How indispensable is your smartphone in your everyday life? Be honest. Could you go more than a day or two without it? How about an hour or two? Whom do you thank for that? Probably a name you know, Qualcomm. Qualcomm connected the smartphone to the internet and that changed everything.
It’s impossible to imagine a smartphone without features Qualcomm made possible. Every time you snap, shop, navigate, stream, download, store something or even just talk, you’ve got the power of Qualcomm technology to thank. Examples include backlit selfies that showcase image stabilization technology, finding the hottest restaurant on a crowded street even if you have no sense of direction, gorgeous graphics, as well as lightning-fast video streaming and immersive 3D experience.
Qualcomm works closely with the world’s leading network operators – such as China Mobile, Vodafone, Telefónica, AT&T, and Verizon – to help connect new industries, new services and experiences that are changing every day. They also work with the manufacturers who design and sell smartphones with Qualcomm chips to better understand how their chips can be designed more efficiently. This, in turn, allows the service providers to add more phones with newer features! A win-win-win for Qualcomm, the network operators and smartphone manufacturers. We could add a fourth ‘win’ here and that is we, the consumer too!
Qualcomm is in partnership with the network operators refining those networks with a renewed culture that relies on;
- Sharing and allowing full access to analytics and data;
- Sophisticated analytics; and
- An innovative ecosystem that relies on data NOT moving.
“I have data scientists in our group, and a lot of people with machine learning expertise and everything else, but the biggest impact we’ve had is just sharing information. Changing from a need-to-know company where people would share data because, ‘I’ll give it to you if you need to know it’, to almost like a need-not-to-know, like, ‘Why can’t I share it?’ And so that was really the driver…to share data across the company. Then the more advanced insights you can get from the machine learning, the data scientists, and everything. But number one, get people access to as much as possible.” Craig Brown, Senior Director of Technology
The key to number two above is actually number three. Okay, we’ll clarify. For Qualcomm, sophisticated analytics depend on the simplicity of the ecosystem. Eliminating movement and leaving data in its place ensures efficiency.
“The Teradata system used by Qualcomm is currently on AWS. It gave us the benefit of instant access to it. We can just turn it on, right? It gives us the thing where, in our development systems, we can use it in various environments, in various places. It’s not just in our Las Vegas data center…Building a great end-to-end ecosystem that best meets our needs, best meets our existing momentum, or best suits our existing momentum as well as giving us the best results in the future. And it, again, comes down to simplicity. Don’t move data around. And a much bigger thing is that Teradata is more than a Teradata box. Teradata is huge.” – Craig Brown, Senior Director of Technology
But we have buried the lead. The goal state is really innovation in the chips and the networks that our smartphones rely on and that Qualcomm gets when they add it all together.
Partnering with network operators and manufacturers around the globe, Qualcomm strives for innovation. Together, they test networks in existing and emerging markets to understand “total network profile.” There is software running on the phones to determine how they are operating. With geo-location knowledge, Qualcomm and the network operator collect data off cell phone towers to analyze multi-structured data about handoffs from cell tower to tower. For example, if you have 100M phones/subscribers with 2B phone calls in a month, you can expect to collect 30 – 40 different types of data including cell phone year, manufacturer, format, service, etc to reveal:
- Where does the network misbehave?
- Is there an area of the network running much better than expected?
- Where is the switching station?
- Do certain phones and technologies perform better with certain cell towers and technologies?
- How old is the cell tower? What parts are being used in the cell tower?
- Where are the anomalies?
- Can we exploit those cases?
Working alongside consultants from Think Big Analytics, Qualcomm data scientists discovered the key lesson from this engagement was that simple things often work the best. They found that a manually created decision tree or a simple probability model would solve many cases. When more advanced techniques were needed, they tended to stick with supervised machine learning approaches such as support vector machines (SVMs) or random forests where the model will indicate the criteria it used to make a decision. In this particular engagement, deep learning was a last resort for situations where it’s often difficult to show why a decision was made, and thus, more difficult to get adoption. For example, Qualcomm previously had the goal of clustering cell towers into groups, and this lent itself to an unsupervised approach as well as the use of k-means clustering.
“Now what we can do is we can say, ‘Ah, let’s go and measure what we found here, and let’s go and measure the entire network and see.’ One of the machine learning techniques is where you do unsupervised learning, where you cluster things. And so I’m not teaching the machine what to do; it’s just looking for patterns, and it says, ‘Those things are acting the same way.’ And so the thing I just found, the thing that was playing up, if I can now, across an entire country, say, ‘Where else is it playing up the same way? Go out and try and fix them the same way.’” – Craig Brown, Senior Director of Technology
Congratulations to Qualcomm! Your equation of culture + analytics + ecosystem = business outcomes is your success!