A small takeaway should choose the EPOS that sends orders from delivery apps and its own channel straight to the kitchen printer or screen, prints a clean ticket, reports sales by item and hour, and has fair contract terms. AI features matter last: score each on a real task from your own menu, not on the vendor's claims.
Vendor pages rank for this search and they mostly list features. Your decision is better made by tracing one order from a delivery app, through the till and kitchen, into the end-of-day report, and seeing where staff touch it twice. Every retype is a mistake waiting for a Friday night.
Trace one order through each system
Before comparing vendors, write your current route on a sheet of paper. A typical illustrative takeaway has three channels: walk-in and phone orders at the till, two delivery apps on tablets, and its own online ordering page. Mark each point where a person copies information from one screen to another. Those points are where an EPOS earns or loses its fee.
Then ask each vendor to demonstrate the same four journeys, with your menu loaded:
- A delivery-app order arrives and prints in the kitchen without anyone touching a tablet.
- A phone order with a modifier ("no onions, extra sauce") prints those modifiers clearly.
- An item is sold out. How fast does it disappear from every channel?
- End of day: the report shows sales by channel, item and hour, and matches the cash drawer and card totals.
If a vendor cannot do all four in a demo with your menu, assume a gap and cost it.
What the vendor pages actually say
Two widely used systems show the pattern. Epos Now's site lists its own "Epos Now Delivery" app, says its hospitality system can integrate with delivery apps without naming them on the home page, and sells a kitchen display system (KDS) as a hardware item. Its only AI reference there is the Enterprise tier, described as an AI-driven option for businesses running several POS systems or opening more locations. No takeaway-specific AI feature is named.
Square for Restaurants has Free, Plus and Premium plans. The Free plan charges only processing fees and includes menu management, pickup and local delivery, and a branded online ordering page; Plus adds inventory tracking and close-of-day reports. The kitchen display app is a paid add-on per device ($30 a month on Plus, $20 on Premium, as of October 2026), and the pricing page names DoorDash among its delivery partners. Square does have an AI feature: Square AI, labelled beta, which answers plain-language questions about your sales and can pin charts that update. Its page does not say which plans include it, and Square also mentions AI voice ordering without detail, so confirm both for your account before counting them.
The lesson is that integration, printing and plan add-ons are spelled out, while AI features are either tier-specific or in beta. Take that as a signal: the dependable value is in the plumbing. Check prices, which change and differ by region, on each vendor's own page for your market, and compare total cost, not the headline.
A scoring sheet for real takeaway tasks
Score each task 0 (fails), 1 (works with workarounds) or 2 (works cleanly). Weight the first three double.
| Task | What to look for | Score A | Score B |
|---|---|---|---|
| Delivery-app orders into the till | No retyping; status and acceptance synced | ||
| Kitchen ticket printing | Modifiers legible; one ticket per station | ||
| Item sold-out across channels | One switch, all channels | ||
| Own online ordering | Fees, menu sync, collection slots | ||
| Reports | By channel, item, hour; export to file | ||
| Staff permissions | Discount and refund controls | ||
| Hardware | Reuses your printer; offline mode | ||
| Contract | Term, exit fee, price rises | ||
| AI extras | Works on your menu; cost; can switch off |
The AI row is last on purpose. A system scoring 14 out of 18 on the first eight rows beats one scoring 10 with a clever assistant.
A worked comparison for a 60-order-a-night takeaway
An illustrative takeaway sells about 60 orders a night, six nights a week. 25 come from delivery apps, 20 by phone, 10 by walk-in and 5 from its own site. At present delivery-app orders are re-entered at the till by whichever staff member is free, taking about 90 seconds each. That is 25 orders times 90 seconds, or roughly 38 minutes a night, around 3.8 hours a week once you count 6 nights.
Option A is an EPOS that takes app orders directly and prints them. It removes most of that retyping, saves about 3.5 hours a week, and removes the transcription errors that lead to remakes. At $15 an hour that is $52 a week, around $225 a month, before counting remakes. Option B is cheaper per month but does not integrate with one of your two delivery apps, so half the retyping stays. It saves about $110 a month. The $115 difference exceeds most subscription gaps, which is why integration outranks price for a takeaway with heavy app volume.
Now add remakes. If the retyping errors cause one wrong order every other night, and a remake costs $9 in food and driver time, that is three remakes over six nights, about $27 a week or $115 a month, roughly the same again as the time saving. It also matches the lesson from pricing menus for delivery-app commission: the margin on app orders is already thin, so waste has a larger effect.
Delivery integrations: questions with yes or no answers
- Which delivery apps does it integrate with directly, and which need an aggregator such as Deliverect or Otter, which adds a fee?
- Does the order arrive with the customer's notes and modifiers, or only item names?
- Does accepting or rejecting in the EPOS update the app status?
- Can you pause an app when the kitchen is overwhelmed, from the EPOS?
- Do prices on each channel come from the EPOS, so you can set a higher app price without editing three menus?
- Who fixes it when an app changes its connection?
