Video
meets the Internet of Things
Video-analytics
technology is transforming the Internet of Things and creating new
opportunities. Are companies prepared to capture growth?
Some of the most
innovative Internet of Things (IoT) applications involve video analytics—a technology that applies
machine-learning algorithms to video feeds, allowing cameras to recognize
people, objects, and situations automatically. These applications are
relatively new, but several factors are encouraging their growth, including the
increased sophistication of analytical algorithms and lower costs for hardware,
software, and storage.
With video analytics
becoming more important to IoT applications, we
decided to examine this technology more closely. We first looked at the general
factors propelling growth and then examined opportunities by use case, setting,
and technology. To gain more insight into go-to-market models, we investigated
two out of the many areas where the use of video analytics is relatively well
established: surveillance (for multiple consumer and business use cases) and retail.1Our work builds on a
June 2015 report, “Unlocking the potential of the Internet of Things,” by the McKinsey Global Institute
(MGI). It also draws on expert interviews for insights on video-analytics
technologies and customer segments.
We found that the opportunities for
video-analytics applications will vary by setting and function. Similarly,
opportunities to make money vary along the IoT value chain, with companies
positioned to capture most revenue from software and application development.
What
settings offer the best opportunities for video analytics?
MGI predicts that IoT applications will
have a potential economic impact of $3.9 trillion to $11.1 trillion a year by
2025. Video-analytics applications, which are expected to have a compound
annual growth rate of greater than 50 percent over the next five years, could
significantly contribute to the expansion of IoT applications.
To identify the areas where video
analytics will be in the greatest demand, we followed an approach similar to
the one MGI employed when estimating the growth of IoT technology.
Specifically, we looked at potential applications that could be developed for
nine settings—all well-defined physical locations—rather than looking at
opportunities by industry.
Video-analytics applications will be
greatest in the city, retail, vehicle, and work-site settings by 2020. The most
common use cases will involve optimizing operations, enhancing public safety,
increasing employee productivity, and improving maintenance. Overall, the
largest total available market for video analytics will involve optimizing
operations in cities and factories—currently its most popular application—and
enabling various functions in autonomous vehicles, such as blind-spot
monitoring.
In all settings, video-analytics providers
will have to address privacy issues, since many users are uneasy about being
monitored, especially in situations where the data might be used against them.
For instance, production-line workers may object to video monitoring if they
believe that the footage will be used to criticize their work habits.
What
are the advantages of today’s video-analytics applications?
IoT applications usually offer more value
when they incorporate video analytics, since the technology allows them to
consider a wider range of inputs and make more sophisticated decisions. For
instance, some typical IoT applications use beacons that transmit location data
each time they connect with a consumer smartphone in a store. While this data
can help retailers track the number of visitors, a video-analytics application
would provide more detailed demographic information, such as the genders and
ages of the shoppers.
Recent
advances that have increased demand for video-analytics applications
Over the past few
years, several important developments have increased the sophistication and
utility of video-analytics applications, turning them into a much more
important growth driver for the IoT. First, computer-processing power has
improved while simultaneously becoming less expensive. For instance,
entry-level pricing for cloud-computing services is 66 percent lower than it
was two years ago. There have also been important advances in IoT connectivity
and the number of large data sets available for analysis. Finally, video
analytics have benefitted from the development of more complex software
algorithms and machine-learning technology—a type of artificial intelligence that gives computers
the ability to learn without being explicitly programmed.
Some of the most important recent advances
in video-analytics technology involve the following:
·
Real-time processing. Today’s applications can process a high volume of
video footage in real time. This feature allows users to see evidence of
potential problems as soon as it is available and take immediate corrective
action, such as deploying store personnel to monitor shoplifters.
