People counting for stores. You know how many receipts you issued, not how many people left without buying.
We count customers on the cameras you already have and compare them with receipts, hour by hour, in every store. You see the real conversion rate, the zones customers avoid and the hours when the queue costs you sales. No new hardware. We only recognize the employees you register; customers remain silhouettes, with no name and no profile.
Centru store · cam 1 · overhead viewlive simulation
walk-ins today0
bought0
no purchase0
in store0
customerboughtno purchasestaff, not counted
Centru store · 18:00 · 146 visitors · conversion 31%Queue 7 at checkout 2 · alert sent in 10 sPromo zone B · 9 out of 100 customers get thereConversion today 41% · yesterday 39%Evening report · sent at 20:30Aisle blocked 9 min · zone D · 13:40Chain · 8 stores measured the same way
What Vizor sees in a store
Three things the cash register doesn’t know.
The cash register counts receipts. Vizor counts the people who came in, where they went and how long they waited. Decisions on shifts, layout and checkouts come from the gap between visitors and receipts.
0
Conversion by hour
How many come in, how many reach the checkout.
Visitors against receipts, for every hour and every store. The peak hour brings the most people and, almost always, the weakest conversion.
Zones and promotions
Where customers walk and what they skip.
Visit rate by section and by promo zone. A section that doesn’t sell has two possible causes: nobody wants the product, or nobody walks by. The map tells you which.
queue 7 · checkout 2 · open checkout 3
Checkout queue
How many wait, and for how long.
Once the threshold is crossed, the shift supervisor gets an alert within ten seconds. In a supermarket, five minutes at the checkout means abandoned baskets and customers who choose another store next time.
Theft and shrinkage
We see when goods leave without passing the checkout.
The same cameras that count customers also watch for three theft patterns. When one occurs, an alert goes out — with the exact time and the exact camera. Checking it takes ten seconds instead of an hour of searching through recordings.
time · camera
Exit that bypasses the checkouts
A person goes from an aisle straight to the exit without stopping at the checkouts. The alert includes the camera and the second.
18:42:07 · camera 3 · exit without checkout
expensive goods
Unusually long dwell
Someone lingers in the aisles with expensive goods, outside the usual pattern for that hour. An employee stepping into the area usually settles it.
16:15:40 · camera 2 · 6 min at cosmetics
gesture, not product
Item hidden in a pocket or bag
The system watches the gesture, not the product: a hand moving from the shelf to a pocket, bag or coat instead of the basket. It is active only in the zones you choose — cosmetics, alcohol, small expensive items. If the person then leaves without passing the checkouts, the alert gets top priority.
11:03:22 · camera 5 · reach toward a bag
Store dashboard
All your stores. One picture.
The dashboard layout from our real installation. Switch stores and go through Dashboard, Alerts, Analytics and Reports; the dashboard cycles through them on its own until you click. The numbers are illustrative; in a live installation they come from the store’s cameras and the POS.
app.vizor.md · stores · dashboard
V
Vizor
On the phone
The same numbers, for the shift supervisor and the store manager.
The mobile app is for whoever has to decide in the next minute: open a checkout, move a person, clear an aisle. Not tables, but the three things that matter right now.
The alert lands on the lock screen, with the store, the camera and the exact minute.
Visitors, people who left without buying and time at the checkout, for the current day, in the selected store.
The evening report takes a minute to read, in the app or in Telegram.
The day’s numbers reach your phone, without customer names.
Store reports
This is the report that arrives every evening.
Not presentation charts: the document the store manager receives in Telegram at 20:30, with the day’s numbers, the zones and two masked frames. Same format as in the real installation; the numbers here are illustrative.
