Should I go freelance in the UK? An honest financial check.

Executive Summary & Verdict

Promising — 7/10. Serve the notice. This plan works not because the market is kind but because three of the things that usually sink a freelance launch are already removed: revenue committed before you resign, a second income in the household, and fixed costs near zero.

The arithmetic is in section 6 and it is close but sound. To stand still — to cover the £1,747 a month your household needs from you after your partner's salary — you need roughly £28,800 of revenue a year, which at £480 a day is 60 billable days. Your two committed clients are worth about £32,540. You are above the survival line before making a single sales call, which is unusual and is the basis of the score.

Two things keep this off a Strong. Matching your old package is a much higher bar than surviving: about £470 a day across 165 billable days, and 165 is the honest number, not 260. And the first substantial Self Assessment payment arrives as a balancing payment plus a payment on account in the same twenty-four hours — roughly £28,430 due on 31 January 2029. Reserve for it from invoice one, or the third year of a working business will feel like a crisis.

Idea Validation

The premise is validated in the only way that counts: two organisations have said they will pay. Everything else in a freelance plan is a hypothesis; a committed client is evidence.

What has not been validated is the differentiator. "Fixed-price rescue of half-built Azure platforms" is a narrow, sellable proposition — and neither first client is buying it. Northbank is buying eight days a month of a competent pair of hands. Kestrel is buying a greenfield build. Your positioning and your pipeline are two different businesses, and that gap closes one of two ways: you sell the rescue offer within six months, or you drift into day-rate contracting, which is a decent living and a much weaker business.

The honest risk in the answers is "no prior experience of selling". Your technical risk is nil. Your commercial risk is total. Scoping a fixed-price engagement, writing a proposal that survives procurement, and asking for money when an invoice is forty days late are three skills you have never used — and the third is where solo consultants actually fail.

Market & Target Customer

Your market is not "UK companies needing data engineering". It is the band of firm that has bought cloud data tooling, has one or two people trying to run it, and has no platform engineer. Below twenty staff there is no budget. Above two hundred there is an internal team and a preferred-supplier list you will not join as a sole trader.

What to verify in the first sixty days

• Client size, deliberately. Both committed clients are small private-sector companies. That is not a coincidence to waste — it materially changes your tax position and your risk, for reasons in section 11.
• Where the rescue work originates. Rarely the company itself — it is the implementation partner whose project overran. Map the five North West firms that create this work and go to them, not to end clients.
• Whether the buyer is technical. A finance director buying a "reporting warehouse" and a lead developer buying "pipeline rescue" need entirely different proposals.

One structural point. Contract day rates for data engineers span roughly £375 at the tenth percentile to £663 at the ninetieth, around a £500 median [1]. That spread is not mostly about skill; it is about who you sell to and whether you sell days or outcomes. Your £480 anchor is the right place to start and the wrong place to stay.

Competitive Landscape

You have three categories of competitor and only one of them is another freelancer.

The boutique consultancy is your rival for Kestrel-type work. A five-person Manchester data shop will quote the same build at two to three times your price and win, because it can show a team, a case study and someone to call when you are on holiday. You beat it on price, on speed of starting, and on the client dealing with the person who writes the code.

The contractor market is your rival for Northbank-type work, against people with ten years of it and an agency relationship. Day-rate work is a commodity market with a published price [1]; there is no differentiation in it, which is the argument for the fixed-price offer.

The in-house hire is the one you lose to silently. A client choosing between your six-week engagement and a permanent junior engineer at £45,000 [1] will often choose the hire, because a salary is a familiar cost and a consultant's invoice is scrutinised. Price against a hire's first six months, not against another freelancer's day rate.

Where you have a hard-to-copy edge

• You can quote a fixed price on a rescue because you have seen enough broken platforms to scope one. A consultancy cannot risk that; a junior freelancer cannot estimate it.
• No overhead. With no staff and no premises your break-even is a fraction of any agency's — so you can take a £9,500 job profitably and they cannot.
• Availability. "I can start in three weeks" wins more small-company work than any credential.

