The right question isn't "does this POS sell faster?". It's "when the fibre goes down at 6pm on a Saturday during the sales, does the till keep scanning products?". Most chains choose point-of-sale software for the pretty screen and the e-commerce integration — and discover the wrong architecture at the first network incident, with a queue at the door and a store manager on the phone looking for someone to answer. This guide's thesis is uncomfortable: almost all the usual buying criteria are secondary. What separates a retail POS from an expensive toy comes down to three tests of two minutes each, and none of them appears on the first pages of a sales proposal. We give you the criteria to assess before you sign — and the right order in which to impose them.
What you need before you start
Before comparing proposals, gather these elements. Without them, any demonstration is theatre — the vendor runs the script and you applaud.
- Exact number of stores, tills per store and peak hours by region.
- Map of real connectivity — how many stores have redundant fibre and how many depend on a single operator that already failed last December.
- Average transaction volume per till/hour on the busiest day of the year.
- Inventory of invoicing programs already certified by the AT (obligation under DL 28/2019).
- List of active sales channels: physical store, e-commerce, marketplace, telephone.
- Loyalty rules and promotions you need to keep at go-live — not the ones you'd like to have one day.
- Who decides: usually the trio of CEO + CFO + head of systems.
The criterion that separates a POS from a toy: resiliente
A synchronous POS needs the connection to the central server to close each sale. It seems efficient until the day the router at the Braga store reboots on its own and the queue grows while the till shows an hourglass. An resiliente POS processes the sale locally, issues the receipt, and reconciles with the back office when the connection returns. The cashier doesn't even notice what happened on the network.
This distinction isn't technical — it's commercial. Every minute a till is stopped on a sales Saturday is revenue that doesn't come back, and the customer who left the basket on the floor rarely returns on Tuesday. MAXIRETAIL was designed with this logic: the till works without a connection and reconciles afterwards in the back office, with no manual intervention from the operator. Note the detail that separates the promise from reality — it isn't enough for the POS to "cache" the sale. It has to keep issuing a certified document with a valid ATCUD even offline, and then integrate it into the sequence without duplicating or skipping numbering. This is where many systems that claim to be resiliente fall apart: they sell without a network, but break fiscal integrity when they reconnect.
A POS that needs the internet to sell is not a retail system. It's a web form with a cash drawer.
Shared stock is not the same as synchronised stock
Here lies the most expensive confusion in omnichannel retail. Synchronisation means each channel has its own copy of the stock and the systems exchange periodic updates — every ten minutes, every hour. Between two synchronisations, the e-commerce sells the last piece of a reference that the Guimarães store has just sold over the counter. The result: cancelled order, irritated customer, scratched reputation and an apology e-mail nobody likes to write.
Shared stock is a single source of truth that all channels consult in real time. The difference shows up in BOPIS (buy online, pick up in store) and ship-from-store, where the promise to the customer depends on knowing exactly where each unit is — not where it was eight minutes ago. If you're assessing solutions for a chain, the omnichannel retail in Portugal guide goes deeper into the fulfilment models this enables.
Decision matrix — what to demand of each proposal
Take this table to the meeting. Mark each vendor with yes/partial/no and ask for a live demonstration of anything marked "yes". "Partial" is almost always a polite "no" — treat it as such.
| Criterion | Why it matters | How to validate |
|---|---|---|
| POS works without a connection | Till doesn't stop in a network failure | Disconnect the network in the demo and close a sale |
| Automatic reconciliation | No double keying when the connection returns | Reconnect and confirm the sale appears in the back office |
| Offline fiscal integrity | ATCUD and numbering don't break without a network | Make three offline sales and check the sequence on reconnecting |
| Real-time stock, not synchronised | Avoids selling what no longer exists | Request a simultaneous sale on two channels |
| AT certification + ATCUD | Legal obligation (DL 28/2019) | Software certificate number |
| Monthly SAF-T | Reporting to the AT (Ordinance 195/2020) | Sample export of the file |
| Architecture for franchises | Scales without multiplying administration | How do you open store 21? |
| Omnichannel BI | Deciding with data from all channels | Dashboard cross-referencing store + e-commerce |
On the second-to-last point: a POS that produces reports only per isolated store forces you to consolidate in spreadsheets on Sunday night. The Qlik Sense layer lets you cross-reference sales, margin and stock rotation by channel and by region on the same screen, with self-service BI for the operational manager to build their own views without waiting for a request to the systems department.
The detail nobody measures: the cost of opening the next store
We see this repeatedly in retail projects: the company assesses the POS by the pilot store and ignores the marginal effort of opening the tenth store. In a poorly designed architecture, each new store means reconfiguring product tables, re-importing promotions and training the manager from scratch. In an architecture built for chains, the new store inherits the central configuration and goes into production with a short onboarding.