Daily report
Centru storeThursday, September 4, 2026 generated automatically at 20:30 · 4 cameras
Visitors1,240+4% vs. last Thursday
Receipts512conversion 41%
Left without buying72859% · 312 of them between 17:00 and 19:00
Peak hours18:00146 visitors · conversion 31%
Time at checkout2:50max 5:40 at 18:12 · queue 7
First · last movement07:58 · 21:47open 08:00–22:00
By hour · visitors and receipts
Hour
Visitors
Receipts
Conversion
Max queue
08
26 / 14
54%
1
09
44 / 22
50%
2
10
62 / 29
47%
2
11
76 / 34
45%
3
12
92 / 40
43%
3
13
105 / 47
45%
4
14
101 / 46
46%
3
15
88 / 39
44%
3
16
103 / 43
42%
4
17
126 / 42
33%
6 · 17:40
18
146 / 45
31%
7 · 18:12
19
114 / 47
41%
4
20
82 / 37
45%
2
21
55 / 27
49%
1
Zones · of 100 customers entering, how many pass through the zone
Zone
Visit rate
Average time
Vs. last week
Bakery
1:10
+2 pp
Dairy
1:35
0
Meat and deli
2:05
−1 pp
Beverages
0:50
−3 pp
Household
0:40
0
Promo B · far back
0:15
9 of 100
Moments of the day · masked frames
CAM 2 · 2026-09-04 Peak of the day · 18:12 · queue of 7 at checkout 2faces maskedCAM 4 · 2026-09-04 Promo zone B · 16:00 · no customersfaces masked
For tomorrow
1Open checkout 3 between 17:00 and 19:00. Yesterday, 312 people left without buying in this window, and the queue crossed the threshold twice.
2Move promo B next to the main path, between dairy and bakery. Measure the effect over seven days, before and after.
3Clear aisle D of pallets before 12:00. Yesterday it was blocked for 9 minutes at 13:40.
Queue of 7 at checkout 2Centru store · cam 2 · 18:12. Two of four checkouts open, wait 5:40. Open checkout 3.18:12
Queue back below threshold18:31 · wait 2:10. The previous alert closed on its own.18:31
Daily report · September 4Visitors 1,240 · receipts 512 (41%) · no purchase 728 Peak 18:00, 146 visitors · max queue 7 at 18:12 Promo B: 9 of 100 · aisle D blocked 9 min Tomorrow: checkout 3 at 17–19, promo B by the main path.
PDF
report_centru-store_2026-09-04.pdf1 page · 468 KB
18:1216:00 · promo B
20:30
Message
18:12
Thursday, September 4
Vizor · Centru storeQueue of 7 at checkout 2 · cam 2. Two of four checkouts open, wait 5:40. Open checkout 3.
now
Vizor · Botanica storeAisle blocked 9 min · zone D · pallets in the main aisle.
Hours with the queue over threshold6144 of them between 17:00 and 19:00
Conversion ranking · same method in every store
Store
Conversion
Visitors
Δ wk
Queue > threshold
Centru
8,680
+2 pp
6 h
Ciocana
7,420
+1 pp
5 h
Râșcani
6,950
0
7 h
Botanica
7,810
+3 pp
6 h
Buiucani
6,240
−1 pp
8 h
Telecentru
5,980
0
9 h
Poșta Veche
7,560
−2 pp
10 h
Sculeni
7,760
−3 pp
10 h
Common dead zones · under 15% of customers in all stores
Zone
Visit rate · chain average
Stores affected
Promo B · far back
8 of 8
Household
6 of 8
Frozen · corner
5 of 8
Three recommendations
1Sculeni and Poșta Veche: the same traffic as Centru, with 20 points less conversion. Open a third checkout 17–19 and review the shift schedule.
2Move promo B next to the main path in all stores: the zone is dead in 8 of 8. Measure the effect for a week.
3Centru: conversion rose 2 points after the third checkout. The same decision can be replicated in 5 stores.
Vizor · all stores measured the same way · aggregated numberspage 1 / 2
Analysis examples · stores
Three scenarios, three decisions.
Our first real installations, in production, are in pharmacies in Moldova. For stores and supermarkets we don’t yet have a client we can name publicly, so the scenarios below are modeled on the patterns measured there and on the retail analytics literature. They are marked as such, and we measure your numbers together in the pilot.
Promoted products were selling poorly, even though the price beat the competition. The supplier wanted an explanation for the promotion budget.
What it showed
The promo zone was at the far back: 9 out of 100 customers got there. The main path ran 12 meters away.
Decision
Zone moved next to the main path. Same products, same price, same campaign. Cost: zero.
×2.3visits to the promo zone; promo sales +18% in two weeks
of 100 customers, how many reach the zonebefore / after
Modeled scenarioSupermarket · 4 checkouts
18:00 brought the most people and the fewest receipts.
Problem
Evening sales weren’t growing, even though evening traffic was the highest of the day. Reviews complaining about the queue, baskets left at the checkout.
What it showed
Two of four checkouts open between 17:00 and 19:00, a queue of more than six people on three days out of five. Conversion dropped by 11 points exactly then.
Decision
Third checkout opened 17–19, second shift moved by an hour. Zero new hires.
+6 ppconversion at peak hours, after four weeks
conversion by hour · before / after%
Modeled scenarioStore chain · 8 locations
Two stores with identical traffic. Sales 35% apart.