Write the answers down. "We can look into it" is a no. Integrations are a frequent source of silent failures, such as an order that reaches the till but never the kitchen printer, so ask how failures are flagged.
Printing and the kitchen
Kitchen printing sounds basic and is where systems fail on a busy night. Test with a long order, a modifier-heavy order and two orders arriving in the same second. Check that tickets show channel, order number, time due and customer notes in a size cooks can read at a glance. If you use a kitchen display instead of paper, ask what happens when the screen loses connection: does the system fall back to a printer? Epos Now sells a kitchen display as hardware, and Square charges for its display app per device; confirm for any vendor what is included in your plan and what is an add-on.
A realistic mistake: a takeaway installs a new system, loads the menu and tests with a single burger. On the first Saturday, a combo with three modifiers prints "Combo 2" on the ticket, with the sides on a second slip that gets stuck to the wrong order. The fix was to design ticket templates before opening, with combo contents grouped under the parent item, and to test every combo type. Do that in the trial week.
Reports: ask for the five you will actually read
Most systems offer dozens of reports. You will use five: sales by channel, sales by item, sales by hour, discounts and refunds by staff member, and end-of-day cash against card. Ask the vendor to produce each from sample data in the demo, and to export it as a file. If the reports cannot be exported, you cannot analyse them elsewhere, and any AI analysis tool you might add later has nothing to read.
A useful illustration of what to do with them: sales by hour shows that between 5pm and 6pm you take 25 orders and between 8pm and 9pm you take 35. If your pizza oven limits you to 30 an hour, the 8pm peak is where you lose orders and where delivery-app "busy" pausing matters.
Sold-out items and one-switch menu control
Before: the kitchen runs out of lamb at 7.40pm. A cook shouts, the person on the till stops selling it, someone edits the website, and nobody tells the two delivery apps for another 20 minutes. Four app customers order lamb, and four orders have to be cancelled or substituted, with refunds and poor ratings to follow. After: one tap in the EPOS marks lamb unavailable, and the integration removes it from each channel within a minute. The second version is worth real money on a busy night, and it is a plumbing feature, not an AI one. Ask each vendor how long the change takes to reach each app and whether you can schedule items off at a set time.
When the internet drops
A takeaway cannot stop at 7pm because the connection failed. Ask what the system does offline: can it still take orders and print tickets, take cash, and store card payments until the connection returns? Delivery-app orders will not arrive without a connection, so find out how the EPOS or the app warns you. A sensible test is to pull the router's plug during the demo, or the trial, and watch. A vendor who says "it won't happen" has not run a takeaway on a Saturday.
Staff permissions and till discipline
Voids, discounts and refunds are where losses hide. Check that you can limit those actions by role, require a manager code above an amount, and see a report of every void with the staff member and time. This is also where simple "AI" anomaly alerts, if offered, would help: a flag when one person's voids jump. Ask to see such an alert, because many are marketing phrases for a standard report. If you do the check yourself, run the void report weekly and look at the three highest staff totals. A takeaway with five staff and a void total that is 3 percent of sales for one person deserves a conversation, not an assumption.
A sample AI sales summary, and what to fix in it
Suppose an EPOS offers a plain-language summary, or you paste a week's export into a general assistant. A prompt to try:
Below is a CSV of last week's takeaway sales: date, hour, channel,
item, quantity, price, discount. Summarise in under 150 words:
total sales by channel; top five items by quantity; the busiest
hour each night; any discount over 15% of an order. Use only the
data given. If something is unclear, say so.A realistic output: "Delivery apps made 41 percent of sales. The top item was the chicken wrap. Friday was busiest at 8pm." Check each statement against the file. Common faults are percentages that use the wrong denominator, a "top item" that counts a modifier line as an item, and an invented "trend" that the data does not show. If the summary is wrong on a week you know well, do not rely on it for a week you do not.
A trial week plan
- Days 1 and 2: load the full menu with modifiers and combos, print every ticket type, and fix layout.
- Day 3: connect delivery apps and place test orders from each, with notes.
- Day 4: train staff on voids, discounts, sold-out and refunds; give each a card of steps.
- Day 5: run a quiet-night live trial in parallel with the old system, comparing totals.
- Day 6: test offline mode, card failure and a printer jam.
- Day 7: review faults, write them to the vendor and decide whether to proceed.
Do not switch on your busiest night. Choose a Tuesday, and keep the old system for a week. The double entry is annoying and costs less than one bad Friday.
Which AI features are worth a trial
Treat these as categories to test, not features every vendor has or charges the same for.
- Sales summaries in plain language. Useful if accurate and drawn from your own reports. Test with a week you know well.
- Menu or pricing suggestions. Risky; they ignore your food cost. Treat as ideas to check against a margin sheet.
- Phone ordering. An AI that takes calls and enters orders can save labour at peak; see the test checklist for AI phone ordering for takeaways before you trust it with numbered menus and delivery postcodes.
- Stock or prep forecasting. Needs months of data; ignore it in the first quarter.
- Item descriptions and photos. You can do this with a general assistant for free; see how to rewrite delivery app menus.
