Requirements Guide for People Counting Tender

Most people counting tenders fail before a single vendor responds. The failure point is almost always the same: the requirements document specifies hardware — camera models, mounting heights, sensor brands — instead of outcomes. The result is a tender that locks the buyer into one supplier's ecosystem for a decade, kills competitive pricing at renewal, and makes it commercially painful to replace a sensor that underperforms in year three. If you write only one thing into your next people counting tender, make it this: the software layer must be sensor-agnostic, and accuracy must be contractually guaranteed, not marketed.

Start with the accuracy clause — and make it contractual

Every vendor deck claims "up to 99% accuracy." A tender is not the place for "up to." Your requirements should demand a contractual minimum accuracy figure, a defined audit methodology, and a remedy if the vendor misses it. Be realistic in what you ask for: serious providers will commit to a contractual minimum of 96%, typically achieving 98–99% in practice when lighting, entrance layout and visitor behaviour allow. Vemco Group, whose counting software has run since 2005 and now processes more than 85 million counts per day, offers exactly this structure — a 96% contractual floor with typical performance well above it. A vendor who refuses any contractual accuracy commitment is telling you something important.

Specify the audit method in the tender itself. A workable clause: manual video-based validation over a minimum sample of visitors per entrance, conducted at agreed intervals during peak and off-peak periods, with accuracy calculated per entrance rather than averaged across the estate. Estate-wide averages hide a broken entrance behind twenty good ones.

Separate the hardware decision from the software decision

This is the structural mistake that costs the most money over a contract lifecycle. When your tender bundles sensors and analytics into a single proprietary stack, every future negotiation happens with zero leverage on your side of the table. A device-independent platform lets you run stereo cameras at high-traffic mall entrances, time-of-flight sensors in privacy-sensitive public buildings, and cheaper devices at back-of-house doors — all reporting into one analytics layer with one data model.

Write it as a mandatory requirement: "The proposed analytics platform must support sensors from multiple manufacturers, and the buyer must be able to add or replace sensor hardware from third parties without loss of historical data continuity." Then score vendors on how many sensor brands they support in production today — not on a roadmap slide.

Integration requirements: name the systems, name the method

Footfall data is worth little in isolation. Its value appears when it sits next to transactions, staffing rosters and marketing calendars — conversion rate, revenue per visitor, labour-to-traffic alignment. Your tender should name the specific systems the winning platform must connect to: your POS vendor, your BI environment (Power BI, Tableau, Qlik), your ERP and CRM. Ask each bidder to describe an existing, live integration with each named system, including whether it is API-based, file-based or via an integration hub. Vemco handles this through VemFusion, which connects counting data with POS, BI, ERP and CRM systems — the point for your evaluation is that the integration layer should already exist, not be built for you at your cost during implementation.

Include a data ownership clause: all raw and aggregated count data belongs to the buyer, exportable in a standard format at any time, including at contract exit. Vendors rarely refuse this when asked; they simply do not offer it unprompted.

Hosting, privacy and public sector specifics

For public sector buyers and regulated enterprises, hosting flexibility is often the deciding factor. Require both options in the tender — vendor-hosted cloud and private cloud deployment — and score the vendor's ability to deliver either without functional differences. Ask explicitly where data is processed and stored geographically, and how GDPR compliance is achieved at sensor level: does the device process anonymously on the edge, or does video leave the premises? For libraries, transport hubs and municipal buildings, edge-based anonymous counting is usually the only politically survivable answer.

Request certifications, SLAs and a documented support model as mandatory attachments, and ask for reference customers of comparable size and sector — then actually call them. A vendor with 2000+ customers and partners across 95+ countries should be able to produce a reference in your industry and your region without hesitation.

The requirement almost everyone forgets

Here is what implementers know and procurement documents rarely capture: counting accuracy is decided during commissioning, not manufacturing. The same sensor can deliver 99% in one entrance and 88% in the next if nobody accounted for a glass façade throwing afternoon reflections, a promotional table pushed under the counting zone, or staff cutting through a customer entrance forty times a day. Your tender must therefore specify who owns commissioning quality: require a documented calibration and validation report per entrance before acceptance, and define acceptance as measured accuracy at each entrance — not installation completed. Also require an exclusion mechanism for staff traffic, because unfiltered staff movement quietly inflates footfall and deflates your conversion rates, and you will make bad staffing decisions on top of bad data for years without noticing.

A scoring model that actually differentiates bidders

If your evaluation weights price at 60%, you will get the cheapest sensors and the thinnest support. A structure that produces better long-term outcomes:

• Contractual accuracy and audit terms (20%) — the floor, the method, the remedy.
• Platform openness (20%) — sensor independence, data export, named live integrations.
• Scalability and track record (15%) — proven single-store-to-enterprise deployments, reference calls completed.
• Total cost of ownership over 5 years (25%) — licences, hosting, support, sensor replacement, integration fees. Not unit price.
• Implementation and support quality (20%) — commissioning methodology, per-entrance acceptance criteria, SLA response times.

One final procedural tip for consultants and system integrators drafting on a client's behalf: include a proof-of-concept stage for the two highest-scoring bidders — two or three entrances, thirty days, accuracy validated against manual counts. It costs a few weeks and removes nearly all delivery risk from the award decision. Vendors confident in their platform accept it readily; the others self-select out, which is precisely the point.

Preparing a people counting tender and want to pressure-test your requirements against what a platform provider actually sees in RFPs across 95+ countries? Contact Vemco Group for a requirements review — including sample accuracy clauses, integration specifications and acceptance criteria you can adapt directly into your tender documents.