Business Model & Monetisation

The model you have described is a hybrid. Here that is the right answer, but only if you are deliberate about which part is the business and which part is the float.

Day-rate work is your float. It smooths cash, requires no selling, and has a hard ceiling: income is hours multiplied by a published rate. You cannot pass about £80,000–£90,000 a year on it without raising the rate, and you cannot take a holiday without cutting revenue.

Fixed-price engagements are the business. A six-week rescue priced at a deliverable decouples your fee from your hours; the second time you scope one you do it faster and the margin is yours. This is the only version of the plan with a path past a senior salary.

Revenue lines, in the order they should mature

• Anchor retainer. Northbank's eight days a month. Keep it, do not grow it, and never let it exceed half your revenue past month twelve.
• Fixed-price builds and rescues. Two to four a year at £9,000–£20,000 — where the margin and the positioning live.
• A small monthly support agreement attached to every build you finish. Three clients at £450 a month is £16,200 that arrives whether you work or not, and it is the most valuable line you can add in year two.
• Subcontracting to the boutiques you cannot beat. Lower rate, zero sales cost, useful gap filler.

Note what is absent: productising into software, building a course, hiring. All three are the standard advice and all three are wrong for at least two years.

Finances & Budget

Three numbers matter and people conflate them: the rate you need to survive, the rate you need to match the old job, and the days you will actually bill.

The billable-days number is not 260

Start with 260 weekdays and take things away. Eight bank holidays. The twenty days of leave you used to be paid for — statutory entitlement is 5.6 weeks, about 28 days including bank holidays, and you now fund every one yourself [1]. Five days of illness, with no employer sick pay. Roughly one day a week on selling, invoicing and bookkeeping, which is 46 days. And an allowance for the weeks between engagements.

260 − 8 − 20 − 5 − 46 − 16 = 165 billable days in a steady year. Plan year one at about 130. Build on 220 and everything downstream is fiction.

The rate that matches the old job

Your £68,000 salary is worth about £50,000 net in 2026-27 in England — £12,570 personal allowance, 20% to £50,270, 40% above it [2], less employee Class 1 National Insurance at 8% on earnings between £242 and £967 a week and 2% above [3]. Add the employer pension: the statutory minimum is 3% of qualifying earnings [4] and this plan assumes 5% of salary, £3,400. Total value of the job: about £53,400 in cash and pension, plus 28 paid days off and sick pay.

A sole trader pays the same income tax bands but Class 4 National Insurance instead of Class 1 — 6% on profits between £12,570 and £50,270 and 2% above, with Class 2 treated as paid once profits reach the £7,105 small profits threshold, so nothing to pay [5]. That saves about £750 a year. The profit that nets £53,400 is about £72,600. Add roughly £4,900 of annual running costs the employer used to absorb — insurance, accountant, software, cloud sandboxes, travel, training — and you need £77,500 of revenue.

At 165 billable days that is £470 a day; at 130 it is £596. Your committed rate is £480. The steady state works with a little headroom, and year one is a pay cut of £6,000–£10,000 that the buffer absorbs. A fair trade — but not the same thing as "freelancing pays more".

The survival number, which is much lower

Your household needs about £3,900 a month net. Your partner's £31,000 salary produces roughly £2,153 of that after tax and NI [2][3], leaving £1,747 a month for you — about £21,000 net, which is about £23,900 of profit and, with running costs, £28,800 of revenue. That is 60 billable days at £480. Northbank at eight days a month for six months is £23,040; Kestrel's build is £9,500. £32,540 committed against a £28,800 floor. That comparison is why this report says Promising and not Risky.

That leaves £6,061 unallocated, which looks comfortable until you put the tax timetable beside it.

MVP & What to Build First

There is no pivot to recommend here, which is rarer than you would think. Sequencing is what matters, and the temptation with three months of notice is to spend it building a brand. Build the documents that decide whether you get paid instead.

