Cash register vs POS system: the honest line between them

The cash register versus POS system question is usually presented as old versus new, and that framing is wrong. Plenty of shops running a cash register in 2026 are making the correct decision, and plenty running a POS system are paying a subscription for capabilities they have never opened.

The real difference is one sentence:

A cash register records what you took. A POS system records what you sold.

Everything else follows from that. If you only need to know how much money came in, a register does it, and it does it faster, cheaper and more reliably than any software. If you need to know which things left the shelf, a register cannot help you and never will, because it does not know what a product is.

This is where the line actually falls, what pushes a shop across it, and what upgrading costs you beyond the price.

What a cash register genuinely does well

It is worth stating this properly, because most comparisons are written by people selling the alternative.

It works. No internet, no account, no updates, no vendor. A power cut is the only outage.
It is fast. A trained hand on a mechanical keypad beats a touchscreen, and there is no loading.
It costs nothing to run. One purchase, no subscription, no per-terminal fee, no percentage.
Nobody has to learn anything. Staff turnover is a real cost in retail, and a register can be taught in ten minutes.
It is nobody's data but yours — because there is barely any data. That is a limitation and, in a small way, a freedom.
Nothing to migrate, ever. No lock-in, because there is nothing to be locked into.

If your shop is one counter, a limited range, cash-heavy, and you already know your margins because you set them, a register is not a compromise. It is the right tool, and replacing it with software will make your day slower before it makes anything better.

What it structurally cannot do

Not "does badly" — cannot, because it has no concept of a product.

Stock. It does not know you sold three of something, so it cannot tell you that you have four left. Every stock question becomes a physical count.
Per-product margin. Total takings tell you nothing about which lines make money. Most shops discover their worst-performing line is one they assumed was fine.
A return against the original sale. With a register, a refund is money out. Nothing links it to what was sold, so nothing goes back into stock and nothing stops the same receipt being refunded twice.
Reordering. No reorder points, no low-stock list, no purchase history to buy against.
Loyalty or customer history. No customer exists in a register's world.
Theft detection. This is the important one and it is covered below.
More than one location. Two registers are two islands with no shared view.

None of that matters until it does. The list above is not an argument for upgrading; it is the list of things you will be doing manually, and whether that is acceptable depends entirely on how many lines you carry.

The four triggers

In practice, shops cross the line for one of four reasons — plus a fifth, further down, that comes from the bookkeeping rather than the shop floor. If none of them describes you, the register is still the right answer.

1 · You carry more lines than you can hold in your head. This is the most common trigger and the most honest one. Somewhere between fifty and a few hundred SKUs, physical stock counting stops being a job you can do properly and starts being a job you skip. The moment you are guessing what to reorder, you are paying for the register in lost sales and dead stock.

2 · You want to know what is actually selling. Not what you feel is selling. Almost every retailer who moves to a POS reports the same surprise in the first month — the ranking is not what they expected, and one line they were about to drop was carrying a category.

3 · You are not always in the shop. This is the trigger that matters most and gets stated least, so it gets its own section.

4 · A second location, or a second sales channel. The moment stock exists in two places, a register cannot give you a consolidated view, and a spreadsheet reconciled weekly is a job nobody wants.

The trigger nobody says out loud

A cash register cannot tell you that a member of staff rang up a sale, took the cash, and voided the transaction afterwards.

This is the single most common form of retail loss, it is invisible in a register's totals — the drawer balances, because the sale was removed — and it is the reason the "not always in the shop" trigger matters more than the stock one. If you are behind the counter every hour you are open, you have a control. If you are not, you need the system to have one.

What a POS gives you here, if it is built properly, is threefold: a void marks a sale rather than erasing it, so voids appear in the day's report; a refund needs approval from a second person who is not the one performing it; and a shift close records expected versus counted cash with the variance attributed to whoever counted.

The word "properly" is load-bearing. We went through 60 real POS products and counted what each one ships. A full register and cash-close lifecycle appears in only 18 of the 60. Staff permissions with approval at the point of work appear in 16. These are the two least common capabilities in the market, and they are precisely the two that address this trigger.

So the honest advice is not "get a POS". It is: if this is why you are upgrading, check for those two capabilities specifically, because most of the market does not have them and none of the marketing mentions it. A POS without them is a register with better reports and the same blind spot.

The fifth trigger: what your accountant needs

This one belongs on the list and almost never appears there, because the person doing the bookkeeping is rarely the person choosing the till.

A cash register gives you a daily total, and if it is a decent one, a breakdown by department and by tender. That is genuinely enough for a simple shop with a single tax rate and a bookkeeper who is happy with a monthly envelope of Z-readings.

It stops being enough at a specific point: mixed tax rates. A shop selling both standard-rated and zero-rated goods — most food retail, a lot of general stores, anywhere with a mix of products and services — needs takings split by rate, every day, reliably. A register can do this only if every product was assigned to the right department every time it was rung in, by whoever was on the counter. In practice that assignment is where the errors are, and they are invisible until a return is being prepared.

A POS assigns the rate to the product once, at setup, and applies it automatically thereafter. That is the whole difference, and for a mixed-rate shop it is worth more than the stock reporting.

