Incorporate Video Analytics and IoT into Your Enterprise Data Strategy

By Keith Roscarel, Director Public Safety and Smart Cities – Asia Pacific, Hitachi Data Systems

Whether for enterprise solutions, data warehouses or big data platforms, data strategy is very similar. Information is ‘harvested’ for enterprise data strategy from existing data sources, and is ultimately reconstructed to provide business insight.

But enterprises usually consider data strategy in the final mile, except during greenfield projects and major upgrades. And it’s regularly positioned as an assumed benefit that is over-shadowed by operational and strategic benefits that have already been achieved.

Enter the Street Light Effect

We also tend to assume that data that was created for other purposes, can be repurposed to provide a new dimension of knowledge. This is called “The Streetlight Effect” – in which people tend to look only where it is easiest – i.e. under a streetlight even though they lost their keys in the dark.

Some examples of where this is done today, include customer behaviour, employee retention, campus utilisation, performance optimisation, and in the prediction of future events.

Keith Roscarel

Keith Roscarel

For example, analysis of customer behaviour is usually derived only from transaction history and demographic assumptions, not from the actual behaviour of the shopper. A customer can spend more than thirty minutes browsing and evaluating a shop’s contents. However we only really understand which products were extracted from shelves, at the checkout.

Similarly limited, employee retention – is usually measured on tenure, exit interviews and performance feedback procedures. Whereas campus or office utilisation is generally measured by overall campus access and/or pre-defined schedules such as meeting times, lecture hall bookings etc.

In a production environment, performance optimisation is measured by throughput, total cost of manufacture and qualitative checks at points-of-assembly.

And prediction of future events are generally derived from historic events, and usually rely on regression techniques or spatio-temporal models for location-based prediction.

All of these have one shared characteristic. The data sources they use, are already in place and serve a different mission in terms of delivering business insights. Which means we’re searching ‘in the light’ – because there are no other available data sources.

Enter Video Analytics as a Real-Time Data Source

But, what if you had a whole new class of data source that can be readily implemented to bridge these gaps in your data strategy?

Video Analytics can provide these new data sources, for retail, employee, campus, production and other environments.

Data can be extracted from video streams and ingested into the enterprise data lake. This information can then be ued to deliver a more accurate view of real-world interactions, with capabilities such as facial recognition, geo-locationary –based apps and IoT sensors to provide identity, presence and real-time sensor feeds; and the new information can be combined with existing data sources to provide insights that are determined according to your required business outcomes.

Now Leverage These Sources to Transform Your Organisation

This means, for example, that the customer journey can be redefined, by starting when the customer creates a shopping list on an in-store app with geo-locationary features. The app could provide a ‘pick list’ for when the customer arrives at a supermarket, and by measuring the customer’s actual route, deviation from suggested route, dwell time and many other actions, an accurate profile of shopping behaviour and store utilisation, can be provided – resulting in a significantly richer insight than transaction data alone, could give.

For employee retention, the identification of a big trend – such as employee churn at 18 months, is easy to locate. But understanding why can be more difficult. Facial recognition can be utilised to understand how employees use a campus, and a ‘company app’ could be deployed to give insight into how employees are interacting with each other. This is useful because employees at risk will tend to disassociate. With monitoring, one organisation was able to correlate the reduced amount of supervisor/employee time with employee churn. While supervisors thought less time spent was a positive reinforcement, it actually had the opposite effect.

Similarly, it is easier to understand real-time campus utilisation with people counters in and out of each room. And facial recognition can be utilised in situations where you want to verify the attendees in a room.

Performance can also be better optimised in a production environment, by correlating visual activity analysis with production check points. Exceptions, defects and slowdowns can be associated with the actual real-life events. This supports real-time optimisation techniques, in contrast to infrequent process reviews: which means that lead indicators can be provided, which can be modeled for predictive intervention for maintenance, replacement etc.

Don’t be afraid to look in the dark!

Existing data sources are instrumental in providing intelligence, but, don’t be afraid of constructing new sources to enrich, evolve and potentially disrupt your business insight.

I have only touched on only a few business insights and enablements that can be expected from video analytics. There may be many other innovative applications of video analytics that can come from multiple industry touchpoints and initiatives, as big data analytics and IoT develop.

There are no comments

Add yours