Build first, inside the notice period

• A consultancy agreement you own — not the client's paperwork. Scope, deliverable, acceptance criteria, change control, payment terms, IP assignment on payment rather than on delivery, and an interest clause. The one thing worth paying a solicitor for.
• A written scope for the Kestrel build. A fixed price with a vague scope is not a business model, it is a donation. Define "done" before you quote.
• The tax reserve account, open before the first invoice.
• One case study from your employed work, with the employer's permission and names removed. Without it the rescue offer has nothing behind it.

Defer deliberately

• The limited company. Sole trader is right for year one — see section 11 — and incorporating later is easy. Now it adds cost and filings for an unquantified benefit.
• VAT registration. You are under the threshold; watch it rather than volunteering, for reasons in section 12.
• A real website. One page saying what you fix, for whom, with a case study and a phone number will outperform anything larger for two years.

The first ninety days of trading have one job, and it is not revenue — the revenue is committed. It is one sale from neither Northbank nor Kestrel. Until that exists you do not have a business; you have two clients who have not yet left.

Go-to-Market & Launch Plan

You have the easiest launch of any business type in this collection. Nobody needs to discover you: you need about eight clients over three years and you probably know forty people who could introduce you to them.

• Leave well. Former employers and colleagues are the largest source of freelance work in this discipline. Check your contract for non-compete and non-solicit clauses before you speak to anyone, and assume the client-poaching clause is enforceable in spirit even where it is weak in law.
• Convert the verbal commitments to paper during the notice period. A signed statement of work with a start date is worth ten times a conversation — and if either commitment evaporates you want to find out while still on salary.
• Go to the five implementation partners from section 3, not to end clients. They have the overrun projects and no capacity, and they introduce work at no sales cost.
• Two conversations a week, forever. Not a campaign. The freelancers who get into trouble stop selling when they are busy, then have nothing when the engagement ends.
• Do not launch with a discount. A first client at £380 a day is a permanent reference point, and everyone they introduce you to will hear that number.

Marketing & Growth

For a solo consultant selling £10,000–£20,000 engagements to a few dozen viable buyers a year, almost all marketing advice is written for the wrong business. Three things move it.

Referrals from finished work. Your only reliable growth engine. Engineer it rather than hoping: at the end of every engagement ask for one specific introduction to a named type of person, and ask while the client is pleased rather than three months later.

Writing about the failure modes you fix. Not thought leadership — six short pieces a year on why a particular kind of Azure pipeline breaks, with the diagnostic, aimed at the lead developer living with that problem. It is the only content that produces inbound enquiries here, and it takes a year to work, which is why it starts now.

A LinkedIn profile that reads as a service, not a CV. "I rescue half-built Azure data platforms for companies without a platform team" is a position. "Senior Data Engineer | Azure | Databricks | Python" is a job application. Growth past £90,000 comes from raising the fixed price, adding the support agreements in section 5, and eventually subcontracting delivery — the first is a lifestyle, the last is a company.

Pricing Strategy

Your £480 anchor is defensible and slightly too low. The UK median is £500 with a ninetieth percentile of £663 [1], and you enter with a committed client and specialist positioning rather than as an unknown.

• Publish £550 and let Northbank keep £480 as an agreed volume rate for a committed eight days a month. A real discount for a real commitment, and it stops the anchor client setting your price for everybody else.
• Price fixed-scope work against the client's alternative, not your day rate. The Kestrel build at £9,500 is roughly twenty days at £480. If the honest scope is fourteen days, £9,500 is correct and the extra margin is yours for carrying delivery risk. If it is twenty-eight days, reopen it before you start, not after.
• Never quote a fixed price without change control. The clause defining what happens when requirements move is worth more than the number.
• Invoice in stages — a third on signature, a third at a midpoint, a third on acceptance. Never do six weeks of work and then ask for £9,500.
• Payment terms of 14 days, stated on the quote. Thirty days is a convention, not a law, and the interest position in section 12 only helps if your terms are written down.
• Raise the rate for new clients, not existing ones. A 5–8% annual step compounds invisibly; asking a three-year client for 20% does not.

