retail analytics

Dwell time analytics that show where attention actually goes

In short

Ayonix dwell analysis measures how long people remain in defined areas of a camera view. Reported as distributions rather than single averages, dwell shows which displays, counters and waiting areas hold attention, how that changes by hour, and how a layout change altered behaviour.

Request a pilot Send a sample video

Conceptual illustration of dwell and waiting behaviour in a terminal. AI-generated conceptual image, not a customer deployment.

Scope

What this analytic does, and what it does not

What it detects

Time each tracked person spends inside a defined zone
Dwell distribution per zone, not only the mean
Number of distinct visits to a zone per interval
Change in dwell between comparable periods

What it does not guarantee

Why a person stayed — dwell measures duration, not interest or intent
Accurate dwell where the person is occluded for long periods
Individual-level tracking between zones or across visits
A causal link between dwell and sales

The problem

What customers are actually dealing with

These are the situations that lead teams to look at this analytic in the first place.

A merchandising change is made and nobody can say whether it altered behaviour, only whether sales moved for other reasons too.
Average dwell hides the distinction between many brief passes and a few long stops, which are very different situations.
Waiting areas are sized from assumption rather than measured waiting behaviour.
Teams debate which displays work using anecdote because there is no consistent measurement.

How it works

Detection workflow

Every event carries the rule that produced it, so an operator can see why they were alerted.

1

Ingest and detect

Zone cameras are read and people in frame are detected and tracked.
2

Accumulate dwell

Each track accumulates time inside each zone polygon it enters.
3

Close visits

A visit is closed when the track leaves the zone beyond the brief-exit tolerance or the track ends.
4

Aggregate

Visits are aggregated into distributions, medians and percentiles per interval.
5

Publish

Zone dwell series are written to the dashboard and made available for export.

Configuration options

Zone polygons per camera, aligned to displays, counters or waiting areas
Brief-exit tolerance before a visit is considered closed
Minimum dwell before a visit is recorded, filtering pass-through traffic
Reporting intervals and the percentiles reported alongside the median
Comparison baselines for before-and-after measurement
Trading-hours schedule so closed periods are excluded

Alert quality

How irrelevant and duplicate alerts are reduced

No video analytic eliminates false alerts. These are the mechanisms that reduce them, and the residual rate is measured on your own cameras.

A minimum dwell threshold separates genuine stops from people simply walking through the zone.
Brief-exit tolerance prevents one visit being split into several by a momentary step away.
Reporting distributions rather than a single average prevents a handful of very long dwells from distorting the picture.
Excluding closed hours prevents cleaning and restocking activity from contaminating the series.
Occlusion breaks tracks and truncates dwell; the effect is quantified per zone during the pilot.

Prerequisites

Camera requirements and environmental limits

Camera suitability decides more of the outcome than any software setting. These are assessed per camera before commitment.

Camera requirements

A view that keeps the subject visible for the whole expected dwell, not just at the zone boundary
An elevated angle that limits occlusion between people standing close together
Person height of at least 8–10% of frame across the zone
Stable, consistent lighting across trading hours
A zone that fits comfortably within the field of view with margin at the edges

Environmental limitations

Dwell is systematically under-reported where fixtures, pillars or other shoppers occlude the subject.
In crowded zones, tracks merge and dwell measurement degrades.
Dwell alone does not indicate interest — a queue produces long dwell for reasons unrelated to the display beside it.
Comparison between zones is only fair when camera views are comparable.

Delivery

Alerts, evidence and where they land

Event and evidence fields

Zone name and camera
Interval start and end
Visit count in the interval
Median dwell and selected percentiles
Mean dwell
Track break rate for the interval

VMS integration

Dwell analytics is a reporting series rather than an operator alarm. Where the same zone also needs a security threshold, the dwell alerting configuration is used alongside it and delivers events to the VMS.

See compatibility states →

Deployment options

Edge processing per store or terminal so only aggregate dwell series leave the site
On-premise processing where zone cameras already aggregate locally
Hybrid deployment for estate-wide comparison of comparable zones

Compare architectures →

Measurement

Dashboard metrics and pilot acceptance criteria

Acceptance thresholds are agreed with you before the pilot starts. This site publishes no benchmark figures, because they do not transfer between sites.

Dashboard metrics

Median and percentile dwell per zone per interval
Visit count per zone
Dwell distribution shape, shown as a histogram
Before-and-after comparison against a baseline period
Zone ranking by median dwell
Track break rate as a data-quality indicator

Measurable pilot criteria

Dwell accuracy against stopwatch ground truth for a set of scripted stays of varied length
Track break rate per zone and its measured effect on reported dwell
Stability of the median across two comparable days
Behaviour of the minimum-dwell filter against scripted pass-throughs
Data completeness across trading hours

Governance

Privacy and governance

Dwell is measured per anonymous track within a single camera session and aggregated before reporting.
No individual is identified and no profile is built across visits.
Aggregate dwell series can be retained without retaining imagery.
Where zones cover staff positions, employee monitoring obligations may apply.

Frequently asked questions

Does long dwell mean customers are interested?

Not necessarily. Long dwell at a counter may mean a slow service process rather than engagement, and a queue beside a display produces dwell unrelated to it. Dwell is a measurement of duration; interpreting it requires knowing what else is happening in that space.

Why report distributions instead of an average?

Because an average hides the difference between a hundred three-second passes and ten thirty-second stops. The median and upper percentiles usually describe behaviour more usefully, and the histogram makes the pattern visible.

Can you follow one person between zones?

Not on this site. Tracks are scoped to a camera session and aggregated. Cross-camera re-identification is not part of the published Ayonix video analytics capability set and is not offered here.

Keep reading

A controlled pilot establishes what this analytic actually does in your environment, against acceptance criteria we agree before it starts. Send a sample video first if you would rather see an assessment before committing to a pilot.