Do the maths with your own numbers. If you open four stores a year, the difference between two days and two weeks of go-live per store is six weeks of central team work that no longer goes to putting out configuration fires. Retail commerce turnover in Portugal grew 4.7% in 2024 (INE, 2024), with total commerce reaching €201.8 billion. Those with operations ready to scale capture that demand without hiring a systems administrator for every five stores; those who don't, throttle it — or pay in overtime invisible in the budget.
Common mistakes and how to avoid them
Five patterns that recur in POS assessments, in order of cost. The first is the most seductive and the most expensive.
- Choosing by the sales screen. The pretty screen deceives. Ask for the demonstration of a network failure and end-of-day closing — that's where weak systems reveal themselves, and no salesperson offers that test of their own accord.
- Accepting "synchronisation" as omnichannel. Demand that the vendor explain whether the stock is shared or copied. The question "when do the channels stop being out of agreement?" kills the ambiguity in seconds.
- Ignoring compliance until the end. AT certification, ATCUD and SAF-T are not extras. Confirm the certificate number before proceeding, not on the eve of go-live.
- Underestimating store manager training. The most capable system fails if the cashier doesn't know how to process an omnichannel return of an item bought online and returned over the counter. Include operational scripts in the go-live plan, not just two hours of training.
- Buying modules you won't switch on. Loyalty, dynamic pricing and marketplace are useful — if they're on your roadmap for the next 12 months. If not, postpone and negotiate.
Reduce the decision to three tests
Choosing a retail POS doesn't need a hundred-line grid. It comes down to three tests any serious vendor will agree to do in front of you: disconnect the network and see if it sells, make a double sale across the two channels and see if the stock reacts, and ask for the AT certificate number before discussing anything else. A system that links BackStore and till in a truly resiliente operation passes all three — and that's the line that separates selling more from selling with control. For the strategic framing of the channel, the omnichannel retail trends for 2026 show where customer expectations are pushing Portuguese chains. If a vendor refuses any one of the three tests, you already have your answer.
Sources
- INE — Commerce Enterprises and Turnover, 2024 (retail commerce +4.7%; total commerce €201.8 billion).
- Decree-Law No. 28/2019 — invoicing and fiscal documents regime (AT certification, ATCUD).
- Ordinance No. 195/2020 — reporting of the SAF-T (PT) file to the Tax Authority.
Frequently asked questions
What is an resiliente POS and why does it matter in retail?
An resiliente POS processes sales locally without depending on the connection to the central server. When the network fails, the till keeps working, issuing receipts with a valid ATCUD and reconciling automatically afterwards. This avoids queues and lost sales in connectivity failures, common on sales Saturdays.
What is the difference between synchronised stock and shared stock?
Synchronised stock means periodic copies (every 10 minutes or hour) that can go out of sync between channels. Shared stock is a single source of truth consulted in real time by all channels. In BOPIS and ship-from-store, shared stock avoids selling what no longer exists and order cancellations.
How do you validate whether a POS is really resiliente in a demonstration?
Disconnect the network during the demo and close a sale. Check that it issues a receipt with a valid ATCUD. Then reconnect and confirm that the sale appears automatically in the back office without double keying and with correct sequential numbering. If this doesn't work, the system is not resiliente.
What legal certifications should you demand of a retail POS in Portugal?
AT (Tax Authority) certification and ATCUD in accordance with DL 28/2019 are mandatory. The system must generate a monthly SAF-T for reporting to the AT (Ordinance 195/2020). Always ask for the software certificate number and validate it directly with the AT before signing a contract.
Why does a POS with franchise architecture matter?
In a good architecture, opening store 21 means inheriting the central configuration and going into production quickly. In a weak architecture, each store requires reconfiguration of tables, re-importing of promotions and training from scratch. With four openings a year, the difference can be six weeks of central team work.
What is omnichannel BI and how does it help with the decision?
Omnichannel BI cross-references sales, margin and stock rotation across physical store, e-commerce and other channels in a single dashboard. It lets the operational manager create custom views without waiting for IT. A POS that only reports per isolated store forces you to consolidate in spreadsheets manually.
What is the most important test to do before signing with a POS vendor?
The two-minute test: disconnect the network, close a sale, and check that the receipt has a valid ATCUD. Reconnect and confirm automatic reconciliation in the back office. This simple test separates a real retail POS from a "web form with a cash drawer".
What elements should I gather before assessing POS proposals?
Exact number of stores, tills and peak hours; map of real connectivity (redundant fibre or not); transaction volume on the busiest day; AT-certified invoicing programs; active channels (store, e-commerce, marketplace); loyalty rules to keep; and identifying who decides (CEO, CFO, head of systems).