Legal Setup & IP

Registration is trivial. You must register for Self Assessment as a sole trader once you earn more than £1,000 in a tax year — the trading allowance — and you can register earlier by choice, which you should [1]. There is no licence to obtain.

Off-payroll working, which is the item that actually matters

Get this right and most of the horror stories about UK freelancing do not apply to you. The off-payroll working rules — IR35 — bite where a worker supplies services through their own intermediary, typically a personal service company. Who decides status depends on the client. For a medium or large private-sector client or a public authority, "the client will be responsible for determining the employment status of the worker" and should issue a status determination statement [2]. Where the client is small and in the private sector, "the worker's intermediary is responsible for deciding the worker's employment status" [2]. Since 6 April 2025 a non-public-sector organisation is medium or large if it exceeds two of: turnover above £15 million, balance sheet total above £7.5 million, more than 50 employees — the 50-employee limit unchanged [3].

Both committed clients — 40 staff and 25 staff — are comfortably small. Two consequences. First, if you later incorporate, the determination and the liability sit with you: the intermediary must calculate the deemed employment payment, pay employer National Insurance, and account for income tax and employee NI [4]. Second, as a sole trader invoicing directly you have no intermediary, so the off-payroll chapters do not apply to you at all — but employment status for tax still does, and it is then your client who carries the PAYE risk if HMRC decides the relationship was really employment. That is why some clients refuse to engage sole traders.

Either way, run HMRC's Check Employment Status for Tax tool for every engagement and keep the output. "HMRC will stand by all results given by the tool, as long as the information you give remains accurate and is in accordance with our guidance", and the saved result can be used as a valid status determination statement [5].

Sole trader or limited company

Sole trader for year one. The company case has narrowed. Corporation Tax is 25%, with a 19% small profits rate below £50,000 and Marginal Relief between £50,000 and £250,000 [6]. Dividends above a £500 allowance are taxed in 2026-27 at 10.75% basic, 35.75% higher and 39.35% additional rate [7] — the first two of which rose this April. Against that, £100 to incorporate online and £50 a year for the confirmation statement [8], plus statutory accounts, a Corporation Tax return, a payroll scheme and typically £700–£1,500 a year of accountancy you do not currently need.

At a profit near £74,000 the remaining company advantage is real but modest and easily eaten by that admin. The genuine reasons to incorporate are limited liability on fixed-price delivery risk, and clients who will only contract with a company. Revisit at the end of year one with actual figures.

Also settle before the first engagement

• Intellectual property. Assign IP in deliverables to the client on payment and expressly reserve your pre-existing tools, templates and reusable code. Without that reservation you can sign away the library that makes you fast.
• Professional indemnity insurance before you start, not after the first claim. Many clients require it contractually and will ask for the certificate; cover is inexpensive for a consultant [9].
• Your employment contract. Read the restrictive covenants and the notice clause before you tell anyone anything.
• Business name. Check the trade mark register before you pay for a logo; a sole trader's business name gives you no exclusive rights by itself.

Tax, Invoicing & Admin

The admin burden of a UK sole trader is genuinely light. The cash-flow burden is not, and the difference between those two sentences is what catches people.

The dates

• 5 October after the end of your first tax year: tell HMRC you need to file a return. Register late and HMRC gives you a different filing deadline, typically three months from the letter, but the tax is still due on 31 January [1].
• 31 October for a paper return; 31 January for an online one, with a late filing penalty either side of it [1].
• 31 January to pay, and 31 July for the second payment on account [1]. This is the mechanism behind the £28,430 in section 6.
• Making Tax Digital for Income Tax applies by qualifying income in an earlier year: over £50,000 for 2024-25 meant starting from 6 April 2026; over £30,000 for 2025-26 means 6 April 2027; over £20,000 for 2026-27 means 6 April 2028 [2]. You are not in the first wave, but on these profits you will be brought in — so choose MTD-compatible software now rather than migrating later.