Problem
The difference was blamed on the neighborhood. The marketing budget went to the weak store, with no effect.
What it showed
Same traffic, conversion 28% vs. 44%. The difference was in-store execution, not location.
Decision
A weekly chain-wide ranking, training at the weak store, shifts adjusted to actual traffic hours.
+9 ppconversion at the weak store in six weeks, with no marketing budget
conversion ranking · 8 stores%
Our reference installation
The first location we measure daily, with a parallel manual count and an automatic report every evening. The method is the same for a store; only the numbers differ.
The entrance and checkouts are almost always covered already. For zones and promotions we add one or two ordinary cameras if needed.
zero investment
02
Automated analysis
The video stream is analyzed continuously. Staff are told apart from customers by behavior and position; we draw the aisles and zones together with you.
own method
03
Numbers, not identities
Images are processed on the Vizor server; the report gets numbers: visitors, zones, waiting times. Customers are not identified.
data minimization
04
Link to the POS
Hourly receipts are matched with visitors, which gives you the real conversion rate. Daily export or integration, whichever is easier for you.
conversion
Privacy in the store
We don’t identify customers. We count silhouettes.
The system tracks silhouettes from frame to frame and turns them into numbers: how many came in, where they went, how long they waited. Images are processed on the Vizor server. We provide the privacy notice for customers and the compliance documentation. Individual monitoring applies only to employees you explicitly register, with prior notice to them and the documented legal basis required by Law No. 195/2024.
Pilot for store chains
Three stores. Three months. Numbers checked together.
One large store, one small one and one neighborhood store, so the result says something about the chain, not about a single location. In week three we count manually in parallel, at hours you choose. You can stop the pilot at any time, with no penalty.
We don’t ask you for
New cameras
We work on your existing cameras. Where coverage falls short, we fill the gap.
We don’t ask you for
An IT project
Installation takes one day per store and doesn’t stop sales. No integration required to start.
We don’t ask you for
Long contract
Three months. Stop whenever you want. The comparison report for the three stores is yours to keep.
We don’t ask you for
Customer data
Only the video stream and, for conversion, the number of receipts per hour. Nothing about what anyone bought.
Calculator · what a conversion increase is worth
In an ordinary supermarket, the conversion gap between 14:00 and 18:00 is more than ten points. Here we calculate conservatively, with the figure you choose. An estimate, not a promise.
Extra sales per month648,000 MDL
extra receipts / month3,600
per year7,776,000 MDL
Calculation: visitors × days × conversion points × average basket × stores.
Usually not. The entrance and checkouts are almost always covered by the existing cameras. For zones and promotions we add one or two ordinary cameras if needed.
How is conversion calculated?
Receipts from the POS divided by the counted visitors, for every hour and every store. We take the receipts from a daily export or through an integration, whichever is easier for you. Staff aren’t included in the count.
Does it work with several entrances and several checkouts?
Yes. Each entrance has its own counting line, each checkout its own queue zone. The numbers are added up per store and shown separately in the dashboard.
Does the system tell staff apart from customers?
Yes, by behavior and position: who stands at the checkout or behind the counter, who comes in and goes out through the door. Employees aren’t counted as visitors. If you register them in the system, they are recognized by face and can be reported on individually.
Does it use facial recognition?
Yes, for the employees you register. Customers aren’t registered, so they remain counted silhouettes, with no identity. Camera images are processed on the Vizor server. The privacy notices for customers and employees and the compliance documentation are prepared together, as part of the pilot.
How long does installation take?
One day per store, without stopping sales. We draw the zones and queue thresholds together on the first day; the first report arrives that same evening.
How accurate are the numbers?
In the third week of the pilot, we count manually in parallel, at hours you choose, and check the numbers together. We promise numbers you can verify, not brochure percentages.
How do we start?
With a paid three-month pilot in three stores, which you can stop at any time, with no penalty. Before that, 30 minutes and a look at the cameras in one store.
Contact
Show us a store. We’ll tell you what we can measure.
We don’t ask for an hour-long meeting. You pick a store, we look at the existing cameras, and we reply within 24 hours, in writing, with what can be measured there.
What happens next
Three steps, no pressure.
1We reply within 24 hours, in writing, with what can be measured at your location.
2Thirty minutes on a single location: together we look at what your existing cameras cover.
3If it makes sense for you, we propose a pilot on three locations. You can stop it at any time, with no penalty.