5 Reasons Telematics Projects Fail - Reason #2: Data Overload

Consider a system that has access to thousands of data points from multiple sources including vehicle sensors, the internet, the environment, vehicle operators, dispatchers, and supervisors. It is not obvious which parameters will provide the most value, so the telematics module collects, stores, and transmits all of it. You might think, “It’s always possible to filter the data later, but it’s important to have as much data as is available.” Some time passes and the system slows down, the database becomes unsearchable, and your cellular data costs become unaffordable.

 

One way to visualize this is to consider a fisherman casting a net into the ocean. That net has the potential of catching a variety of sea creatures. Some might be valuable as tropical pets and others might be desired for food, but many of the creatures in the net will be tossed back or discarded.

 

 

Just like with casting a net into the ocean, some telematics (or vehicle) data is more valuable than others and much of it will be discarded or ignored, but it is not obvious how to best discriminate. Consider Tesla, a driver-assisted car equipped with 8 cameras, 12 ultrasonic sensors, and forward-facing radar. The car rapidly processes the data to make decisions about how to safely avoid obstacles, obey street signs and lights, and navigate to the driver’s destination. Many sensors are necessary to help the vehicle run effectively. The cameras and some of the sensors are useful to the operator of the vehicle. Some are helpful to a mechanic. Others might be important to a rental fleet manager, or an Uber dispatcher. Finally, some of the data is beneficial to access remotely. If the system collects and transmits data points without discriminating it will quickly become unmanageable and impossible to separate the valuable information from the noise.

 

One factor when determining a data strategy is size. Sensory data like video, audio, and tactile feedback are becoming more common as autonomous vehicles become a more mainstream reality. But one minute of compressed high-definition video with audio at 30 frames per second is about 250 megabytes. Uploading videos to the cloud is tempting, but it might make sense to keep them local to the telematics module until a specific video is requested. It is less convenient, but it significantly reduces costs associated with cloud storage and data transmission.

 

Another costly dataset is time-series data.  Time-series data includes sensor data that changes frequently over time such as the vehicle’s GPS location, accelerometer values, ultrasonic sensor data, or vehicle velocity. Each message is usually small but reporting these values dozens or hundreds of times per second could become death by a thousand cuts to the usability of the system. Each datapoint adds bloat to the database and potentially slows access for generating reports or searching for interesting insights.

 

 

Preventing Data Overload

Data overload can be a crippling problem for a telematics system, but there are some steps that can be taken to create a successful data strategy. One tactic is to organize a system discovery session with people that understand the responsibilities, needs and challenges of system users. The more specific the purpose of the IoT system, the easier it is to target specific data points and derive immediate utility.  One way to accomplish this is to do it by referencing use cases.

Consider the following examples:
 

1. An auditor needs the ability to confirm that a vehicle safety check was completed every day a vehicle was operated.
2. A supervisor should be notified when the driver forgets to buckle their seatbelt.
3. A supervisor should receive an alert when a driver diverges too far from their assigned route. The report should include velocity, GPS location and seatbelt status.
4. The driver should be alerted if the cameras or proximity sensors detect obstructions around the vehicle.
5. A supervisor needs access to 10 seconds of video before and 10 seconds of video after a collision is detected as well as headlight status, seatbelt status, vehicle speed, and brake status.
6. Alert the supervisor when a vehicle operator exceeds the posted speed limit by 7 miles per hour. Include a picture of the driver from the dashcam and the vehicle velocity.

 

Once effective use cases are established, you will be able to create a more efficient data collection strategy since all data being collected will serve a clear purpose. 

 

Assigning Value to Parameters

Another way to control data collection is by associating a value proposition with each parameter that gets stored or transmitted. If a data point does not have an obvious reason to be transmitted and stored in the cloud, it should not be uploaded.

 

Here are some examples of value-based questions:

1. How will this parameter improve safety?
2. How will it improve efficiency of the vehicle or driver performance?
3. How will it decrease expenses or cost of ownership?
4. How will it provide important insights into the business?
5. How will it provide a competitive advantage?

 

The more event driven a telematics system is, the more usable it will be. Creating a data strategy around use cases is a great way to provide value to the users and reduce database bloat, making the system more responsive. One last piece of advice is to get to market as quickly as possible to learn from the users of the telematics system. Maybe target the initial product launch to one type of user or a specific use case.

 

It is easy to get consumed with engineering a complex feature and it is disheartening when the market does not adopt the feature. My experience in the telematics space has led to the following general observation: telematics systems that focus on process improvements through automation and exception-based reporting provide more value than a system that focuses on real-time charts and graphs.  Specifically, customers can expect telematics systems to increase up-time, improve productivity, and simplify ease of use. These essential functions of a well-devised telematics solution improve user engagement and, at the same time, help to avoid data overload!

 

Conclusion

A sound data collection strategy is an important consideration for a successful telematics solution. Similar to casting a net in the ocean and hoping for a valuable catch, untargeted data collection only shifts the burden of identifying valuable insights to a later date. Therefore, the only way to keep the system performant is to discard any unnecessary data and if completed upfront, the system will be much easier to maintain and extract value from moving forward.