VAT, and why the Flat Rate Scheme is probably a trap for you

You must register once taxable turnover in the last twelve months exceeds £90,000, or when you expect to exceed it in the next 30 days — within 30 days of the end of the month you crossed it, effective from the first day of the second month after [3]. On £480 across 165 days plus fixed-price work you are under it in year one and plausibly over in year two. Check the rolling twelve-month figure monthly; late registration means VAT owed on past sales plus penalties [3].

The Flat Rate Scheme is available below £150,000 of VAT turnover and lets you pay a fixed percentage and keep the difference, but you cannot reclaim VAT on purchases except certain capital assets over £2,000 [4]. The published rate for computer and IT consultancy or data processing is 14.5% [5] — except that a business whose goods cost less than 2% of turnover, or less than £1,000 a year, is a limited cost business and pays 16.5% [5]. A data engineer working from a laptop buys almost no goods, so you would almost certainly be one; 16.5% of gross is very close to the 20% you charged, and the advantage disappears. There is a 1% discount in your first year as a VAT-registered business [5], which is the only part worth having. Standard VAT accounting, reclaiming input tax on cloud and software bills, will normally beat it.

Getting paid

Under the Late Payment of Commercial Debts (Interest) Act 1998 you can charge statutory interest at 8% above the Bank of England base rate [6]. With the Bank Rate at 3.75% [7] that is currently 11.75%. The reference rate is fixed six-monthly — the rate at 30 June applies to debts becoming overdue between 1 July and 31 December, and the rate at 31 December to those overdue between 1 January and 30 June — and once fixed it does not move with the base rate [6]. You can also claim a fixed sum for debt recovery costs: £40 below £1,000, £70 from £1,000 to £9,999.99, and £100 at £10,000 or more, once per payment [8]. If no date is agreed, payment is late after 30 days for public authorities and 60 days for business transactions [9].

Be clear-eyed about what that is for. On the £9,500 Kestrel invoice 45 days late, statutory interest is about £138 plus the £70 fixed sum. It is not compensation; it is a credible, unemotional thing to put in a chaser, and its real value is that clients who know you charge it pay you first. Put the clause in the contract, state it on the invoice, and stage payments so no single overdue invoice can break your month.

Tech, Hosting & Security

You are technical, so the risk here is not incompetence. It is holding client production credentials on a personal laptop with a consumer backup, and having nothing to say when a security questionnaire arrives.

• Separate the machine, or at least the accounts. Full-disk encryption, a password manager, hardware-backed multi-factor authentication on every client tenant. A contractor with a compromised laptop is a supply-chain incident, and it ends the engagement.
• Never hold long-lived client credentials. Ask for a named account inside the client's own tenant, time-boxed, removed at the end. It protects them and removes you from the blast radius.
• Your own backups, versioned and offsite — contracts, invoices, insurance certificates, CEST results, bookkeeping. Losing the paperwork is worse than losing the code, because the code is in the client's repository.
• Answer the security questionnaire once, properly. Three pages on encryption, access control, backup, subprocessors, incident response and insurance. A day to write, reusable for a decade.
• A hard budget alert on cloud sandbox spend. A forgotten Databricks cluster is a four-figure lesson.
• Keep client data out of your own systems where you can. If you never hold a copy of a client's customer table, most of section 14 becomes straightforward.

Regional Compliance & Privacy

You are not selling to consumers and you are not regulated, so your compliance surface is narrow and almost entirely about other people's data — which a data engineer touches by definition.

The ICO fee. Organisations including sole traders that use personal information must pay the data protection fee unless exempt. Tier 1 covers micro organisations with turnover up to £632,000 or no more than 10 staff and costs £52; tier 2 is £78 and tier 3 £3,763, with a £5 discount for direct debit [1]. You are tier 1. Pay it and expect clients to look you up on the register — several procurement checklists include it.

Controller or processor. This distinction determines your obligations and is routinely mishandled in consultancy contracts. When you build a client's warehouse and touch their customer data on their instructions, you are a processor and they are the controller — a written processing agreement is required, specifying subject matter, duration, purpose, security measures, subprocessors, deletion and audit. When you hold your own prospect list, you are the controller and need a lawful basis, a privacy notice and a retention rule. Both are true at once. Most freelance contracts address neither.

