Setting up a video analytics system for foot traffic counting
Australian retailers are entering a period where knowing exactly how many people walk through the door is no longer a luxury reserved for flagship stores in Sydney or Melbourne. With shop rents climbing across suburban shopping strips and Westfield centres, even a single-venue operator in Parramatta or Geelong can justify spending a few thousand dollars on hardware that turns an entrance into a live data feed.
A video analytics system for foot traffic counting uses computer vision and, increasingly, edge AI to identify and track people as they move through a store. The most basic setups simply count heads at the doorway, while more sophisticated deployments can distinguish staff from shoppers, measure dwell time at end-caps, and even flag queue lengths at service counters.
The aim of this walkthrough is to give store owners, operations managers, and tech integrators a practical path from "we should probably do this" to a working system that delivers reliable numbers every day.
Mapping out what your store actually needs
Before looking at any hardware, walk the floor with a notebook and sketch the customer journey. Where do most people come in? Is there a secondary door near the loading dock that gets propped open on hot arvos? Are there pinch points near the fitting rooms or the deli counter where a single camera could cover a wide section of aisle?
A typical 200-square-metre store might only need two or three devices to cover its main entrance, the queue area, and one high-value zone. Larger format shops, such as a JB Hi-Fi or a Bunnings warehouse, often deploy ten or more sensors because their floor plates are sprawling and they want granular data by department.
Equally important is deciding what success looks like. If the goal is simply to compare weekday and weekend trade, a basic counter on the front door will do the job. If the goal is to measure conversion against POS data, the system needs to integrate with the till and capture enough demographic detail to be useful without crossing privacy lines.
Comparing the hardware options available
Once the brief is clear, the next decision is which sensing technology fits the site. Different approaches suit different layouts, lighting conditions, and budget levels. The summary below captures the most common options used in Australian retail today.
| Technology | Typical accuracy | Best suited to | Indicative cost (AUD) | Privacy profile |
|---|---|---|---|---|
| 3D stereo overhead cameras | 95–99% | Storefronts and entrances | $800–$3,000 per unit | Low – no personal data captured |
| 2D wide-angle cameras | 85–92% | Open-plan floors | $300–$1,200 per unit | Moderate – resolution matters |
| Thermal sensors | 90–96% | Low-light or outdoor zones | $600–$2,500 per unit | Low – no imagery recorded |
| Wi-Fi and Bluetooth probes | 70–85% | Large venues, malls | $200–$900 per unit | Higher – tracks personal devices |
For most small to mid-sized Australian stores, 3D stereo cameras offer the best trade-off between accuracy and privacy because they capture depth data rather than identifiable video. They are also forgiving of shadows and harsh afternoon sun coming through plate glass, which is a recurring issue in west-facing shopfronts.
Selecting software and AI analytics tools
Hardware is only half the story. The software layer turns a stream of raw detections into counts, heat maps, and trend reports. Most modern platforms run some form of person detection on the device itself, sending only anonymised counts to the cloud rather than full video frames.
When evaluating vendors, look for a few practical features: real-time dashboards accessible from a laptop or phone, hourly and daily aggregations that can be exported to CSV, and integration hooks for existing tools such as a POS system, a workforce scheduler, or a marketing platform like Mailchimp or Klaviyo. Pricing models vary widely, with some charging a one-off licence per camera and others running a monthly subscription per site.
It is also worth asking whether the vendor supports on-premise deployment. Retailers handling sensitive customer data, or those operating in regions with strict data residency requirements, sometimes prefer to keep footage and analytics entirely on a local server rather than pushing it to an overseas cloud. This is a common request from Australian operators working with health, government, or financial services clients.
Meeting Australian privacy obligations
Foot traffic counting sits within the scope of the Privacy Act 1988, particularly the Australian Privacy Principles (APPs), if the business has an annual turnover above the current threshold or handles specific categories of personal information. Even where the legal threshold does not apply, adopting privacy-by-design practices is sensible and builds customer trust.
The first step is a clear signage strategy. A sign at each entrance stating that anonymous visitor counting is in use, along with a contact email for questions, satisfies most reasonable expectations. The second step is technical: ensure that the chosen system does not capture or store facial imagery unless there is a separate, documented purpose that has been communicated.
Independent retailers should also publish a short privacy notice on their website explaining what data is collected, how long it is retained, and who it is shared with. Reviewing this document annually helps keep the operation aligned with any updates to the Office of the Australian Information Commissioner's guidance.
Installing cameras and network infrastructure
Installation day usually takes a few hours per camera and is best scheduled outside trading hours. Overhead mounts at the entrance typically sit between 2.4 and 3 metres, which gives a clean top-down view and minimises the chance of shoppers blocking each other. Cabling should run through existing conduit where possible, with PoE (Power over Ethernet) switches keeping the wiring simple.
For stores without structured cabling, wireless IP cameras are an option, although they require a solid Wi-Fi signal at the mounting point. Running a quick speed test before mounting the bracket can save a return visit later. Outdoor entrances, such as those leading from a Westfield car park, may need weather-rated housings and surge protection.
Configuration is mostly done through the vendor's web interface. Setting the count line correctly, defining the entry direction, and excluding staff-only doors are three tweaks that noticeably improve data quality from day one. A short walk-through test with a colleague entering and exiting several times confirms the logic is sound.
Calibrating and validating accuracy
No system is perfect out of the box. The first week of operation should be treated as a calibration period, with manual spot-checks at different times of day to compare the reported counts against reality. A simple clipboard tally at the door for an hour each morning during the trial helps establish a baseline.
Environmental factors can throw off readings. Sunlight streaming through glass at midday, a banner promotion, or even a popular footy match on the telly can change the patterns the system sees. Most platforms allow sensitivity adjustments and exclusion zones for known trouble spots, such as a poster display that reflects light strangely.
Once the numbers stabilise, document the configuration. Future staff members will thank you for a short runbook explaining where each camera is mounted, what it counts, and how to reset it if it goes offline after a power outage.
Turning visitor counts into retail decisions
A reliable count is valuable, but the real return comes from acting on it. Comparing entry counts against POS transactions reveals the conversion rate, which is often more telling than total sales. A store with steady traffic and weak conversion might need a layout refresh, while one with strong conversion but light traffic might benefit from a local marketing push.
Hourly patterns inform staffing. If the data shows a sharp lift between 11am and 1pm on weekdays, scheduling an extra team member during that window reduces queue times and lifts basket size. The same data can be shared with suppliers to justify premium end-cap placement or to renegotiate lease terms with a landlord based on proven footfall.
Heat maps generated from camera footage add another layer, showing where customers linger and which aisles they skip. Pairing this with planogram adjustments tends to deliver quick wins, particularly in stores where the merchandising team has been working from intuition rather than evidence.
Ready to take the next step
For retailers who would rather hand the project to an experienced integrator, look for a provider that handles site survey, hardware procurement, installation, software configuration, and ongoing support under a single agreement. Bundled offerings of this kind have become common, and even operators outside Australia, such as a Y!mobile shop in Aizu, now pair their core retail business with on-site ICT deployments. The same model works well for an Australian independent looking for one accountable partner rather than three separate contractors.
Start with a clear brief, choose technology that matches the floor plan, respect customer privacy from day one, and treat the first month as a learning period. The result is a system that pays for itself many times over through smarter rosters, sharper merchandising, and the confidence that comes from knowing exactly how the store is performing.