Buying POS software in 2026: the tests to run before you pay for anything

Most POS advice reads like a feature-list war. On a vendor website, any two systems look identical: both scan barcodes, both print receipts, both promise a real-time dashboard. The differences show up later — in week three, when the internet drops during a rush, or in year three, when an accountant asks for a closed-day report from a quarter nobody remembers. So this is not another comparison. It is the sequence of tests that separates a system you will still like next year from one you quietly abandon. The complete version, with vendor notes and the pre-purchase checklist, sits in the full buyers guide on tanit365.com; what follows is the part you can actually rehearse during a trial.

What the demo has to prove, not show

Vendors demo on their own sample catalog. Flip it: bring your twenty fastest-moving products, including the size and colour variants that always break imports, and run the demo on those.

• Kill the Wi-Fi mid-demo. A counter that stops selling the moment the router blinks is disqualified — offline mode that actually works means sales continue and sync when the connection returns.
• Run a two-item sale and count the taps. A trained cashier completes a sale like that in under 15 seconds; every extra confirmation screen gets paid for hundreds of times a day.
• Ask for a refund and a discount, each with a manager PIN. If either can happen without one, and no audit log records who did it, that gap eventually becomes your shrinkage.
• Open the reports screen and look for three answers without scrolling: what sold today, what is running low, and what earns the most margin. Reports that need an accountant to interpret do not get read.

The year-one number almost everyone undercounts

Monthly licence math hides the real bill. A realistic spread for a small retailer looks like this:

• Licence or subscription: about $10–$50 per device each month — and watch per-user pricing that multiplies with staff count.
• Setup and data migration: $50–$300 one time. A "free" setup that skips product variants costs more later.
• Hardware: $200–$600 per counter, with proprietary terminals as the lock-in trap.
• Payment fees: 1.5–2.5% per card or wallet transaction. Bundled acquiring with poor rates quietly eats margin.
• Support: sometimes included, otherwise $10–$30 per month.

Add it up for a one-counter shop in Vietnam and year one lands around $300–$700 with hardware included. A quote dramatically below that range usually means offline mode or support got quietly removed. Price year one, not month one.

The records you will need in three years

Licence fees are easy to compare. What decides whether you can answer a question three years from now is the archive behind the till: daily close-outs and Z-reports, the e-invoice data your tax filing relied on, supplier invoices proving cost of goods, price-change history, and the staff log showing who discounted what.

When the whole system lives on one counter tablet — exports in a Downloads folder, or inside a vendor portal that shuts the day the subscription lapses — one wiped device takes a year of history with it. Whatever storage you choose, keep three copies of every export that matters: one on the working machine, one on separate storage, one off site in the cloud. Version history is worth checking too, because it is what pulls back a report that was overwritten. And before paying for any plan, verify whether the capacity is genuinely a one-time purchase or renews at a higher price, and whether it holds several years of exports rather than one season's worth.

Five ways buyers talk themselves into the wrong system

1. Buying for badges. Restaurant modules are useless in a clothing store — pay for fit, not feature lists.
2. Skipping the trial with real products. Twenty real SKUs with variants will break a system that a glossy demo sails straight through.
3. Ignoring the exit. Ask in writing how you get your data out. Proprietary formats are hostage risk, not a feature.
4. No training budget. Half a day of training per cashier is the realistic minimum; adoption decides ROI more than feature count does.
5. Hardware first. Pick the software that fits the workflow, then buy compatible hardware — never the reverse.

The exit test: run one real week

Before signing anything, run the trial the way a real week runs. Sell, refund, discount and receive stock. Then export everything and open the files on a different machine. If the exports open cleanly and the stock numbers still match the counter, the system passes. If anything exists only inside the vendor dashboard, treat that as a risk, not a feature. The same principle the backup routine guide on tanit365.com hammers on: a backup you have never restored is a wish, not a backup.

Two questions that come up every time

Should the POS and the inventory tool come from the same vendor? Not necessarily. The only non-negotiable is two-way stock movement with no manual patching in between: ring something up at the till and the warehouse count drops; receive goods against a note and the counter immediately knows it can sell them. Whatever setup survives the trial with the fewest extra steps is the one to keep.

What happens if the counter tablet is lost or wiped? Nothing bad — if the habits above are in place. Exports on a fixed schedule, three copies, older versions kept, and files opened on a second machine to prove they read back. The habit matters more than the tool; it is what lets you answer a tax or supplier question long after the sale.

Where the full guide lives

This is the compressed version. The full walkthrough — feature requirements for retail in 2026, the complete cost table, the pre-purchase checklist and the back-office test — is on the original post at tanit365.com. Take one week, test two systems against it with your own products, and the decision makes itself.