The premise
Most organizations run on two numbers: what they sold, and what it cost them. Both are recorded after the fact, and both describe outcomes rather than causes. The number almost nobody has is the one that explains the other two; how many people were actually there. Without it, a bad month is unexplainable. Was traffic down? Did people come and leave without buying? Were you understaffed at the moment demand peaked? Did the promotion bring anyone through the door, or did it just discount customers who were coming anyway? Each of those has a completely different remedy, and each costs a completely different amount of money to get wrong. People counting supplies the missing denominator. On its own it is a curiosity. Connected to your point-of-sale or ERP system, it becomes the basis for most of the operational decisions you currently make on instinct.
That connection is the point of this article.
What the system measures
Parcy deploys overhead sensors, video analytics or a hybrid of both, depending on site geometry and accuracy requirements. The raw outputs are:
- Entries and exits, directional, timestamped to the second
- Occupancy, live and historical, at building or zone level
- Zone-to-zone movement, where multiple sensors are deployed
- Dwell time, derived from entry and exit timestamps per zone
- Queue length and wait duration at service points
- Passer-by counts at the facade, where an outward-facing sensor is installed
These feed into Fusion, Parcy’s reconciliation and analytics platform. Fusion is API-enabled by design: it does not assume it is the system of record for anything, and it is built to be joined against the data you already hold.
The integration is the value
Counting hardware is widely available. What distinguishes a useful deployment from an expensive one is whether the count is joined to your commercial data, and the ability to generate strategic insight and value from the data.
Fusion exposes a documented REST API and supports scheduled or event-driven exchange with:
- Point of sale: transaction counts, line items, basket values, timestamps
- ERP: inventory movement, stock position, purchasing, cost data
- Workforce and rostering systems: scheduled hours, actual clock-in data
- Marketing and campaign platforms: campaign windows, spend, channel
- BI tooling: Power BI, Looker, Tableau, or direct database access
Data can move in either direction. If your POS can push a webhook on transaction close, Fusion consumes it; if it cannot, Fusion polls, ingests a scheduled export, or reads directly from the database. Where you would rather keep everything in your own warehouse, Fusion exposes counts as an API for you to pull and does not require you to adopt our dashboard at all.
The practical consequence: every metric below is available because two systems were joined, not because a sensor was installed.
The metrics that change decisions
Conversion rate (transactions ÷ visitors)
The single most important retail metric that most operators do not have.
Two branches with identical revenue can have entirely opposite problems. One draws 4,000 visitors and converts 12%. The other draws 1,400 and converts 34%. The first has a sales floor problem; staffing, product availability, layout, service. The second has a traffic problem; location, signage, marketing reach. The interventions are unrelated and the budgets are unrelated, yet the P&L shows the same figure for both.
Conversion tells you which conversation to have.
Capture rate (entries ÷ passers-by)
Of the people who walked past your frontage, what proportion came in? This isolates the effectiveness of your facade, window display, signage and street-level presence from everything that happens once someone is inside. It is the only clean way to evaluate a window change or a new fascia.
Sales per visitor (revenue ÷ visitors)
Revenue normalized by opportunity. It separates a genuinely improving store from one that is merely benefiting from a busier month, and it makes stores of different sizes and locations legitimately comparable.
Dwell time and zone engagement
How long people stay, and where. Long dwell with low conversion indicates confusion, poor product findability or insufficient staff coverage. Short dwell at the entrance indicates people are looking in, not finding a reason to proceed, and leaving. Zone counts show which departments actually pull traffic deeper into the floor and which are dead space paying rent.
Queue length, wait time and abandonment
Wait time correlates directly to abandonment, and abandonment is measurable lost revenue. Joined to POS timestamps, Fusion can estimate the value of transactions lost to queueing, usually the fastest ROI argument in the whole system, because it converts a service complaint into a number.
Traffic to labor ratio
Labour is the largest controllable cost in most retail and service operations, and rosters are usually built from habit rather than demand. Overlaying arrival curves on scheduled hours almost always reveals hours sitting in the wrong part of the week. Reallocating them costs nothing.
Occupancy and utilization
For offices, clinics, event spaces and facilities: what you are paying for versus what you are using. A tenancy costing KES 400,000 a month with a peak occupancy of 45% is a renegotiation waiting to happen, but only if you can evidence it.
By sector
Retail chains and franchise operators
Rank branches on execution rather than revenue. Compare conversion across the estate to find the stores where the traffic is arriving and the sale is being lost. Measure promotions properly e.g. a campaign can move footfall, conversion or basket size, and knowing which one it moved determines whether you run it again. Use zone data to plan layout and adjacency with evidence rather than planogram convention.
Shopping centers, landlords and property managers
Footfall is the currency of leasing negotiation, and it is almost never independently verified. Parcy provides:
- Centre level and entrance level traffic reporting, auditable
- Tenant level spillover analysis which anchors actually generate traffic for which tenants
- Evidence for rent differentials by position within the center
- Verified figures for tenant reporting and marketing-levy justification
For tenants, the same data supports the other side of that negotiation: rent per visitor, and rent per converted visitor, by site.
Quick service, restaurants and hospitality
Arrival-rate curves at 15-minute granularity drive shift design. Queue analytics quantify abandonment. Occupancy supports capacity planning and licensing compliance.
