted after tax reserve, priced offers instead of expressed interest, and what your employer really spends on you: three numbers to check first.
The standard rule is six months of expenses in the bank. It sounds cautious and it usually isn't, because both halves of it are measured wrong: the savings figure is overstated and the income it is meant to replace is understated.
Take the second half first. In March 2026, benefits accounted for 30.1% of total compensation for US private-industry workers — $14.01 of every $46.60 per hour that employers spent, against $32.60 in wages and salaries (BLS, Employer Costs for Employee Compensation, released 12 June 2026). Your salary is roughly seventy cents of every dollar your employer spends on you. The other thirty bought paid leave, insurance, a retirement contribution and payroll taxes. Part of that you are about to buy at retail, and part of it simply stops.
What follows is the go/no-go arithmetic only. How to leave without damaging the relationship is a separate problem with its own mechanics, and the wider question of what independence does to a career belongs in the milestone guide.
Test one: runway counted after the money that was never yours
Most people compute runway as bank balance divided by monthly costs. That treats the tax you already owe on income you already earned as though it were savings. It isn't. It is a liability sitting in a current account.
Take an illustrative UK case — these are made-up figures to show the shape of the calculation, not observed data. Say you hold £26,000 in cash and your fixed costs run £3,200 a month. Bank balance over costs gives 8.1 months. Now suppose £7,400 of that balance is tax: a balancing payment for 2025/26 and a first payment on account, both due 31 January 2027, with a second payment on account on 31 July 2027 (HMRC Self Assessment deadlines). Spendable cash is £18,600, and runway is 5.8 months. The same correction applies in the US against the estimated-tax dates of 15 April, 15 June, 15 September and 15 January (Form 1040-ES).
There is a second deduction most people miss, and it is structural rather than arithmetic. Work delivered in your first month is invoiced at the end of that month and paid on terms. Under the EU late payment directive the default B2B period is 30 calendar days from receipt of the invoice (Directive 2011/7/EU, Article 3), and 30 days is a floor rather than a norm — plenty of clients run on 45 or 60. So your first real cash lands in month three at the earliest, and that is if nothing slips. Runway has to cover the gap between starting work and being paid for it, not the gap between quitting and starting work.
Compute it honestly with the runway calculator, and if you are unsure how much of your balance is reserve rather than savings, the tax set-aside calculator will separate them for the US, UK or Germany. A rolling forward view is the thing that actually changes behavior here — 13 Weeks Out exists because a bank balance tells you where you are and a forecast tells you where you are going. The spend and expense tooling you had at work was built for a finance department, and what replaces it needs to be a Ramp alternative sized for one person rather than a smaller version of the same thing.
Pass line: cash you could spend tomorrow without creating a liability, divided by fixed monthly costs, of at least six months. Six is not a measured figure — it is derived from the payment lag above plus a conservative ramp, and it should go higher if test two is thin.
Test two: count priced offers, not expressed interest
Almost everyone quitting has some version of "a few people said they'd use me." That sentence has no predictive content. Interest is free to express and costs nothing to withdraw, and it is systematically inflated by the fact that the person expressing it likes you.
The unit that means something is a priced offer: a named buyer with budget authority received a scope, a price and a date, and did not say no. It doesn't have to be accepted yet. It has to have a number attached, because the number is the point at which a conversation converts from goodwill into a purchasing decision, and that conversion is where estimates break.
Make the offers while you are still employed. This is also the moment to check that your evidence is current rather than remembered. In the eight months after ChatGPT launched, job posts in writing and coding fell 21% relative to manual-intensive work, and image-creation posts fell 17% after image models shipped (Demirci, Hannane & Zhu, CESifo WP 11276). Fiverr reported 2.7 million annual active buyers in Q2 2026, down 21.9% year over year, while annual spend per buyer rose 15.6% to $368 (Fiverr Q2 2026 results). Fewer buyers, each spending more, is a market consolidating upward. Demand evidence from 2022 is describing something that no longer exists, and the way to find out where you sit is to run a cheap priced test before you commit.