Three questions worth asking your bookkeeper before you choose, since they will be living with the output:

What do you want to receive, and in what format? The answer is usually a categorised monthly summary and, increasingly, a file rather than paper.
Do you reconcile card settlements to sales? Card takings arrive net of fees and often a day late, and a system that records the gross sale and the net settlement separately saves real time every month.
What do we do about cash discrepancies now? If the answer is "nothing, we do not track them", that is a finding in itself.

Whatever your jurisdiction requires by way of receipts and record retention is worth checking before you buy rather than after — the rules vary widely, some places mandate specific fiscal hardware, and this is one of the few areas where a local supplier genuinely knows something a comparison site does not.

What about the hardware you already have?

A practical point that decides a surprising number of these choices. Most POS software will drive a standard receipt printer and a cash drawer connected through it, and most will work with a standard barcode scanner, because these are old, boring, well-standardised devices.

What is usually not reusable is anything proprietary — a register's own printer, an integrated keyboard-and-display unit, or a card terminal tied to a specific processor. Ask about your existing kit by model number before you assume either way. Being able to keep a working printer and drawer takes a meaningful chunk out of the switching cost, and being told late that you cannot is how a modest upgrade becomes an unplanned purchase.

The middle option, and where it breaks

Most shops do not go straight from register to POS. They run a register for money and a spreadsheet for stock, and this is a genuinely sensible intermediate step that nobody recommends because nobody sells it.

It works, for a while. It breaks in one specific way: the two never agree, and there is no way to find out which is wrong. The register knows takings; the spreadsheet knows what you think you have. Neither knows what was sold, so a discrepancy has no explanation and no trail. You cannot reconstruct it, because the connecting fact — the sale line — was never recorded anywhere.

If you are running this today, the highest-value change is not buying software. It is restructuring the spreadsheet so stock is derived from a movement log rather than typed into a cell, which at least gives you a history to argue with. We have written that up in detail in inventory management in Excel — the same rule that separates good retail software from bad, expressed as three tabs.

What upgrading actually costs you

Beyond the money, which is covered in POS system cost, there are four costs people meet after they have committed.

A dependency you did not have. A register does not care about your internet, your vendor's uptime, or whether the subscription lapsed. A POS does. Ask what happens when the connection drops mid-transaction, and get a demonstration rather than an answer.

A subscription that scales with you. The pricing is usually per terminal. Growing means paying more, which is fair, and it means the second counter costs twice what you budgeted.

Staff training, repeatedly. Retail turnover is high. Every new hire has to learn it, and the cost is real and recurring in a way the one-off training estimate never captures.

Your data becomes somebody else's problem to give back. Ask what a full export contains before you put two years of history in. This is the single most-skipped question in the category, and the one people regret.

There is also a cost that goes the other way and is worth naming: you have nothing to migrate. Coming from a register, there is no historical data to move, which makes this the cheapest and lowest-risk moment you will ever have to change systems. If you are going to move, moving now is strictly easier than moving in three years.

If you do move, what to insist on

The counted core across 60 products, in order of how many ship it: the sell screen (60 of 60), stock visibility with movement history (58), reporting that drills to source rows (51), customer context (44), tender handling including split payments (36), multi-location (34), returns and exchanges (29), cash close (18), staff approval (16), purchasing (13).

Read that list against whichever trigger moved you. If you are upgrading for stock, the top three matter. If you are upgrading because you are not always in the shop, the bottom two matter and the top three are table stakes — and the bottom two are exactly where the market thins out.

The fuller version, with a test to run for each capability, is in POS system for small business.

Building your own till

Worth mentioning because it has genuinely changed. A coding agent will produce a working till in a couple of days — we have three complete ones, built from a single specification by three different models, running as live demos.

What it will not produce unprompted is the bottom of that list. One of the three ran its own review and found a hole where a single person could ring up a sale, take the cash and erase the record afterwards; it closed that off, routed corrections through returns so money, stock and points reverse together, and made self-approval impossible. It found that because the specification made it look. Left to guess, a model builds the sell screen and the stock list and stops, because that is what "build me a POS" has always meant in the material it learned from.

So building is right if you can name those controls, and it reproduces a register's blind spot with a nicer interface if you cannot.

The decision, in one table

Your situation	Answer
One counter, few lines, you are always there, cash-heavy	Cash register
Few lines, but you want takings by product	Register + a properly structured spreadsheet
More lines than you can count in your head	POS
You are not always in the shop	POS — and check for cash close and second-signature approval
Second location or a second sales channel	POS
You sell across more than one tax rate	POS — the assignment happens once, not every sale
Your products do not fit a standard catalogue, or you run several tills	Consider building
You are upgrading because it feels dated	Stay where you are and revisit at one of the four triggers

The last row is not a joke. "It feels old" is the most expensive reason to change a system that works, and it is the reason most often given.

POS system for small business has the full counted list with a test for each capability. POS system cost works through the arithmetic, and free POS software covers where the free tiers stop.

If stock rather than the till is your real problem, what a retail inventory management system includes is the counted inventory for that side.

And if you decide to build, the Retail POS planner is the specification we hand our own agents — cashier-first, with the cash close, the approval path and the reversal rules written in so they cannot be skipped.