For each, ask three questions: can I see the output before it reaches a customer, does it work from my data, and what does it cost on top of the base plan? Anything charged per use should be capped.
Hardware bundles, minimum terms and exit costs
A discounted hardware bundle is attractive and may be fine, but read the commitment. Epos Now's offers, for example, show a discounted hardware bundle plus a monthly fee with "terms and conditions apply"; ask any vendor in that position whether the discount depends on a minimum payments or support term. For every vendor, find out:
- The minimum term and what ending early costs.
- Whether card processing is tied to the vendor, and the rate for different cards.
- Who owns the hardware if you leave.
- What support covers and when it is available. A takeaway needs help on Friday evening, not Monday morning.
- How price rises are notified and capped.
- Whether you can export your menu, customers and history.
Put the answers in the same table as the scoring sheet, with a total cost for the first 12 months. That is the number to compare.
Per-device extras that decide which system is cheapest
Line up the extras, because they decide which system is cheapest in practice: a customer-facing screen, a kitchen display, a second printer, extra tills, a card reader per driver, text messages, online ordering fees and any aggregator. Square's pricing page, for instance, lists its kitchen display app and kiosks as per-device monthly add-ons. Write each as a monthly or one-off cost for your set-up, and add them to the 12-month total.
Three takeaways, three choices
A single-shop takeaway with mostly phone and walk-in orders
Delivery integration matters less. A simple system with a reliable printer and clear reports, possibly on a free plan, is enough. Spend your attention on phone-order speed, and see how to set up an AI phone line for takeaway orders if calls cause queues.
A takeaway where half the orders come from apps
Integration is the whole decision. Pick on the delivery-app journey, price the aggregator if needed, and test pausing and sold-out across channels. Accept a higher monthly fee if it removes the retyping.
A small group of three takeaways
Central menu control, per-site reports and permission levels matter. Look at multi-site features and any "enterprise" tier. Ask for a demo using your three sites, because pricing at this level is often negotiated.
Retyping, remakes and misprints: the one-month review
After a month, compare against your baseline: minutes spent retyping app orders, wrong or remade orders, tickets that printed incorrectly, and the time to produce an end-of-day total. Add a fifth number, how often staff called support. If retyping is near zero and wrong orders fell, the integration works. If staff still ask "where is that order", the printing or status sync is faulty and you should raise it while you are still inside any trial or cancellation window. Keep a one-line log of every fault with date and channel; it is your evidence for support or for leaving.
For a wider view of which tasks are worth automating in a takeaway at all, read AI for takeaways: what is worth automating and what is not. If you want help mapping your own order route before choosing, that is what my AI implementation consultation covers; start from the contact page.
Takeaway EPOS questions owners ask
What matters most in a takeaway EPOS?
Orders from delivery apps and your own channel landing in one place, and a kitchen ticket that prints correctly every time. Reports and AI extras come after those. If an order has to be retyped from a tablet into the till, the system is costing you staff time every night.
Do I need an EPOS with AI features?
Not to run a takeaway well. AI features such as sales summaries, menu suggestions or phone ordering may help, but score them against real tasks and fees. Many vendor pages describe automation without giving details, so ask to see each feature working on your menu.
Should I take the cheap hardware bundle?
Read the contract first. Discounted hardware often comes with a minimum term or a tie to the vendor's card processing, set out in the terms rather than the headline. Calculate total cost across the term, including card fees and add-ons, before comparing it with a free-software option.
Can I switch EPOS later?
Yes, but plan for it. Ask whether you can export your menu, customers and sales history as files, whether hardware is tied to the vendor, and what notice the contract needs. A system that holds your data hostage is a cost you will pay once, at the worst moment.
Further reads
- AI Software for Restaurants: POS-Built AI or an Add-On Tool? — Built-in POS AI versus add-on tools for restaurants.
- Food Truck POS: Offline Mode, Fees and Which AI Features Pay Off — Which POS AI features a small food business should pay for.
- AI Ordering vs Delivery Apps: What a Takeaway Actually Pays — Own AI ordering compared with delivery-app commission.
- Kiosks AI for Cafés and Takeaways: Self-Service Ordering Costs — Kiosk costs if you add self-service ordering.
- Can a Takeaway Take Orders on WhatsApp With AI? — Taking orders on WhatsApp alongside your EPOS.
- Pizza Delivery Software: Which Options Handle Dispatch and Texts? — Which delivery tools assign drivers, promise realistic times and send on-its-way texts for a pizza shop, with list prices and a Friday-night walk-through.
- AI for Restaurant Order Management: One Queue for Every Channel — One queue for every channel: choose where it lives, map each menu with AI, test phone ordering, and run the checks that catch dropped or doubled orders.
- Pizza Shop Inventory: Plan Dough and Toppings From Weekend Orders — Turn eight weeks of orders into dough balls, cheese and topping quantities for Friday to Sunday, then tune the plan from a simple waste log.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Epos Now website; Square for Restaurants pricing page and Square AI page (viewed October 2026); vendor contract and plan pages to be checked on the day.