Build these into how you work. Do not copy production data onto your laptop to "test locally" — the most common real-world breach in this discipline; use synthetic or masked data. Have the written agreement in place before you are given access, and keep a short record of what you processed for whom, because that is the first question after any incident.

Two cautions. UK data protection law has been amended in recent years and continues to be; check current ICO guidance for your exact processing rather than any summary, this one included. And if an engagement takes you into special category data — health, biometric, criminal offence data — the obligations step up materially and are outside this report's scope. Take advice before accepting it.

Risks & Common Pitfalls

• The 31 January 2029 bill. A balancing payment plus a payment on account, roughly £28,430 on these figures [1]. It ends otherwise healthy freelance businesses. Reserve 30% of every receipt from invoice one.
• Client concentration. Northbank is over 70% of committed revenue. Consultancies cut subcontractors first when their own pipeline softens, with a fortnight's notice. Get a third client before month six, at a lower rate if necessary.
• Verbal commitments. Neither client is contracted. People change their minds during a three-month notice period and budgets get reallocated. If both are not in writing before your last day, delay the resignation.
• Fixed-price scope creep. Your differentiator is also your largest delivery risk. The £9,500 build becomes a £4,000 build the moment scope expands by half and you absorb it. Change control in writing, from the first engagement.
• Selling once and never again. Bill hard for four months, do no selling, then face six empty weeks. Two conversations a week regardless of workload.
• Status drift. A day-rate engagement running three years with your own desk, a line manager and a rota starts to look like employment. Keep your CEST results, review them when the engagement changes, and keep at least two clients [2].
• No sick pay, no employer pension. There is no statutory sick pay for the self-employed, and the 3% employer contribution [3] stops. Both are real pay cuts. Start a personal pension in month one and price income protection — a broken wrist is a continuity event when you are the only engineer.
• Isolation and no ceiling on the hours. Unglamorous, and the most common reason people go back. Nobody will tell you to stop working; set the working week before you have to defend it.

Action Plan: 30 / 60 / 90 Days

This plan assumes you serve notice now and trade from January 2027. Days 1–60 fall inside the notice period — the most valuable stretch you will have, because you are being paid to de-risk.

Days 1–30 — convert the commitments first

• Read your contract's notice, non-compete and non-solicit clauses. Then serve notice.
• Get both verbal commitments into a signed statement of work with a start date, a rate or fixed price, a defined deliverable and 14-day payment terms.
• Instruct a solicitor on one reusable consultancy agreement: change control, IP assignment on payment, reservation of your own tools, late-payment interest clause [1].
• Run HMRC's CEST tool for both engagements and save the results [2].
• Scope the Kestrel build honestly in writing. More than twenty days of work means reprice it now.

Days 31–60 — set the machinery up

• Register for Self Assessment as a sole trader rather than waiting for the £1,000 threshold [3].
• Business bank account plus a separate tax reserve account. Standing rule: 30% of every receipt moves across the day it arrives.
• Professional indemnity and public liability cover in place, certificates saved [4].
• Pay the ICO tier 1 data protection fee, £52 [5].
• Choose MTD-compatible bookkeeping software now, not when mandated [6].
• Write the security statement and the case study. One-page site up, LinkedIn profile rewritten as a service.
• Start the personal pension and get an income protection quote.

Days 61–90 — trade, and sell while busy

• Deliver. Invoice Kestrel in three stages, not one.
• Two sales conversations a week from week one, targeting the five implementation partners from section 3.
• Publish the first two pieces on the failure modes you fix.
• Monthly reminder to check rolling twelve-month turnover against the £90,000 VAT threshold [7].
• At day 90, check three numbers: billable days achieved against the 130-day plan, the tax reserve against 30% of receipts, and whether one enquiry has come from neither committed client. The third decides whether year two is a business or a slow return to employment.