Sightora

Platform

What We Read From a Camera View.

Not “AI watches your store”. A zone you draw, a measurement you chose, a threshold you set, and a person who decides what to do. Here is each of those, on three real retail scenes — with the boundary stated as plainly as the capability.

Read this page knowing

  • Every number here is sample data. None of it is a customer result, a benchmark, or a measurement of the photograph it sits on.
  • Nearly everything is still pilot. Each measurement names the capability behind it, and the register says where that capability stands.
  • No face, no name, no intent. Not on this page, and not in the product. Those are recorded as prohibited claims, not as features awaiting a release.

Frame by frame

Three views, and what each one is configured to report

A camera view becomes useful once someone draws the areas that matter in it and says what each area should measure. These are three such views, with the zones numbered and the measurements listed beside them.

  • Counting — People passing through an area you marked.
  • Dwell and queues — How long people stay in an area.
  • Where a rule fires — The threshold you set, and the signal it raises.
  • Restricted and evidence — An area with access rules, and the clip kept for review.

A checkout at 18:42

Five shoppers holding bags wait in an orderly line at a store checkout while a cashier serves a customer

Illustrative zones — sample values — Zones and values are an example configuration, not an analysis of this photograph.

What the picture shows: Five shoppers holding bags wait in a line at a store checkout while one cashier serves.

Checkout queue

Needs a pilot to confirm

The floor space a line forms in. Measured as an area, so the count does not depend on people standing in a tidy row.

People in the area
5
Length held for
6 min
Threshold configured for this site
4 people for 5 min

Measured under: Queue and wait time, Real-time alerts.

Till line

Needs a pilot to confirm

Counts staffed positions, so a long queue can be read against how many tills are actually open.

Positions open
1 of 3
Signal raised
Open a second till

Measured under: Queue and wait time, Real-time alerts.

Approach path

Needs a pilot to confirm

The route into the checkout area. Separating it from the queue stops someone walking past being counted as waiting.

Passed through, last hour
214
Entered the queue
38%

Measured under: Foot traffic and occupancy.

What it reads

  • How many people are standing in the area you marked as the queue.
  • How long the line has been that length, which is what turns a busy moment into a signal worth sending.
  • How many tills are open, counted as staffed positions rather than identified staff.

What it does not read

  • Who any of these people are. No face is matched, stored or compared.
  • Whether anyone is a repeat visitor. Re-identification across visits is outside the scope agreed for a pilot.
  • What anyone intends to do. No behaviour here is scored, ranked or flagged as suspicious.

A display wall on a weekday afternoon

A shopper examines a leather loafer in front of a wall of shoes on wooden shelves

Illustrative zones — sample values — Zones and values are an example configuration, not an analysis of this photograph.

What the picture shows: A shopper examines a leather loafer in front of a wall of shoes on wooden shelves.

Display front

Needs a pilot to confirm

The standing space in front of the shelving. Dwell in this area is the measurement; the shelf itself is the reference, not the subject.

Median dwell
48 sec
Stops, last 4 hours
63
Longest single dwell
4 min 10 sec

Measured under: Zone dwell and movement patterns.

Aisle approach

Needs a pilot to confirm

The walkway past the display, so a stop rate can be calculated against people who had the chance to stop.

Walked past
402
Stopped
16%

Measured under: Foot traffic and occupancy, Zone dwell and movement patterns.

What it reads

  • How long people stay in front of a display, and how that changes when you move the display.
  • How many people walk past the aisle versus stop at it.
  • Which of two displays holds attention longer, measured the same way on both.

What it does not read

  • Which product anyone picked up. Item-level interaction is not a registered capability and is not inferred from a frame.
  • Age, gender or any demographic estimate. None is produced, and none is planned.
  • Anything about this person as an individual. The measurement is of an area over time, not of a shopper.

A stockroom bay you have marked restricted

An employee pushes a stock trolley of boxes along orderly shelving in a retail stockroom

Illustrative zones — sample values — Zones and values are an example configuration, not an analysis of this photograph.

What the picture shows: An employee pushes a stock trolley of boxes along orderly shelving in a retail stockroom.

Restricted bay

Planned, not orderable

A bay you mark restricted - high-value stock, or the way through to the cash office - with the hours and the two-person rule that apply to it recorded alongside.

Entries today
7
Outside configured hours
1
Clip retained for review
Yes

Measured under: Restricted-area access events, Video-backed incident review.

Stockroom aisle

Planned, not orderable

Movement through the stockroom, used to tell a normal replenishment round from an out-of-hours visit.

Replenishment passes
22
Out-of-hours passes
0

Measured under: Zone dwell and movement patterns, Restricted-area access events.

What it reads

  • That the door to an area you marked as restricted was opened, and at what time.
  • Whether that time falls inside the hours you configured for that door.
  • Whether one person entered where your own procedure asks for two.
  • How long the door stood open, which is usually a propped door rather than anything else.

