I've set up all three of these platforms for UK small businesses, and I've also been the person called in on a Friday evening because a workflow quietly died in March and nobody noticed until the VAT return was wrong. So trust me on this: the choice of automation platform matters far less on day one than it does in month twelve, when the renewal invoice and the maintenance questions arrive together.
If you're researching this, you're probably drowning in admin, and you're in good company. The latest American Express SME Business Barometer, covered by Startups Magazine, surveyed 1,000 UK small business owners and found they spend an average of 11 hours a week on admin and finance tasks, roughly six working days a month, and 54 per cent say paperwork actively gets in the way of running the business. Workflow automation is the most realistic way I know to claw a chunk of that back.
The Short Answer
Here it is before we go deep. Make is the best value for most UK small businesses with real, branching processes and someone willing to give up a weekend to learn it. Zapier is the right call when your workflows are simple, your volumes are low and your time is worth more than the subscription. n8n wins when you have technical confidence in house, high volumes, or customer data you'd rather keep on a server you control than on a cloud in America.
That's the honest one paragraph version. The rest of this article is the evidence, in pounds, with the traps I've watched people fall into.
What You're Actually Choosing Between
All three tools do the same basic job. Something happens in one app, a new enquiry lands, an invoice gets paid, a form gets filled in, and the platform automatically performs actions in your other apps without anyone touching a keyboard. The differences are in how they're built, how they bill and who they expect to be driving.
Zapier is the veteran, and it's enormous. Parseur's sourced statistics work puts it at around 3.4 million businesses, and its catalogue covers more than 9,000 apps. You build "Zaps" as a simple chain: trigger, then actions. It's the easiest of the three by a distance, and everything runs on Zapier's cloud with nothing for you to look after.
Make, formerly known as Integromat, swaps the chain for a visual canvas. Each step is a module, each workflow is a scenario, and you can see branches fork and rejoin in front of you. It handles conditional logic, loops and messy data far more gracefully than Zapier, and it costs considerably less for the same work. It's also cloud only, hosted in the EU.
n8n is the interesting one. It's a Berlin built platform you can either pay to use as a hosted service or run yourself on your own server, free of licence fees. It's grown from open source roots into serious infrastructure: SAP took a strategic stake in May 2026 at a valuation of 5.2 billion dollars, double its figure from less than a year earlier, and the company reports 1.7 million monthly active builders and more than 1,400 enterprise customers. It assumes some comfort with data and APIs. An API, if the term is new, is simply the doorway one piece of software uses to talk to another.
A Quick Word About Money Before The Numbers
The prices below were checked in early September 2026 against each vendor's published pricing and recent independent breakdowns, and they're annual billing rates unless I say otherwise. Monthly billing typically costs a fifth to a third more across all three.
One localisation note, and I'll only say it once. Zapier bills UK accounts in sterling. Make publishes its prices in dollars and n8n publishes its prices in euros, so my pound figures for those two are approximate conversions at the early September 2026 rates of roughly 1.35 dollars and 1.16 euros to the pound, and your business card may add a small non-sterling transaction fee on top. All list prices exclude VAT, so add 20 per cent unless you're VAT registered and can reclaim it.
What Zapier Costs, And How It Counts
Zapier bills per task, and a task is one action that actually runs. The trigger itself is free, and so are a set of built in helper steps. This is worth flagging because plenty of comparison articles still claim every filter and formatter burns a task, and that's out of date. No Code MBA's August 2026 breakdown lists the steps that don't count: filters, paths, formatter, delay, looping, digest and Zapier's own tables and forms. Every real external action, creating a contact, sending an email, adding a row, does count.
The Free plan gives you 100 tasks a month but limits every Zap to two steps, one trigger and one action. Treat it as a demo of the editor, nothing more, because almost no genuinely useful workflow fits in two steps. Professional starts at $19.99 a month on annual billing, or $29.99 monthly, for 750 tasks, and adds multi-step Zaps, webhooks, conditional paths and premium apps. According to the pricing guide by Electric Monk, a UK Zapier specialist, that entry Professional tier comes to £182.76 a year for UK accounts, which is £15.23 a month.