·
Greater accuracy. Video-analytics applications are capable of much
more precise image analysis. Consider the revolution in surveillance
applications. The first ones were only capable of basic motion detection, using
pixel matching and frame referencing to detect changes in the position of
objects within their view. As a result, any movement might be flagged as a
possible problem. By contrast, current video-analytics applications can
recognize and disregard motions that previously triggered false alarms, such as
a leaf falling in front of a security camera. In addition, users can program
surveillance systems to detect specific visual patterns, such as movements
associated with retail theft or the appearance of flames.
·
Better business
insights. With their advanced image-processing
capabilities, video-analytics applications can consider multiple visual inputs,
some of which may be ambiguous and require careful processing. For instance,
they can assess the demographics and behaviors of retail customers and turn
this information into business insights that assist with product assortment and
placement, potentially improving store efficiency, customer conversion,
customer loyalty, and other metrics.
·
Access to large data
sets and more nuanced analyses. The
software algorithms in video-analytics applications are now capable of
gathering and analyzing video footage from multiple sources, thereby generating
more detailed insights. For example, surveillance applications can identify
people based on physical characteristics from video feeds collected at multiple
locations at different times. Similarly, retail applications can aggregate data
from multiple video feeds to determine the shopping patterns characteristic of different
demographic groups.
·
More innovative use
cases. With better video-analytics
applications, new use cases are emerging. For instance, some cities are
examining aggregated data from city and highway video cameras for the first
time, looking at volume, timing, and distribution of traffic. This information
may help improve traffic management and could even be used when designing
future roadways.
Such improvements have helped business
executives recognize the value of video analytics across sectors, from city
planning to healthcare. Retailers, for instance, are using IoT applications
with video analytics to assess the age range, demographic profile, and
behaviors of their customers. The software within these applications then makes
multiple recommendations about product assortment and placement.
How do
companies use video-analytics applications?
To explore the potential opportunity for
video-analytics IoT applications in more detail, we examined two of the top use
cases. First, we examined surveillance applications, including those for motion
tracking, object counting, and detection of object removal and abandoned
objects, across all settings. We then looked at retail analytics, such as heat
mapping, people counting, shopper-demographics analysis, loitering detection,
and dwell-time analysis.
Surveillance
Across settings, IoT applications can
reduce crime and protect the public. By 2025, for instance, cities are expected
to capture $14 billion to $31 billion in economic value through improved crime
detection and monitoring. Although video-analytics technology is already
central to many surveillance applications, it may play an even greater role in
the future.
Most mature surveillance companies still
specialize in simple video analytics, such as motion detection, where cost is
the main differentiator. The more advanced surveillance applications, which
have advanced detection capabilities and high accuracy, are marketed by
start-ups.
Despite recent advances, video-analytics
applications for surveillance still have many technological limitations. In
particular, they would benefit from greater video-compression capabilities to
ease transmission and storage demands, as well as better integration with other
IoT systems. For instance, it would be helpful if a video-analytics application
could detect fire on a video feed and then notify another IoT device that
activates the sprinkler system or calls for firefighters.
As with most video-analytics applications,
software development appears to offer the best opportunities for capturing
value for surveillance use cases. This layer may be lucrative because customers
typically need customized applications, rather than off-the shelf solutions. In
addition, surveillance software is often protected by intellectual-property
rights or strict licensing agreements, so companies with strong offerings may
have little competition.
The
current market and winning business models.
The surveillance market is composed of
many small customers and a few large retailers. Companies that develop video-analytics
applications for surveillance can follow several different business models, but
most now do one of the following:
·
Integrators. Under this model, companies offer solutions across
the entire IoT value chain, from solution integration to hardware, giving
customers a single source for all their surveillance-technology needs. This
model may give integrators a competitive edge, since most surveillance
customers are not security experts and prefer end-to-end solutions that cover
installation, hosting, analytics, and other tasks. In addition, many
surveillance customers, such as casinos and government agencies, must meet
strict regulatory requirements and want assistance in fulfilling them.
Integrators often assist with these tasks by subcontracting with other
providers, such as companies that install cameras.
