30 July 2019, 12:19
Based on large scale as well as small scale IoT projects (both at ML6 and before), I wanted to share 5 important lessons learnt.
These lessons are applicable in various industries for projects with a lot of data, machine learning and new innovative technology.
IoT projects are complex. Instead of focusing on a single application or business process you have to focus on your equipment, device management and communication, large scale data storage and processing, data visualisation, machine learning and data integration.
We notice new IoT projects quickly turn into 100 % IT or machine learning projects. This is a risk because it’s very likely you will face issues with for example the equipment and access to non-IoT data.
-Do we have the correct sensors?
-What are the differences between firmware version 1.1 and 1.2?
-To improve the predictive maintenance model we need ERP data but we don’t have the correct business key in the IoT database?
Make sure multiple teams are on board and you have management support to tackle these issues. Spend enough time to break down the walls between the teams and data silos.
In a lot of cases, the IoT team will be the first to capture, process and store large amounts of data. Often the tools and knowledge available within the organisation are not enough to support these workloads. At the beginning of an IoT project, it’s also not easy to define which tools and methodologies will be used in production.
In some cases, the IT organisation refuses software they haven’t approved to be installed so the IoT team will lose valuable time by using suboptimal tools.
Releasing the final solution in production will take a lot longer due to redesigns triggered by architecture changes and tool selection in the later stages of the project.
In other cases, the IoT team will start running their own tools on several personal cloud projects linked to the credit card of anybody who is willing to finance it 🙂
The best way to tackle this is to give enough autonomy to the IoT team to experiment with multiple tools. While making sure to have an IT architect embedded in the team. The IT architect needs to support the IoT team with the tool selection. The focus should be on security, scalability, costs and align with the overall IT strategy if the strategy supports the IoT use cases.
Scaling an IoT platform is not an easy task. Keeping everything up and running in production is even harder.
In some cases, companies have a strong IT team with all the skills to do this successfully. However, most companies struggle. It’s important, after the first proof of concepts to decide how you are going to tackle this.
Several options are available on the market. You can go all in and build your own IoT platform or use a selection of open-source projects. It’s also possible to use the IoT building blocks, data and ML services available in larger cloud platforms. On the other end of the market, you could also use a fully managed SaaS IoT solution. This will enable your company to fully focus on your specific use cases and domain knowledge.
Pick your battles based on the strengths of your organisation. It’s often more productive to rely on a specialized partner for large parts of the IoT infrastructure. However make sure your data is available in a scalable affordable data store and it’s easy to integrate your own ML models or APIs.
The top requirement, and a considerable part of the budget if you build this from scratch, for any IoT project are real-time dashboards.
Great to use in marketing presentations, to show off to other managers while networking but in real life, we notice the audience is actually very limited.
Real-time information is extremely valuable for process and maintenance engineers if the data is not yet available in a local SCADA system or on the machine itself.
Other audiences are less likely to pro-actively check the dashboards.
Mainly because of a lack of time or knowledge to take action immediately.
The real-time dashboards are just another application they have to use in a wide range of other tools they need every day.
Tackle this by finding the best way to visualize and communicate the IoT results to each stakeholder.
It can be as simple as sending a well-targeted notification, emailing a weekly/monthly status report or integrating directly with an ERP application or API.
IoT projects get a lot of exposure within organisations. It’s essential to inspire teams that are not yet involved or convinced people IoT will bring value to the business.
We noticed it’s essential to not only focus on the more difficult use cases, for example, predictive maintenance. These projects are complex so it might take a few years before it’s available in production.
Try to inspire people by showing the data that’s now available in real-time. Discuss the more modern data and ML stack used for IoT.
It might solve the scalability issues they face in the tools they have to use today. You will also trigger valuable discussions about data quality, data availability and data integration.
It’s very likely you will be able to pick some “low hanging fruits” and create business value to work on the more complex use cases.
Contact us for more information about our IoT playbook and the services we can offer for your IoT projects and proof of concepts.