Team starts at $69 a month, around £51, for 2,000 tasks, 25 users, shared workspaces and single sign on, which lets staff log in with one company account. Zapier's AI agents are priced separately at $33.33 a month, about £25, for 1,500 agent activities. Two warnings from experience, both backed by TinyCommand's analysis of the official pricing page: teams routinely underestimate their task burn by three to five times, and once you exceed your allowance, extra tasks bill at 1.25 times your plan's effective rate until you hit three times your included volume, at which point your Zaps pause.
What Make Costs, And The Credit System
Make bills per credit, and since 27 August 2025, when it renamed operations to credits, each standard module action, trigger or function consumes one, as Alltomate's plan guide explains. AI features and code steps consume more; Use Apify's breakdown notes that Make's code modules burn two credits for every second of run time.
The Free plan gives 1,000 credits a month with two active scenarios and a minimum run interval of 15 minutes, which rules out anything time sensitive. Core is $12 a month on annual billing, roughly £9, for 10,000 credits and unlimited active scenarios. Pro is $21, about £15.50, and Teams is $38, about £28, and here's the detail that surprises people: all three paid tiers start at the same 10,000 credits. Moving up the ladder buys capability, priority execution, scheduling as often as every minute, log search, team roles, not volume. Extra credits are bought separately, and Trackstack's pricing analysis found the top up packs have carried roughly a 25 per cent premium over in-plan credits since late 2025.
For a solo operator or a two person firm, Core is nearly always the right rung. On Make's published pricing, you're getting more than 13 times Zapier's entry task allowance for under half the entry price, which is the single biggest reason Make keeps winning my client recommendations.
What n8n Costs, Cloud And Self-Hosted
n8n bills per execution, and one execution is one complete run of a workflow, however many steps it contains. A two step workflow and a forty step workflow cost exactly the same per run. Once that clicks, n8n's pricing page reads very differently from the other two.
The hosted Starter plan is 20 euros a month on annual billing, about £17, or 24 euros monthly, for 2,500 executions, five concurrent runs and 2,300 AI credits. Pro is 50 euros, about £43, for 10,000 executions. Every hosted plan includes unlimited users, which quietly settles a lot of arguments, given Zapier caps Team at 25 seats. The catch sits higher up: Business, which is where single sign on, Git version control and multiple environments live, jumps to 667 euros a month, around £575, though firms with under 20 employees can apply for a startup discount of 50 per cent on it.
Then there's the Community Edition, which is free software you run yourself. A virtual private server, a small rented computer in a data centre, costs five dollars or so a month, well under a fiver in real terms, and No Code MBA reckons a typical Docker based setup takes 30 to 60 minutes if you follow a guide. Unlimited executions, no licence fee, and the missing features are almost entirely enterprise concerns like single sign on and advanced permissions. I'll come back to why "free" deserves inverted commas.
A Worked Example With Real Numbers
Abstract billing units are useless, so let's price one workflow every UK service business recognises. A new enquiry arrives through your website form, and the automation creates a contact in your CRM, that's your customer database software, posts an alert to your team chat, adds a row to a spreadsheet and sends the customer a confirmation email. Four billable actions per enquiry, plus the trigger.
At 300 enquiries a month, Zapier burns 1,200 tasks, which already blows past the 750 task base Professional tier and pushes you up its volume ladder. Make uses five credits a run, 1,500 credits a month, which is 15 per cent of the Core plan's allowance at £9 or so. n8n uses 300 executions, about 12 per cent of the Starter allowance at roughly £17, or effectively nothing self-hosted.
Now the twist that decides real budgets. Add three more steps, a duplicate check against your CRM, a postcode lookup, a drafted quote. Zapier's bill rises with every step. Make's rises gently. n8n's doesn't move at all. As the AI Essentials small business comparison rightly insists, tasks, credits and executions are not the same unit, so never compare the three headline prices straight across. Price one real workflow of yours at your real volume, and the fog clears in ten minutes.
The Learning Curve Nobody Puts On The Pricing Page
Zapier is genuinely usable on day one. The trigger and action model is the simplest mental picture in automation, the template library is vast, and a non-technical owner can ship something before lunch. That convenience is exactly what you're paying the premium for.
