What Hardware Do You Need to Capture Retail Data?

When a retailer decides to get more from data, the first instinct is often a shopping list: sensors, counters, new cameras, new terminals. For most shops that list is far shorter than expected, because the hardware already on your floor captures most of the data you need. The real question is what happens to that data after it is captured.

In short: Most Singapore retailers already own the data-capture hardware that matters: the POS, existing CCTV, barcode scanners, and the ecommerce platform between them capture sales, footfall, stock movement, and online orders. The footfall piece is software (computer vision) running on the cameras you already have. New hardware earns its place only in specific cases, such as RFID for fast stocktakes of high-value inventory. The bigger gap, in almost every shop we see, is that the captured data sits in separate systems and never gets combined.

What data does a retail store actually need to capture?

Four streams cover most decisions a retailer makes:

• Sales: what sold, when, at what price, with what discount.
• Stock movement: what came in, what moved between locations, what is on the shelf.
• Footfall and in-store behaviour: how many people came in, where they went, how many bought.
• Customer identity: who bought, when they consented to be known, across visits and channels.

Each stream answers different questions, and the earlier posts in this series cover what becomes possible once the streams are combined into one source of truth.

What hardware do you already own that captures it?

Data stream	Hardware that captures it	Do you already own it?
Sales	The POS terminal	Yes
Stock movement	Barcode scanners and the POS	Usually
Footfall and behaviour	Your existing CCTV, with computer vision in software	Usually
Customer identity	The POS (loyalty at checkout) and your ecommerce platform	Usually

The row that surprises most retailers is footfall. Counting visitors does not require new people-counting sensors in most stores: computer vision runs on standard CCTV feeds, which is exactly how the retail vision setup on existing cameras works. Because cameras that analyse people involve personal data, PDPA applies to how it is set up; done properly, no faces are stored and no individuals are tracked.

When do you actually need new hardware?

Three cases justify buying new equipment:

• RFID tags and readers, when you carry high-value stock and stocktakes by barcode take hours. RFID counts a rack in seconds; it pays for itself through counting time and accuracy, and only above a certain stock value.
• A camera where you have none, such as an entrance with no CCTV coverage at all. The camera itself is standard equipment; nothing exotic is needed.
• Label printers and scanners where stock still moves with handwritten tags.

If a vendor's proposal opens with a hardware list rather than a question about your current systems, ask what each item captures that your existing equipment does not. In most quotes, only one or two items pass that test.

Why isn't the data you already capture doing more?

The missing layer comes after capture. The POS has been logging every sale since the day it was installed, and the CCTV has been recording since it went up. What most shops lack:

• The sales data and the stock data sit in different systems, so sell-through and reorder points cannot be computed.
• The footfall on the recordings is never counted, so conversion is unknown.
• The customer at the till and the customer online are two records, so their combined value cannot be calculated.

Combining those streams is software and integration work on hardware you already own. That is the single source of truth the rest of this series is built on, and it is usually a far smaller spend than a hardware refresh.

Is it worth spending on capture at all?

Buy hardware when:

• A data stream is uncaptured: no camera at the entrance, stock moving on paper.
• Stocktake hours are heavy and the stock value justifies RFID.
• A capture device is broken or so old it cannot export its data.

Hold off when:

• The stream is already captured and unused. Count what the CCTV already sees and join what the POS already logs before buying anything.
• The proposal is a platform bundled with hardware you do not need. If you want a second opinion on a quote like that, a two-week audit (from SGD 4,000) will price the software-only path against it.

What to do next

Walk the shop once with a short checklist:

• List every device that captures data today: POS, cameras, scanners, and where each one's data ends up.
• Mark which streams from the table above are captured but unused.
• Note the one stream that is missing, if any.

Most retailers finish that walk with a list of software and integration work rather than new equipment.