Banking halls, service centers and government offices
Wait times and service-point loading, with counts reconciled against ticketing or CRM records. Supports both service-level reporting and staffing decisions, and provides objective evidence for citizen or customer service commitments.
Healthcare facilities
Patient flow through reception, triage, consultation and pharmacy. Bottleneck identification between stages. Waiting-area occupancy for both experience and infection-control purposes. Most clinics estimate these figures; few measure them.
Public space, transport hubs and urban projects
Pedestrian counts are the evidentiary base for street design, non-motorized transport investment, market planning and donor funded urban programmes. Before and after measurement is frequently a funding requirement rather than an optional extra, and retrofitting it after the intervention is impossible.
Events and venues
Verified attendance rather than claimed attendance for sponsors, for licensing, and for capacity management in real time.
Corporate workplaces and facilities
Desk, floor and meeting-room utilization against lease cost. In a hybrid working environment this is one of the few property decisions that can be made on hard data.
From metrics to strategy
Operational reporting is the entry point. The larger value sits in the decisions that traffic history makes answerable:
Site selection. A multi-site baseline lets you model expected traffic for a proposed location instead of accepting the figures you are given. Capex decisions of this size deserve independent measurement.
Lease negotiation and renewal. Rent per visitor and rent per converted visitor across a portfolio, measured over months, is the strongest position a tenant can hold in a rent review.
Closure and consolidation. Distinguishing a declining catchment from a poorly managed branch is a decision worth millions of shillings, and it is currently made on intuition in most organizations. Traffic and conversion trends separate the two definitively.
Trading hours. Marginal revenue against marginal staff cost, hour by hour. Opening or closing an hour earlier is often worth more than any single merchandising change.
Marketing allocation. Traffic lift per shilling, by channel and by site. Campaigns that move reported impressions but not observed footfall stop being funded. Labour planning. Rosters aligned to measured demand rather than to convention, typically a 5–15% efficiency gain that pays for the system itself.
Reconciliation: measuring claims against observation
Parcy’s core discipline is the comparison of what was claimed against what was observed. Fusion applies the same engine here:
| Claimed | Observed | Why it matters |
| Landlord footfall figures | Independent sensor count | Lease and service-charge negotiation |
| Campaign reach and impressions | In-store traffic lift | Marketing accountability |
| Rostered staff hours | Measured presence | Payroll and contractor verification |
| Event attendance reported to sponsors | Counted attendance | Sponsorship assurance |
| Transactions recorded at POS | Visitors and queue observations | Shrinkage and till-discipline indicators |
Every reconciliation is reported with an explicit confidence band. We do not present estimates as certainties.
Accuracy, and what we will tell you before you buy
Counting accuracy varies materially by technology and installation, and any vendor who quotes a single number for all conditions is not being straight with you.
- Overhead stereo-vision or time-of-flight sensors, correctly mounted, deliver 95–98% accuracy under normal conditions.
- Video analytics on existing CCTV is far cheaper to deploy but depends entirely on camera angle, lighting and crowd density. Where cameras were installed for security rather than counting, which is nearly always and thus accuracy can fall well below the figure above.
We conducts a site survey and error-budget assessment before quoting, and states the expected accuracy for your specific installation in writing. Where existing CCTV is adequate we will say so and save you the hardware. Where it is not, we will tell you that rather than sell you a system that will quietly under-report.
Staff exclusion is handled explicitly, through directional logic, height filtering or RFID/BLE staff tagging using Parcy’s existing RFID stack. Uncorrected staff movement is the most common cause of inflated counts and corrupted conversion figures in poorly implemented systems.
Reports carry confidence intervals. A conversion rate is only as trustworthy as the count beneath it, and we would rather you knew the margin than assumed there wasn’t one.
Privacy and regulatory position
Parcy’s counting deployments are designed as anonymous by default and structured for compliance with the Kenya Data Protection Act, 2019.
- Counts are aggregate. No facial recognition, no demographic profiling, no cross-visit re-identification unless a client specifically requires it and can establish a lawful basis for it.
- Where video analytics is used, processing occurs at the edge wherever technically possible, with counts rather than footage transmitted.
- Retention periods are configurable and defined at contract stage.
- We support clients with signage requirements, records of processing activities, and Data Protection Impact Assessments where the deployment warrants one.
This is a deliberate design position. Analytics that create regulatory exposure are not analytics you can rely on long term.
Commercial structure
Parcy does not only sell counting hardware as a product. We deliver an outcome:
Deployment — site survey, error-budget assessment, sensor specification and installation, integration with your POS, ERP or workforce systems, and metric definition agreed with your team.
Ongoing service — a live dashboard, API access to your own data, a monthly Traffic and Conversion Report against your defined metric set, and a quarterly review session in which we interpret the results with your management team rather than leaving you to interpret a dashboard alone.
Extension — additional sites, additional zones, and integration into wider Parcy capability where relevant: RFID stock accuracy, access control, or reconciliation across the supply chain.
Pricing is per site and per zone, with the analytics subscription structured around the reporting and integration scope rather than the sensor count.
Where to begin
The most productive first step is a single site, fully instrumented and fully integrated, running for one trading quarter. That produces a baseline, a proven integration path into your POS or ERP, and a measured accuracy figure after which extending across a portfolio is a commercial decision rather than a technical experiment.
We are happy to conduct an initial site survey and integration assessment at no cost, and to tell you plainly if the numbers will not justify the investment.