Make is not beginner software, whatever the marketing implies, and I think the "anyone can do it" framing does new users a disservice. The canvas is brilliant once it makes sense, but your first scenario will take an evening rather than an hour, and mapping data between modules is where I watch every newcomer get stuck. Budget a proper weekend and a few YouTube rabbit holes, and after that it pays you back for years.
n8n asks the most. It expects comfort with JSON, the structured text format web services use to exchange data, with APIs, and occasionally with a line of JavaScript. Digidop's comparison estimates full mastery can stretch over several weeks, which matches what I've seen. Hand it to a marketing coordinator cold and it will not go well. Hand it to anyone who's ever written a formula heavy spreadsheet and enjoys that sort of thing, and they'll be dangerous within a fortnight.
Apps And Integrations, And Why The Big Numbers Mislead
The headline counts, checked by Parseur in August 2026 against each vendor's directory, are 9,000 plus apps for Zapier, 3,000 plus apps and a further 350 plus AI apps for Make, and 2,003 listed integrations for n8n. People treat this as a league table. It mostly isn't.
Your business uses perhaps 15 apps that matter, and all three platforms cover the mainstream ones. What actually decides it is the long tail: the niche UK booking system, the trade specific quoting tool, the accounts package your bookkeeper insists on. Check each of those by name in each directory before you commit, because that twenty minutes of homework outranks every headline number.
There's also a webhook shaped escape hatch, and a webhook is simply an instant message one app sends another the moment something happens. As Fixed Labs points out in its comparison, n8n's smaller native catalogue matters less than it looks because generic HTTP requests and webhooks let it reach almost any service with an API. That flexibility favours technical users, which is n8n's whole personality anyway.
The AI Layer, And Whether It's Worth Paying For Yet
All three vendors are racing to bolt intelligence onto plumbing. Zapier offers a Copilot that builds Zaps from a plain English description, and sells autonomous agents as that separate £25 a month line item I mentioned. Make has folded AI apps and agents into its catalogue, with AI features drawing on the same credit pool at different rates. n8n has gone furthest, with native nodes for the major model providers and agent orchestration, which is a large part of why SAP is embedding it inside its own Joule Studio product.
My honest advice for a small firm: automate the boring, deterministic work first, because that's where the reliable payback lives. When you do add AI steps, watch two things. They consume your allowances faster than standard steps, and their output is unpredictable enough that you should route it past a human, or at least a validation step, before it touches an invoice or a customer email.
Keeping The ICO Happy: The UK Data Protection Bit
This is the section most comparisons skip, and it's the one that can genuinely hurt a British business. The moment customer names, emails or order details flow through one of these platforms, that platform is processing personal data on your behalf under the UK General Data Protection Regulation, the data protection law we carried over after Brexit, which is enforced by the Information Commissioner's Office. Practically, that means you need a data processing agreement with the vendor, the contract that sets out what they may do with the data, and you should name them among your processors in your privacy notice. Ask each vendor for its data processing terms before you connect live customer data, and if they can't produce them quickly, treat that as an answer in itself.
Where the data physically lives matters too. Parseur's August 2026 check lists Zapier as cloud only with US or EU hosting, Make as cloud only with EU hosting, and n8n as running wherever you host it. EU hosting is the comfortable case, because the UK treats the European Economic Area as adequate, so data can flow there without extra paperwork. Confirm with the vendor exactly which plans include which region, because hosting options and plan tiers shift.
Sending personal data to US servers is a "restricted transfer" under UK GDPR, but it's a solved problem in most cases. Since 12 October 2023, as DAC Beachcroft and White & Case both set out, UK organisations can transfer personal data to US companies that have self certified under the UK Extension to the EU-US Data Privacy Framework, informally the UK-US data bridge, without needing standard contractual clauses or an International Data Transfer Agreement. The practical step is checking your vendor appears on the framework's public list; the ICO's guidance on the UK Extension walks through exactly that, and through the fallback safeguards and transfer risk assessment you'd need if a vendor isn't certified. Note also that the ICO refreshed its whole international transfers guidance on 15 January 2026 with a clearer three step test, as the law firm Kennedys has analysed, so if your last compliance review predates that, it's due another look.
