What we refuse to know

Many event-analytics systems learn something about individual devices or people in order to measure a crowd. Roam Motion was built around a different idea: measure the space without creating a record of the people in it.

This page explains the difference, because it is the whole point of the company.

What "measuring an activation" usually means

Event and experiential measurement commonly relies on one of several approaches. Some create persistent device-level records. Some retain imagery. Some infer attendance from third-party location datasets. They can all produce useful analytics — but they make very different privacy tradeoffs.

Listening for phones. Small sensors listen for Wi-Fi or Bluetooth signals emitted by nearby devices. Depending on the system, a device-derived identifier is captured or created and used to distinguish one device from another. That can be used to estimate how many devices were nearby, how long they stayed, whether the same device returned, and — in some systems — whether it later appeared near another sensor.

Camera-based analytics. Some systems record the space and analyze the resulting footage. That footage may be retained, at least temporarily. Some systems also infer demographic characteristics from faces or re-identify the same person across cameras to reconstruct movement.

Modeled location data. Aggregated GPS or app-location data is bought from data brokers and modeled to estimate attendance in a geofence. Nobody is measured at the venue at all; the number is a statistical estimate derived from a sample of phones.

Each of these produces a report. Some also produce something else: a persistent device-level or person-level record that allows the same subject to be distinguished over time.

Why Roam Motion refuses

The people in the space did not consent. Attendees came for the event. A sponsor's guest, a family walking past a fan zone, a child at a festival — none of them opted in to having their phone logged or their face analyzed. We do not think a sticker on a fence changes that.

We do not want the argument in the first place. In 2021, the Dutch data-protection authority fined the city of Enschede €600,000 over a Wi-Fi-based pedestrian-counting system. After years of litigation, the fine was ultimately overturned in July 2026 after the courts found that the regulator had not adequately established the alleged unlawful processing in its enforcement decision.

Our takeaway is not that one side was right. It is that counting foot traffic should not require a multi-year legal argument over whether a device identifier can be tied to a person. Roam Motion is designed so that question never needs to arise.

A sample is not the same as a direct count. Phone-based counting can only observe devices that are detectable by the system. Modern phones randomize hardware addresses and use other privacy protections that make passive device observation less deterministic.

Vendors using this approach may describe the result as a "statistically significant sample." That can be useful for behavioral analysis, but it is not a census of the people who actually passed the footprint. A sample-based conversion rate is an estimate from the devices the system could observe. Roam Motion's capture rate is based on directly measured pass-by traffic and directly measured entries.

Our product does not need it. Roam Motion measures a physical footprint: how many people passed it, how many came in, how long they stayed, and where inside it they stood. None of that requires knowing who anyone is, whether they came back, or where they went afterward.

What we measure instead

• Pass-by traffic on the aisle, walkway, or concourse in front of the footprint. This gives capture rate a directly measured denominator instead of an inferred audience estimate.
• Entries and exits across the footprint boundary.
• Dwell time inside the footprint and inside zones defined within it.
• Capture rate: entries divided by pass-bys, by hour and by day.
• Where people stood, shown as a heat map over the floor plan.
• Behavioral segments based only on movement: passers, glancers, engagers, deep-dwellers.

Every number is an aggregate. Every number can be footnoted to the sensor and the method that produced it.

What we refuse to measure

• Recorded video. Video is never recorded to disk, retained, or transmitted. Images exist only transiently in the edge-processing pipeline long enough to perform measurement and are then discarded.
• Faces, or anything derived from a face: no age, gender, mood, or ethnicity estimates, ever.
• Persistent identifiers. No MAC addresses, advertising IDs, device fingerprints, biometric templates, or other identifiers that can follow someone beyond the immediate measurement.
• Paths. We will show you a heat map of where people stood. We will not draw the line one person walked, because that line is a record of a person.
• Repeat-visit or cross-location claims. We cannot tell you whether a visitor came back, or went to your store afterward, because we do not know who anyone was.

The edge processor may maintain a temporary anonymous track in memory long enough to determine that an object crossed a boundary or remained in a zone. That temporary state is discarded. It is never stored, transmitted, or reported.

The test we apply

Before any new metric or visualization ships, it has to pass one question: does this require us to preserve a record of an individual person?

If the answer is yes, we do not build it.

A heat map describes the space; it passes. An individual trajectory describes a person; it fails.

We build the left side. Never the right.

For your legal and compliance team

The short version you can put in an email:

Roam Motion sensors are edge-processing devices that output aggregate measurement data only. No video or imagery is retained or transmitted. No device identifiers are collected. No biometric or demographic inference is performed. Individual trajectories are not stored or reported. The data produced consists of counts, durations, and floor-plan density aggregates for a defined physical footprint.