·
Focused single-step
providers. These companies concentrate on a
single link in the IoT value chain, such as video-management platforms or wire
installation. They frequently form partnerships with integrators to provide
marketing and customer-support capabilities, reducing the need for large
internal sales teams.
·
Consulting services. In addition to providing software and hardware
solutions, consultants also make business recommendations that relate to many
major organizational groups or functions.
Retail
analytics
IoT applications could help retailers
capture between $410.0 billion and $1.2 trillion in annual economic value by
2025 by improving performance in multiple areas, including in-store promotions,
staff allocation, and shop-floor layout. Many of the most valuable applications
could include video analytics, since most large retailers already have
surveillance cameras and can use data obtained from these feeds.
While companies in many segments prefer
end-to-end solutions, this is especially true in retail, where most customers
are not as tech savvy and typically do not have strong opinions about specific
software or hardware providers. Companies that provide retailers with video
analytics fall into two categories: large businesses that tend to compete in
more commoditized areas, such as traffic counting, and niche players that focus
on retail applications, such as those that assist with queue management.
The
current market and winning business models.
Retail customers include enterprise
businesses—large companies with a national or global presence—and small to
midsize businesses (SMBs) with, at most, 50 stores. SMBs tend to request simple
applications, such as those for traffic counting and loss prevention. In
addition to such applications, enterprise retailers also seek applications that
can perform more complex analyses of customer demographics, staffing, and other
factors. In many cases, enterprise customers take advantage of their scale by
aggregating and analyzing video data from many different stores, which leads to
greater insights.
Many enterprise customers can
independently deploy and manage video-analytics applications, but SMBs often
want greater customer support. Another difference is that enterprise players
tend to pick the best application for each use case, so they may work with a
number of different providers, while SMBs typically prefer to have a single
point of contact—either an individual, or a company as solution provider or
reseller—for all their video-analytics needs.
Companies that want to serve the retail
video-analytics segment can play one or more roles. Some, for instance, create
end-to-end solutions that can be readily implemented and fulfill all of a
retailer’s needs for software, analytics, service, and infrastructure. Most of
the SMB retailers rely heavily on such providers, since they lack the expertise
needed to implement their own video-analytics solutions. Alternatively, they
could act as resellers that provide software created by other companies.
Resellers have strong existing relationships with retailers of all sizes and
can assist with implementation, deployment, service, and infrastructure.
The remaining two business models focus on
software or hardware. Software-app developers provide specific retail
applications, such as those for people counting or heat-map analysis. They
sometimes provide end-to-end solutions on a limited scale. Hardware providers
sell video-camera hardware without the software, but often have difficulty
making a profit in retail, since their products are commoditized and margins
are low. In addition, most retailers lack the knowledge and resources needed to
create complete video-analytics solutions. Hardware providers that offer small
or unobtrusive cameras may have an advantage over the competition, since many
retailers, especially high-end luxury stores, do not want to make their customers
feel like they are constantly under surveillance.
What
are the best opportunities within the IOT technology value chain?
Companies may be tempted to develop
multiple technologies for use in video-analytics applications, but our research
shows that the revenue at stake varies significantly by segment. What’s more,
these revenues are often higher or lower than those seen with traditional IoT
applications . For instance, IoT video-analytics applications tend to generate
more revenue than typical IoT applications within software and application
development but lower revenues within solutions integration and hardware.
Although the current market for IoT video-analytics
applications is relatively small, there is a large opportunity in the coming
five to ten years. As IoT video-analytics applications become more popular,
they will provide more value across an even wider range of use cases and
settings. Together, these factors could make video analytics one of the most
important growth drivers for the IoT, opening a new world of possibility to
developers, businesses, and consumers.
By Vasanth Ganesan, Yubing Ji, and Mark Patel
http://www.mckinsey.com/industries/high-tech/our-insights/video-meets-the-internet-of-things?cid=analytics-alt-mip-mck-oth-1612
No comments:
Post a Comment