And this is where n8n's self-hosting stops being a nerd hobby and becomes a compliance feature. Run it on a UK server and your customer data, credentials and logs never leave your own infrastructure, which for anyone handling sensitive records is the cleanest answer available. I'm not a lawyer and this isn't legal advice, but I've sat in enough due diligence conversations to know "it never leaves our UK box" ends them quickly.
Where Each Platform Will Bite You
Zapier bites on cost and on ceilings. The per task model means the bill scales with the complexity of your workflows rather than the value they create, which is the most common reason I'm hired to migrate people off it, and heavily branched logic gets hard to read in a linear editor. The Free plan, as covered, is a demo. Budget for Professional from day one or don't start.
Make bites through credit surprises. Triggers, iterators and badly designed scenarios consume far more than the canvas suggests, and Emergent's cost breakdown has my favourite cautionary example: a scenario that rescans a 400 row spreadsheet through four action modules burns 1,601 credits per run, while the same job built on a "watch new rows" trigger picking up 20 rows costs 81. Same outcome, twenty times the price, purely down to design. Learn trigger discipline early and Make stays absurdly cheap.
n8n bites through ownership. Self-hosting is free in the way an allotment is free: the plot costs pennies and the upkeep costs weekends, because updates, backups, monitoring and the 11pm failure are all yours. The hosted plans remove that but carry a brutal cliff, 50 euros to 667 euros, the moment you need single sign on or version control. And one more subtlety: n8n ships under a Sustainable Use Licence, meaning it's source available rather than strictly open source, which is irrelevant for using it in your own business but rules out reselling it as a service.
There's also the question of who you ring when something breaks at an awkward hour, and the three answer it very differently. Zapier answers it best, with staffed support on paid plans and the deepest documentation in the category. Make sits in the middle, with strong tutorials and a genuinely active community forum. On self-hosted n8n the answer is you, backed by a helpful GitHub and Discord community that quietly assumes you can already read an error log, which is either fine or terrifying depending on who you've hired.
One trap is shared by all three, and Parseur states it plainly: workflows do not port. There's no export from Zapier that Make or n8n will read, so outgrowing a platform means rebuilding, not migrating. Pick the tool you'll still be on in two years.
Which Platform Fits Which Business
Zapier suits the time poor owner whose workflows are short, linear and low volume, and who'd rather pay £15 odd a month than learn anything. If a new enquiry into the CRM and a Slack ping covers your ambitions, it's hard to beat. Best for: solo traders and small teams automating simple workflows across mainstream apps, where speed of setup beats cost per task.
Make suits the growing firm with genuinely messy processes, orders with exceptions, quotes with rules, data that needs reshaping, and at least one person who enjoys a puzzle. It's my default recommendation for most UK small businesses on price and capability combined. Best for: e-commerce, agencies and service firms running branching, multi-step small business automation on a tight budget.
n8n suits the business with a technical person on the payroll, or a founder who is that person, plus either high volumes or a hard requirement to keep data under its own roof. Best for: technical teams, regulated or privacy sensitive work, and anyone whose execution volumes make per task pricing silly.
Five Questions That Settle It
Who will actually build and fix this? Not who could, who will, at half nine on a Tuesday when a field mapping breaks. Choose the platform that person can genuinely operate, because a powerful tool nobody maintains is just a new source of silent failures.
How often will your busiest workflow run, and with how many steps? Multiply those two numbers. If the answer is small, Zapier's convenience is cheap. If it's large, Make or n8n will save you hundreds of pounds a year, and you now have the arithmetic to prove it.
Where is your customers' personal data allowed to live? If the honest answer is "we've never checked", start with the data protection section above before you connect anything. If the answer is "on our own servers", n8n has just been chosen for you.
Which tier will you need in year two? Map your growth against the cliffs: Zapier's task ladder, Make's credit top ups, n8n's Business jump. The entry price is bait on all three; the second year price is the real quote.
Can you afford the rebuild if you outgrow it? Since nothing ports, either budget a future rebuild as a known cost or spend an extra week choosing properly now. The second option is cheaper.
If you want my closing steer, it's this. Don't pick an automation platform from a comparison table, including mine. Take your single most annoying repetitive process, sign up for the free tiers of your two best candidates, and build that one workflow on both over a fortnight. The winner will be obvious, the losers cost you nothing, and you'll have already bought back a slice of those 11 weekly admin hours before you've spent a pound.