Skip to main content
  1. Projects/

IoT based Visitor Tracking Application

·1069 words·6 mins

METRO Market Bangalore
METRO Market Bangalore

Introduction #

Last weekend I went to the Metro supermarket to buy some groceries. Metro is a members only supermarket and sells groceries at wholesale price. Since it was the beginning of the month the store was pretty crowded.

I was wondering “If there is a tool which can track the customer movement within the store and provides real time analytics like google analytics, it would have been helpful for the store management. The system can literally monitor all the customers within the store till they exit and can provide a lot of meaningful insights based on their behaviour.

There are possibilities for automating many of the routine activities depending on customer behaviours. With such a system available, it is possible to act quicker and make decisions confidently. We don’t have to rely on the field staff to know about the actual customer experience and multiple stores can be managed by a small centralized remote team.

Challenges on Running a Supermarket #

Running a supermarket is much more expensive and labor intensive than running an online store. But yet there are only less technology available today which cater their needs.

Challenges on Running a Supermarket

I understand, I’m comparing apples and oranges here, tracking a physical space is not as straightforward as tracking a web page. It requires a technical innovation and a framework which works along with external hardwares like sensors, cameras, etc.

But looking from the business perspective I guess, it is a highly needed product, it can help business owners to learn their customer behaviour better and make informed decisions. It also gives the opportunity to experiment new techniques and measure its performance quickly. For a supermarket like metro, customer satisfaction is at most KPI they would care about.

What’s tracked Today? #

Inside the store, I can see only CCTV cameras fixed everywhere, which I guess will be used mainly for security monitoring. Maybe somebody’s watching all the video feeds somewhere in a closed room, but even if it is yes, this is how the room would look like. We can hardly understand anything about the customer. Lots of valuable information is lost every day.

A control room with 100s of camera feeds.
A control room with 100s of camera feeds.

There was a feedback box kept near the exit gate. It had only the standard questions with 1–5 rating input. I’m not sure how that details would be utilized. Without knowing the actual experience of a customer it is hard to make sense out of it.

With the combination of other customer details like Purchased details, Visit History, Billing Info, Feedback, Store inventory etc, it is possible to identify a lot of helpful information which can help the store owner.

Promising Technology #

One of the promising techniques which I came across is, “Real Time video Stitching’’ It is a technology which helps stitch the video feeds from multiple cameras into one single video . All the cameras are affixed at the roof of the store, to cover the entire floor area without any blind spots. By stitching all those videos into one, we get a single video with all the customer movements. The customer path within the store can be mapped without any disconnect.

The final video can be fed into an AI system which would process and identify the key customer events like, stopping pointsspeed of momenteye directiontime spent on each sectionviewed items etc. such a technology is already used in sports the same can be used here as well.

Realtime Video Stitching and processing
Realtime Video Stitching and processing

However, exploring the technical possibilities are beyond the scope of this project. The current design is based on the assumption, we get all the above mentioned data, and a program identifies the key events and patterns.

Tracking Customer movement
Tracking Customer movement

Combining Data #

As I mentioned already, metro is a members only supermarket, which means details of every visiting customer are known and other information, with the customer traveled path and key stop by points, it is possible to analyze which segments they have visited and the time on each of them. Their feedback can be understood better and necessary action can be taken.

Here are some of the metrics which can be identified from the above combination of data.

Total time spent on the store, Segments visited, Time spend on each segments, Segments visiting segments, Interested Segment — based on the eye direction and speed of moment, Random moments, Cart Abandoned, Billing Queue time, Session replay, Missed opportunities — comparing the time spend on each segment and billing items, Customer Feedback — Feedback with customer moment in the store.

Customer Analytics #

Here is a wireframe which demonstrates the individual customer activity within the store. It also aggregates all the other details about the customer data and provides a complete summary of a customer visit.

Customer Analytics
Customer Analytics

Key metrics #

The key insights about the customer like total visits, Average shopping time etc, can be shown realtime at the top of the dashboard.

Real Time Feed #

The complete aerial view of the store with all the customer moments. Details of the customer, performance of a specific segment can be viewed on a click.

Interactive Store Layout with live customer moment.
Interactive Store Layout with live customer moment.

Store Analytics #

A detailed presentation of the store performance, with interactive graphs and charts.

Store Analytics
Store Analytics

Conclusion #

Technical innovation for a physical business is still infant, many of their needs are still untouched. Most of the tools available today are focusing on the web market but not the real world conventional business. There are a lot of benefits this tool can offer for business owners. Here are some things which I can think of.

A solution like this would reduce the dependency of the field staff input to know the customer needs.

Performance of a campaign can be monitored quickly and accurately.

In the future a moment of a specific customer can be completely replied whenever they want.

Possible to monitor multiple stores remotely by a small team of intelligence.

In depth custom data will take our understanding to a different dimension and a lot can be achieved in the future.

Identify the peak and ideal time of the store, accurately and assign the workforce accordingly.

System to identify the repeated key events and patterns, and alert to the management to make wiser decisions.

Track the speed on each billing counter and take actions to improvise the under performing ones.