In the realm of sensors

The collection of personal data is always a balancing act between what is technically feasible, socially accepted and legally permitted. It is important to take a very
close look at the data required and to consciously choose from the available technologies. This is the only way to choose the right analytics solution.

A look at China shows what is technically feasible with analytics. Smart cameras recognize people wearing face masks, measure body temperature
and distance to another person, and even offer payment by eye contact via sensor. The data gathered can be used in a variety of ways, for example
to redirect customer flows or personalize advertising. To make this work, however, the technology constantly monitors the customers; collecting comprehensive personal data, especially in the Asian region – which, according to European ideas, is an absurdity under data protection law.
One might think so.

The legal framework for the protection of personal data in Europe is provided by the General Data Protection Regulation (GDPR). Similar regulations based on GDPR have been introduced in the US (California, Illinois) and Brazil. In Germany, the GDPR is implemented as a Basic Data Protection Regulation, or DSGVO for short. Although the regulation contains much stricter regulations than those that are customary in China, but there are also many grey areas. Moreover, regulations in this area in particular are constantly being adapted or tightened up – jurisdiction is only gradually
developing a clear line. What makes things even more difficult is the fact that case law actually always lags behind technological development.

Social acceptance is just as important. Cases of abuse in the past also lag behind data collection in line with data protection regulations. And the
current debate about state surveillance and discrimination, for example with regards to the USA and the „Black Lives Matter“ movement or the exit
of large technology companies such as IBM from face recognition technology, is fueling prejudice.

Before organizations choose a technology, they must clarify what information
(objects and attributes) should actually be gathered, and why? Decision makers
should ask themselves the following questions:

Why should measurement be performed?

■ The Why is the central question from a business perspective

■ Typical questions

• Analysis data for decision making
Example: How are shelves/merchandize
optimally placed to maximize sales of
certain products/categories?

• Forecast data
Example: Staff scheduling/shift planning
based on historical visitor numbers, sales
campaign information and weather data

• Live-Trigger
Example: Automated playout and optimiza-
tion of content, depending on present persons
in front of a display, or playout of push
messages to customers

Which objects should be detected?

■ Nowadays, there are two basic ways to detect
people:
1) Optical detection of persons, e.g. by camera,
IR sensor, laser (see below))

2) Tracking of radio signals from smartphones

Advantages:

• Direct customer communication via smartphone possible (requires app)
• Link to online/mobile data profiles, e.g.
  purchase history (requires app or opt-in)
• Inexpensive infrastructure
  
  Disadvantages:
  
• Inaccurate people counting (device ≠
  person) – some people do not have a smart-
  phone, others have several
• Less precision in tracking location
  Area counting, no line counting
• Behaves similarly for goods: optical recog-
  nition (cameras) or recognition via radio
  signals (RFID tags)

Which attributes should be tracked?

• This question is decisive for the analyses that
  can be conducted. The rule is: „Less is often
  more“. Because more data means:
  • not necessarily more or better insights
  • greater analytical effort
  • higher costs for transmission (cellular)
  and storage
  • lower customer acceptance

• It is therefore important to think about what
  decisions can or should be made based on
  which data set. (Often there is hope for a „lucky
  insight“ following the principle: „Let’s measure
  as much as possible, then we will find something“).

• Some examples of which attributes can be
  recorded for persons:
  • Number/frequency/area distribution (heat
  maps)
  • Characteristics (age, sex, mood, etc.)
  • Opportunity to See (Media, Merchandize)
  • Interactions (e.g. social distancing, interac-
  tions with staff, products)

• A whole range of different attributes can also
  be measured for goods, such as:
  • Quantity
  • Position and alignment
  • Brand
  • Interactions

• Tracking frequency (where, how often and for
  how long) is then also decisive for the analysis
  options and data volume

Only after these concept questions have been answered, technology comes into the game.

Hardware

A wide range of different technologies is available for the hardware:

• Hardware for optical detection:
  • Infrared sensors (IR)
  • 2D cameras
  • 3D cameras (stereo)
  • Thermal sensors
  • Time of flight sensors (ToF)
  • Structured Light Sensors
  • Laser/Lidar sensors

Popular AI applications are algorithms for object recognition and attribute determination such as face detection, demography (age, gender) or posture and direction of gaze.

• Hardware for detecting radio signals
  (smartphones or tags):
  • GPS tracking (outdoor)
  • WIFI tracking
  • Beacons (BLE: Bluetooth Low Energy)
  • Visual Light Communication (VLC)
  • Ultra-Wide-Band Radar (UWB) sensors,
  with tags
  • RFID sensors, with tags

Business Intelligence (BI) or dashboard
system

Hardware manufacturers typically provide the data via a cloud interface, and in case of data privacy concerns, there is usually also an on-premise option. In most cases, the online tools also offer simple options for displaying and evaluating the data via dashboards. When combining the
data with other sources, the solutions offered by sensor providers usually reach their limits quite fast.

Of course, data can also be integrated or imported into third-party solutions or existing BI systems. In this way, different data sources and more complex correlations can be analyzed.

Analysis platforms

The weak point of most solutions is the actual data analysis. Useful recommendations or actionable insights are generally not provided. The current analytics industry focuses on providing the infrastructure (especially sensors) but lacks in supporting customers to make sense out of data. Specialized consultants and data analysts like invidis partner MM Customer Strategy are therefore necessary.