Those were the stats Yodit Stanton, CEO of OpenSensors.io, chose to open her presentation with yesterday at the annual ThingMonk IoT developer conference in London. Stanton is a regular at ThingMonk and tends to give an honest, real world view of how the challenges relating to ‘IoT’ are adapting and changing for companies.
And this year’s presentation was no different. Stanton gave a frank rundown of her top 10 reasons why a company’s IoT project will fail. She said:
We need to do better as an industry, get past the hype and start talking about real world implementation.
Analysts and vendors get carried away with the concept of IoT, because ultimately, if they can sell a bunch of sensors and charge for access to the data - that equates to a lot of cash. However, as Stanton’s list shows, this isn’t easy to get right and a lot of mistakes are being made. I’ve kept the list as she presented it and the quotes are her own words on the day.
Failure #1 - Thinking of IoT as one industry
Everyone talks about IoT as this one big thing. There’s no such thing as an IoT company. What I imagine that I will see in the next couple of years is that we are going to have the equivalent of wearables companies, consumable focused companies, smartphone companies - and there’s no commonality between wearables and factories. Let’s not even pretend there is. Let’s start talking about the distinct stacks that are necessary for each and then the trade offs and the needs.
Failure #2 - Everyone is becoming a tech magpie
Just because you talk about blockchain, [doesn’t mean you should use it] - I don’t even know what blockchain is going to solve. If you have IoT problems, I will quote back at you, ‘I have 99 problems, but big data is not one’. If you have tiny, tiny, packets from 100 sensors being sent over the wire, you do not need a Cassandra database. Jesus Christ, you might not even need a database. Just stop it. Let’s not talk about reference architecture, what do you need?
Failure #3 - Talking about that architecture, middleware is only 1% of the solution
Saying, oh yes I have Kafka, I can do IoT, is like saying ‘I’ve got a Dynamo instance, I’m going to create an Amazon store’. If you need Kafka, go for it, but spend 1% of your time and energy on it. You have much bigger problems than your middleware.
Failure #4 - Thinking you can data jujitsu your way out of crap readings from lots of cheap sensors
The other thing I see a lot of is people deploying the cheapest sensors. They think they can use data magic to solve the problem. Your data scientists are not alchemists. They cannot do anything with the shit that you send them. Your Kafka system is just going to be like a sewage system if you have crappy data. Get better sensors, spend money on that, and then do some simple aggregation. IoT is not a commodity. Do it properly. Have one sensor, rather than 1,000, but spend some money on it. Appreciate the hardware, people. There’s a lot of innovation still to be done in this space.
Failure #5 - Not building a multidisciplinary team
People think that they’ll do an IoT project and fill it with data people, because they’re what matter. It’s not true. You need hardware people, software people, networking specialists, QA people, project managers. You need this because you are shipping complex products. Make sure that you tell the team that they need each other, make them respect each other. Don’t have pre-Madonnas.
Failure #6 - Not spending 40%+ of your time and money on logistics
The other mistake is not thinking about logistics. IoT cannot ignore the physicality of it. We have supply chains, we manage complex supply chains. We have an inventory management system, because we have to manage hardware. We have distribution networks, I have to figure out how to get sensors from the UK to other countries. This stuff takes up 40% of our time and money and this stuff should take up 40% of your time and money, at least. These are complex environments, so you can’t just chuck stuff at them. People are too used to deploying SaaS.
Failure #7 - Not understanding the importance of provisioning until after shipping
Provisioning is such a big deal. [You need to know] where the sensor is. If you don’t have a plan for your ongoing maintenance, you will have a problem. If you don’t think about this before shipping, it’s going to cost you a lot of money. I’m not seeing enough people worrying about this problem. It would be nice if we collectively started talking about this provisioning problem, because it’s actually really interesting how you tie a ‘thing’ as an asset to the physical world and maintain it. Your analytics is meaningless unless you know exactly where the device is.
Failure #8 - Trying to run before you can walk
I refuse to take on projects where people say they need 50,000 sensors and ask for a price. No, we don’t want that. I know it’s not glamorous, but start with a few sensors, get it working, assess it, really understand the value. Does it meet the end goal? If it doesn’t, scrap it. Then get 50 sensors in the wild, then 500, and then build it up. No one can tell me how much it’s going to cost or how much the complexity is going to be, unless you do these things.
Failure #9 - Not learning from computing history
We are like ‘IoT is a magical thing’. Well, you know what? In the 1990s we shipped networks, we shipped hardware, and software. Who was a Microsoft administrator? That will come in handy. It’s a network of computers, that have a local area network, that talk to a network server, up to the internet. People have figured out a lot of the problems. Let’s not reinvent the wheel.
Failure #10 - Not appreciating how big of a deal all this is
I could make more money and have a more glamorous life, by doing something else. I don’t do this because I want to sell volumes of sensors. I do this because, for what we do, we are managing data in a physical infrastructure. And we think a lot about how we manage it efficiently. We think a lot about how we deploy sensors. The opportunity for IoT is for us as human beings to use our physical infrastructure much more efficiently. In the next 20 to 30 years, we have to use our infrastructure really well. And for me, this is the opportunity. Let’s not forget how transformational this is, and let’s not forget that.