People counting
Six ways retailers lose trust in their people counters. What is different at Pygmalios.
Disputed counts, staff in the conversion rate, reports built in Excel, sensors nobody noticed had stopped, hidden fees, a number at the door that explains nothing. This is what retailers report about their counters, and how each point is handled for Pygmalios clients.
- Accuracy above 99%, verified by manual counts
- Outages seen within minutes
- One price, data yours to export
- Pilot on one or two stores first
Trusted by leading European retailers
Start here
Three questions to ask before you trust a number from your counter.
- 01
Can you explain a disputed hour after the fact?
When a store manager says Tuesday at 3 pm was not 40 visitors, can you show what the sensor saw minute by minute, whether data was missing, and whether a setting changed that week? Pygmalios clients can.
- 02
Can you say who enters your conversion rate?
Staff, families, a courier, a customer going back to the car: each one changes the denominator. Pygmalios counts buying units, tags staff out, and keeps the raw and the adjusted number side by side.
- 03
Can you tell low footfall from a data outage?
A sensor that stopped counting looks like a quiet day. At Pygmalios an outage is flagged within minutes, the gap is estimated and marked as an estimate, and service starts before the store calls.
The six complaints
What retailers report, and what happens at Pygmalios instead.
The six families come from public retailer reviews, forum threads and supplier case studies about people counting systems, collected through September 2026. They are paraphrased. Under each one: how the point is handled for Pygmalios clients, and the evidence behind it.
- 01
Trust in the numbers
What retailers report
Disputed counts, readings that do not match, no way to check a suspicious hour. In one published supplier case the retailer still distrusted its data after the faulty equipment had been replaced.
At Pygmalios
Every disputed hour can be explained after the fact from minute-level counts, the outage log and the configuration history. Every change to a counting line or to opening hours is recorded with a date. A data gap shows as a gap.
Evidence
- Accuracy above 99%, verified repeatedly by manual counts at client sites
- O2, Slovak Telekom and Humanic have run for more than three years without a sensor swap or a recalibration
- 02
What the KPI means
What retailers report
Staff walking through the door drag conversion down, a family counts as four customers, a smoker inflates footfall, and head office compares stores with completely different visitor mixes as if they were the same.
At Pygmalios
Staff are excluded from footfall, tagged by badge or by position. Groups count as buying units. Raw and adjusted numbers stay visible side by side. Stores are compared through filters by type, size and period.
Evidence
- Repeat entries by the same person are deliberately not tracked: that would mean identifying people, and we do not
- Buying units and staff exclusion are documented on the Traffic Analytics page
- 03
Reporting and integration
What retailers report
Conversion calculated in Excel, slow reports, a POS integration stuck for months, an API that fails now and then. A basic question about last week takes an hour.
At Pygmalios
Conversion is a finished number in the product; POS transactions pair with footfall automatically. The first report arrives within days of installation. A regional manager gets an AI Report by email, with weather, holidays and market context and a recommendation, without opening a dashboard. The API is documented; clients pull data into their own BI.
Evidence
- Sample AI Report available on request
- No manual counting or exporting on the client side after setup
- 04
Operations and service
What retailers report
Placement, calibration and repairs become permanent work. Nobody knows who owns an outage. In public reviews of other systems, sensors stopped counting and nobody noticed for weeks.
At Pygmalios
An outage is seen by Pygmalios within minutes and service starts before the store calls. One owner for hardware, software and the network at the sensor. Service times are part of the contract. No calibration after installation; a new setup only when the store is rebuilt. Store staff do nothing. A missing interval is estimated and marked as an estimate.
Evidence
- Monitoring across every installed sensor, around the clock
- Service levels by package, in writing
- 05
Cost against value
What retailers report
Hardware plus subscription plus your own people's time, and at the end: nice charts, we changed nothing. Hidden fees for integrations, users and exports, and a trap at the exit.
At Pygmalios
One price up front covering sensors, installation, software, service, API and users. The data belongs to the client: the full history exports at any time, at no charge, also at the end of the contract. A pilot on one or two stores comes before any network commitment.
Evidence
- A telecom operator changed staff deployment to match footfall and saves hundreds of thousands of euros a year
- Clients have changed opening hours and store layouts on the same data
- 06
Limits of the insight
What retailers report
A count at the door does not say why sales fell or what to change. Retailers ask for dwell time and paths, and get a total.
At Pygmalios
Journey Analytics carries the count past the door: dwell, paths, zones and cross-shopping in the stores of dozens of clients, on the same platform. AI Reports recommend actions. A change to layout, hours or staffing is evaluated before and after, adjusted for season and weather.
Evidence
- Journey Analytics live with dozens of clients
- We give measurements, context and hypotheses to test. Whether a drop in sales had one cause is what the test shows
All six figures are samples on illustrative data, drawn in the shape the product delivers them.
How to start
Two steps before any contract.
- 01
Free data check
Send an export from your current system: hourly counts per store, opening hours, ideally transactions. CSV or Excel is enough, and store names can be replaced by codes; an hourly count carries no personal data. You get a written note on the definitions in use, the missing and suspicious intervals, and whether your dispute can be settled from the data at all. No installation, no visit.
- 02
Pilot on one or two stores
Sensors go up in one or two stores before any network commitment. Where you want proof of accuracy, we compare against an independent reference, a manually annotated sample, under a written protocol with definitions and acceptance criteria agreed up front.
- 03
Then the network
Installation in two to four weeks per wave, first report within days, one price, data yours. The pilot stores keep running.
From Our Knowledge Hub
Recommended Reading
Opportunity to See vs Verified View In-Store
Opportunity to see vs verified view in-store retail media measurement — close the credibility gap, price inventory by zone, and defend spend cross-channel.
Read article →
How to Price In-Store Retail Media Inventory
How to price in-store retail media inventory using audience tiers, zone and daypart data, and verified views to build a rate card buyers respect.
Read article →
GDPR Camera Audience Measurement in Retail: A Guide
GDPR camera based audience measurement retail done right: build compliant-by-design sensor architecture that scales across EU jurisdictions and wins advertiser trust.
Read article →Free data check
Send us the number you do not trust.
Tell us how many stores, which country and what counts at your doors today. We come back within two business days with what we need for the data check.
Request a free data check
99%+
accuracy, verified by manual counts
minutes
to detect a sensor outage
3+ years
without a sensor swap at O2, Slovak Telekom and Humanic
Questions
What buyers ask before they trust a counter.
How accurate is a Pygmalios people counter?
Above 99 percent under the sensor manufacturer's conditions, and we have verified it repeatedly by manual counting at client sites. Accuracy is confirmed per site.
What happens when a store manager disputes a number?
We look at the disputed interval in minute-level counts, the outage log and the configuration history, and show what the sensor saw. Validation establishes what happened. It does not adjust the number until everyone is comfortable.
Do staff count as visitors?
No. Staff are tagged by badge or by position and excluded from footfall. The raw count stays visible next to the adjusted one, so nobody has to take the adjustment on faith.
How do you handle a sensor outage?
Monitoring flags a sensor or data outage within minutes. Service starts without a call from the store, and the missing interval is estimated and marked as an estimate.
Who owns the data, and what happens at the end of the contract?
The client. The full history exports at any time at no charge, also when the contract ends. There is no exit fee.
Can we try it before committing the whole network?
Yes. Start with a free check of your existing data, then a pilot on one or two stores. A network rollout follows only if the pilot holds.