ฉบับเบต้า · เนื้อหาบางส่วนแปลด้วยเครื่องหรือแสดงเป็นภาษาอังกฤษ
ข้ามไปยังเนื้อหา
DEEPINGSOURCE
Back to case studies
Premium automotive showroomMeasured casesAudience · Brand CX team · Showroom operations

What staff counted by eye, a camera now counts

Family group or not. We moved a rule that changed from person to person onto the machine, then checked it against the manual tally.

From problem to result

Problem
Floor staff observed, sorted and tallied by hand. Visitors were missed during consultations, and the criteria shifted with the observer.
Applied and measured
The showroom was already measuring visitors, paths, dwell time and the funnel with saai insight. What remained was customer type. Staff who greet and consult at the same time could not record every visitor as it happened, and the tally changed with who was watching. We built the AI classification to match the categories the brand's CX platform uses, ran it side by side with the existing manual tally, and switched over after the comparison.

saai insight

Measured result
AI sorts groups from the video and produces the statistics. Staff spend their time on consultation and service.
Illustrative

Customer type · visitors · paths · dwell time · funnel

What it uses

Illustrative

On-site manual sorting and tallying → automatic classification and statistics

What changed

Field-verified

The manual tally and the AI classification were compared side by side before the switch

How it was checked

Field-verified

Analyzed after anonymization. No face recognition, no identification of individuals

Personal data

Evidence and scope

02 VerifyMeasured casesAudience · Brand CX team · Showroom operations

Adoption stage

  1. Pilot
  2. Deployed
  3. Scaled

How this was verified

Before the switch, the manual tally for the same period was compared side by side with the AI classification to confirm accuracy. Site names and the details of the classification criteria are not disclosed.

A real case at a premium automotive brand showroom. Site names and the details of the classification criteria are withheld to protect the client.

Counting people without recognizing faces

Sorting customers by type sounds like it requires looking at people, but what this case looks at is not a person. It is shape and movement. The video passes through anonymization before analysis, so the information that would identify anyone is already gone. There is no face recognition and no identification of individuals. What is left is how many moved together, and how long that group stayed in each zone.

So what this case produces is not a record of a person. It is statistics. What to change is decided by the showroom operations team reading those statistics.

How the anonymization works is set out in Anonymization technology.

Related coverage: DEEPINGSOURCE automates customer-type tallying at an automotive showroom

Get in touch

Let’s define the next improvement for your space

Request a consultation