The till is the most-used piece of equipment in a restaurant. Your team touches it more often than the coffee machine, the card reader or the front door. It is also the piece of software operators consider least, usually because it came bundled with something else, or because the last one broke and you replaced it in a hurry on a Tuesday.
That is a shame, because an EPOS in 2026 is not a glorified cash drawer. It is the system of record for what actually happened in your business: every item sold, every payment taken, every void and every discount, in the order they occurred. Choose it badly and you spend years working around its limitations, paying for hardware you do not need and exporting numbers by hand. Choose it well and it quietly does half your admin and feeds the rest of your systems without anyone typing a thing.
I have changed EPOS more than once, including mid-service when one died on a Saturday, so this guide is written from the wrong side of those decisions. Here is how I would choose a till today.
What an EPOS really is now
A modern EPOS does three jobs at once, and it is worth separating them because vendors blur them together.
It takes the order and the payment. That is the obvious job and almost every product does it adequately.
It records what happened, accurately and in a way you can trust later. This is the job that matters to your accountant, to HMRC, and to you at the end of a month when you are trying to understand why margins moved. An EPOS that keeps a clean, tamper-evident record of every transaction is doing something genuinely valuable. One that lets numbers be quietly edited after the fact is a liability dressed as a convenience.
And it captures data about the guest, or it throws that data away. Every order is attached to a person who is sitting at a table you probably took a booking for. A till that knows nothing about that person is discarding the most useful thing it touches all night.
When you judge an EPOS, judge all three jobs, not just the first.
iPad versus traditional terminals: the honest trade-offs
The biggest fork in the road is hardware. Traditional EPOS runs on dedicated terminals, purpose-built boxes that sit on the counter. Newer systems run on iPads, sometimes on Android tablets.
The honest case for traditional terminals is durability and a certain reassuring solidity. They are built for the environment, they take a knock, and the hardware tends to last. The case against is cost, inflexibility and the fact that you are often locked to one supplier's kit and one supplier's repair timeline.
The honest case for iPad EPOS is cost, familiarity and flexibility. The hardware is cheap to replace, your team already knows how to use a tablet, you can add a station for a busy weekend without a service call, and the software tends to update more often. The case against is that a consumer tablet is not built for a hot, wet, busy kitchen, so you need decent cases and a sensible charging routine, and you are relying on the software handling a dropped connection gracefully.
There is no universally right answer, but for most independents opening or re-fitting in 2026, iPad-based EPOS wins on cost and flexibility, provided the software is genuinely robust. Which brings us to the features that actually matter.
The features that earn their place
Ignore the feature count. Here is what I would not run a service without.
Offline mode that genuinely works. Your internet will drop. It always does, usually at the worst moment. An EPOS that stops taking orders when the WiFi blips is not fit for a restaurant. The till must keep working through an outage and reconcile cleanly when the connection returns. Ask the vendor exactly what happens when the line goes down, and do not accept a vague answer.
A clean, trustworthy audit trail. Every transaction, void and discount should be recorded in a way that cannot be quietly rewritten later. This protects you in a dispute, satisfies your accountant, and means your end-of-day actually reconciles. It sounds dull. It is one of the most important things a till does.
A kitchen display that replaces paper. Orders should land on a screen in the kitchen as they are rung in, legibly, in sequence, without a printer jamming on the busiest night of the year. A native kitchen display is cheaper to run than a printer and far less temperamental.
Reporting you will actually read. Sales by item, by period, by server, with the ability to see what is selling and what is dying on the menu. If you have to export to a spreadsheet to answer a basic question, the reporting is not good enough.
Accounting that exports cleanly. Your EPOS should hand your bookkeeper clean figures in a format their software accepts, ideally with proper accrual accounting rather than a rough cash summary. The hours this saves at month end are real.
Payments that suit you, not the vendor. Some EPOS products tie you to their card processing at their rate and make it awkward to use anyone else. The better approach is choice: use the vendor's payments if the rate is good, or bring your own provider without penalty. Be wary of any system that punishes you for not using its card processing.
The feature that changes the most: does the till feed the guest record?
Here is the argument I care about most, because it is the one that took me longest to understand.
A standalone till tells you what was sold. A connected till tells you who bought it. That difference is enormous and almost nobody sells on it, because most EPOS companies are tills and nothing else, so they cannot.
Think about what happens on a normal evening. You took a booking, so you know who is at table twelve. They order, they pay, they leave. If your till is an island, all you have at the end of the night is an anonymous total: table twelve spent eighty-four pounds. If your till feeds a guest record, you know that Sarah, who has been in seven times this quarter and usually takes the corner booth, spent eighty-four pounds and ordered the same red she always does. One of those is data. The other is a relationship.