What it does not read

  • Who opened the door. The event records a role from your rota, not a recognised face.
  • Whether anything was taken. There is no inventory inference and no theft determination anywhere in this product.
  • Intent, of any kind. An access event is a fact about a door and a clock, and it is addressed to a human to interpret.

Across a whole floor

Dwell, as a shape and as numbers

One zone answers a question about one place. Several zones over a week answer a different question: where attention actually goes, and what changes when you move a table.

Shoppers browse clothing rails and tables in a spacious fashion store while an associate folds garments
Dwelllow to high
Sample floor, one Tuesday 11:00-19:00

Illustrative heatmap — Sample dwell data over an example floor, not an analysis of this photograph.

  • Front table

    Share of dwell
    31%
    Median dwell
    1 min 20 sec
    Change
    +6 pts on last week
  • Centre rails

    Share of dwell
    24%
    Median dwell
    58 sec
    Change
    Level
  • Fitting room approach

    Share of dwell
    18%
    Median dwell
    2 min 05 sec
    Change
    +2 pts
  • Left wall

    Share of dwell
    15%
    Median dwell
    41 sec
    Change
    -3 pts
  • Back corner

    Share of dwell
    12%
    Median dwell
    26 sec
    Change
    -5 pts
Share of dwell across five zones on one floor, totalling 100%. Zone geometry is drawn once per camera during setup; the measurement is of an area over a time window, not of any shopper in it.
A heatmap is the easiest picture in this industry to fake, so it is worth being precise about what one is. It is dwell time per zone, summed over a window, drawn as intensity. It is not a live view, it is not a record of individuals moving, and it is not produced by looking at a photograph.
What makes it useful is the comparison, not the colour: the same zones measured the same way next week, after you changed something. That is also why the numbers are printed beside the picture rather than left inside it.

Back of house

Restricted areas: the cash office and the stockroom

The question every retailer asks about a back room is who went in and when. This is the only shape in which we will answer it.

An access event is a door, a clock and a rule you wrote. It carries the role on the rota, never a name and never a face, because identifying a person from video is a prohibited claim here rather than a feature on a roadmap.
It also carries no conclusion. A door opened at 21:54 is a fact; whether that matters is a judgement for a duty manager with the context, and the clip is attached so that judgement is made on evidence rather than on a feeling.

Restricted-area access events

Planned, not orderable

Record that a door you marked restricted was opened, against hours and rules you configure. The event carries a time, a zone and a role from your rota - never an identity, and never a determination that anything was taken.

Planned and not currently orderable.

  • Cash office

    Awaiting review

    Opened outside configured hours (07:00-21:00)

    When
    Thu 21:54
    Role on rota
    Duty manager
    Evidence
    Clip attached
  • Cash office

    Under review

    Single-person entry where the site procedure asks for two

    When
    Thu 14:08
    Role on rota
    Supervisor
    Evidence
    Clip attached
  • Stockroom rear door

    Resolved

    Held open longer than 3 minutes

    When
    Thu 11:36
    Role on rota
    Stock assistant
    Evidence
    No clip

Illustrative access log — Sample events. Roles come from your rota; no person is identified and nothing is inferred about what happened.

The rules a site configures

Hours.
The hours each restricted door may be opened, per site and per day of the week.
Two-person entry.
Where your own cash-handling procedure asks for two people, a single entry can be recorded as an exception.
Door held open.
A duration after which a propped door is worth someone looking at.

Status

What every number on this page rests on

Each measurement above belongs to one of these. None of them is presented as a finished, demonstrated production capability, because no evidence record has been signed.

Zone dwell and movement patterns

Needs a pilot to confirm

Activity and dwell within zones you define. Zone geometry, camera calibration and privacy scope are agreed before measurement begins.

Confirmed for your sites during a scoped pilot before it informs decisions.

Queue and wait time

Needs a pilot to confirm

Queue events against a defined threshold, where camera geometry allows the queue to be seen. Start and end logic is agreed with you in writing.

Confirmed for your sites during a scoped pilot before it informs decisions.

Foot traffic and occupancy

Needs a pilot to confirm

Visit and occupancy counting where the camera view supports it. Accuracy is benchmarked against a manually reviewed sample before any number is used for decisions.

Confirmed for your sites during a scoped pilot before it informs decisions.

Real-time alerts

Needs a pilot to confirm

Event rules can raise a timestamped alert to an authorised reviewer. End-to-end timing is measured for your site during a pilot.

Confirmed for your sites during a scoped pilot before it informs decisions.

Video-backed incident review

Needs a pilot to confirm

Authorised teams review a defined event with its timestamp, camera and context, record the action taken, and keep an audit trail.

Confirmed for your sites during a scoped pilot before it informs decisions.

Restricted-area access events

Planned, not orderable

Record that a door you marked restricted was opened, against hours and rules you configure. The event carries a time, a zone and a role from your rota - never an identity, and never a determination that anything was taken.

Planned and not currently orderable.

Each capability is confirmed for your own sites before it informs a decision. How validation works.

Bring us one camera view.

The fastest way past a page of sample data is to look at a real view from one of your stores and say what you would want measured in it.

Review My Camera Environment