Pass line: at least two priced offers accepted or in live negotiation in the last 90 days, from buyers who are not your current employer. The exclusion matters, because an offer from your employer tests their convenience, not your market.
Test three: the salary is about 70% of what you cost
The replacement number is not your gross salary. It is what your employer spends to have you, minus the part that never reached you, plus the overhead you are about to start carrying yourself.
Work it in the UK first, with illustrative figures. On a £60,000 salary, employer National Insurance at 15% above the £5,000 secondary threshold is (£60,000 − £5,000) × 15% = £8,250 (HMRC rates and thresholds for employers 2026 to 2027). Employer pension at the automatic-enrolment minimum of 3% of qualifying earnings between £6,240 and £50,270 is 3% × £44,030 = £1,321 (GOV.UK workplace pensions). Total employer cost: £69,571.
Of that, the £8,250 of employer NI was never yours and doesn't need replacing. What reached you was £60,000 plus a £1,321 pension contribution, so £61,321 — and, separately, 5.6 weeks of statutory paid holiday (GOV.UK holiday entitlement) and statutory sick pay cover at £123.25 a week, which as a self-employed person you will not have.
Now convert that into a rate. Fifty-two weeks minus 5.6 leaves 46.4 working weeks, because taking the holiday you used to be paid for now means not billing. At 25 billable hours a week that is 1,160 hours. Add £6,000 of your own overhead (software, accounting, insurance, equipment) and you need £67,321 of revenue over 1,160 hours, or £58 an hour, before a penny of profit. The salary you left looked like £60,000 over 1,950 contracted hours, or £30.77. The replacement rate is 1.9 times the number your payslip implied.
In the US the shortcut is the ratio: total compensation over wages is $46.60 ÷ $32.60 = 1.43, so a $100,000 salary corresponds to roughly $143,000 of employer spend. Strip out the legally required component of $3.38 an hour, which is largely the employer half of payroll taxes, and you'll pick most of that up anyway through self-employment tax at 15.3% rather than the employee's 7.65% (IRS, self-employment tax).
Pass line: your evidenced rate multiplied by realistic billable hours clears the replacement number within twelve months. Not your aspirational rate. The one somebody has agreed to.
What a pass actually looks like
Test | What you count | What people count instead | Pass line |
|---|---|---|---|
Runway | Cash you can spend without creating a liability, over fixed monthly costs | Bank balance over monthly costs | ≥ 6 months, and ≥ 9 if test two is thin |
Demand | Priced offers to named buyers with budget authority, last 90 days | Conversations, encouragement, expressed interest | ≥ 2 accepted or in live negotiation, none from your current employer |
Replacement cost | Employer total spend, less employer payroll taxes, plus your own overhead | Gross salary | Evidenced rate × realistic billable hours clears it inside 12 months |
Two out of three is not a pass, but which one fails tells you what to do next. Failing test one is a timing problem and time fixes it. Failing test three is a pricing problem, and the fix is to requote before you resign, not after — the rate calculator will take a target income, billable hours and a tax set-aside and give you the floor. Failing test two is the serious one, because it means the thing you are about to bet a salary on has not been observed.
The line where the answer flips
Take m as your monthly revenue at the prices buyers have actually agreed to, and f as your fixed monthly costs. Let t be the number of months at your current rate of winning priced work before m exceeds f. Add two months for the invoice-to-cash lag from test one. If your post-reserve runway in months is less than t + 2, the answer is not yet — and the lever that moves it is test two, not the savings account. Saving for another quarter while your demand evidence stays at zero buys you a longer runway to nowhere.
If runway exceeds t + 2 and you have the two priced offers, the remaining question is only whether you priced them against the replacement number or against your old salary. Most people price against the salary and spend their first year wondering why the money feels worse.
Worklyn's runway calculator separates spendable cash from the tax you have already earned but not yet paid, which is the difference between the eight-month figure and the six-month one.