This is why I would now weigh integration above almost every individual feature. A till that talks to your bookings and your guest record, so that every item and payment lands on the right person without anyone typing it in, is doing something a better-specified standalone till simply cannot. A till that talks to your booking system is worth more than a slightly faster one that talks to nothing.
Reading a quote honestly
EPOS pricing hides in more places than booking-system pricing, so take the quote apart.
There is the software licence, usually per till per month. There is hardware, either bought outright or rented, and rental can quietly cost far more over a contract than buying. There are card processing fees, which is where a cheap-looking licence often makes its money back, so always check the card rate and whether you are free to use another provider. There is the contract length and the cost of leaving, including whether you own or merely rent the hardware at the end. And there are the extras: kitchen printers, cabling, support tiers, integration fees.
A till advertised at a low monthly price with a high card rate and a three-year hardware rental can be the most expensive option on your shortlist. Work out the real annual cost across all of those lines before you compare anything.
The bit everyone underestimates: switching and training
Two practical realities decide whether a good choice on paper becomes a good choice in your building.
The first is migration. Moving menus, modifiers, prices and any historical data across is more fiddly than vendors admit, and it is worth asking exactly what they will do for you versus what you will be doing yourself at midnight. The second is training. A till your team cannot learn in half an hour will cost you mistakes and slow service for weeks. The best systems are close enough to what staff already understand that a new starter is productive on their first shift. Ask to put a real member of your team in front of the product before you buy, not just the owner.
How to choose: the questions to ask any vendor
Take these into the demo. What exactly happens when my internet drops mid-service? Can transactions be edited after the fact, and if so, who can see that they were? Does this feed my bookings and my guest record, or is it a standalone till? What is the real annual cost, including hardware, card fees and the cost of leaving? Am I free to use my own card provider? Can a new member of staff learn this in a shift? And: can I export everything, including my sales history, the day I decide to leave?
The answers to the offline question and the integration question will tell you most of what you need to know.
Where Grace fits
Grace's EPOS is not a separate box bolted to the rest of the platform. It is a native iPad till that exists to feed one guest record, so every item and payment lands on the person who ordered it without anyone re-keying anything. It keeps a tamper-evident, hash-chained record of every transaction, exports cleanly to your accountant with proper accrual accounting, runs a native kitchen display so you are not feeding a printer on a Saturday, and keeps working through a WiFi outage. There are no per-cover fees, and you can use Grace Pay at our published rate or bring your own provider with no Grace fee on top.
You can see the EPOS module in detail on the EPOS page, and the wider thinking on our restaurant EPOS overview.
It works the way it does because it is tested every week on real service at The Nuthatch, the restaurant I run. The offline mode exists because the line went down on a Saturday and I needed the till to carry on. If a feature cannot survive a Friday night, it does not ship.
The short version
Decide your hardware honestly, and for most independents in 2026 a robust iPad system wins on cost and flexibility. Demand a genuine offline mode and a clean audit trail, because those protect your service and your books. Weigh integration above raw speed, because a till that feeds your guest record is worth more than one that forgets. Read the quote across every line, including card fees and the cost of leaving. And put your actual team in front of it before you sign.
---
FAQ
What is the best EPOS for a small restaurant in the UK?
The best one is the one that fits your service, exports cleanly to your accountant, keeps working when the internet drops, and feeds the rest of your systems rather than sitting as an island. For most independents in 2026 that points toward a robust iPad-based system, but judge it on offline reliability and integration before you judge it on price.
Is an iPad EPOS reliable enough for a busy venue?
Yes, provided the software handles outages properly and you treat the hardware sensibly with proper cases and a charging routine. The risk is never the iPad itself, it is software that stops taking orders the moment the connection drops. Ask the vendor precisely what happens during an outage and do not accept a vague answer.
How much does a restaurant EPOS cost?
It depends on the software licence per till, whether you buy or rent hardware, and the card processing rate, which is where a cheap-looking licence often recovers its margin. Always calculate the real annual cost across all of those lines, including the cost of leaving, rather than comparing headline monthly prices.
Should my EPOS connect to my booking system?
If you care about knowing your guests, yes. A standalone till tells you what was sold. A till that connects to your bookings and guest record tells you who bought it, so every visit and spend lands on the right person automatically. That connection is often worth more than any single standalone feature.
Can I use my own card provider with an EPOS, or am I locked in?
It varies, and it matters. Some systems tie you to their payments at their rate and make alternatives awkward. The better products let you use their processing if the rate suits you or bring your own provider without penalty. Check this before you sign, because it is hard to change later.
What happens to my data if I leave an EPOS provider?
A reasonable provider will let you export your menus, sales history and guest data in a standard format at any time. Ask the question before you buy, not when you want to leave. If the answer is unclear, that lack of clarity is itself a reason to be cautious.
---