In my last blog I talked about the tail wagging the dog in IoT. The National Audit Office (NAO) recently announced that the Fit for Work scheme is going to cost more to implement than it will save. This shows that this problem isn't just limited to IoT, but also links nicely to the topic under discussion today, which is, how much are the insights worth?
Google has made many people think that data is free. No one owns it and someone else will filter it for no charge. Of course that is because you are the product not the consumer. Also with Google there is no guarantee of accuracy and you have to decide which links to click. In IoT this data is either buried in the enterprise or needs collecting and analysing. This will cost, and the value needs to be realised. If analysing the data costs more than the value to be gained, then the tail is wagging the dog again.
You may have heard of Pareto analysis, or the 80:20 rule. Applying it to IoT suggests that 80% of the value comes from 20% of the analytics that you could do on your data set. In other words, somewhere above that 20% mark it will start to cost more to analyse the data than you will get back in operational savings. The exact point will vary, but it is clear that at some point, not only will the tail be wagging the dog, but the dog will also be chasing its own tail. This is why every IoT project should have board level requirements driving them and real business outcomes driven from those requirements. Without this, the project will not know when to stop. Note this doesn't mean that, as part of the discovery phase you should not have a trawl through all your data, but you should not build a production system that works on the low value analysis unless it is worthwhile.
The reason it is so important is that you cannot point data scientists at a pile of data and say 'go play and let us know what you find', as they will do as they are told until they are told to stop. You have to give them clear goals as to what information you want out, and a measure of the effort to extract it. There are two parts to this, firstly analysing existing data, and secondly providing a list of what additional data is needed to meet the goals. It is this additional data where Pareto can really bite. For every parameter measured there is an additional cost to be balanced against the value that the data brings. And this brings in another dimension for consideration. Do you have to measure every one of the devices you need to manage a trend, or merely sample a subset of your equipment? And can you find the critical parameters that need to be monitored to minimise the number of 'things' you need to install?
Healthcare has been doing this for decades. They run a trial with a few patients heavily monitored, and then analyse the data to find the critical parameters for monitoring and produce new clinical guidelines. Once the staff know what the parameters are, they can start to use this to improve care. Huge strides have been made in both critical care and in early diagnosis of new treatment pathways, but without needing a complex and expensive analytics solution running all the time. With IoT you have to do the same and monitor everything on a small scale to understand the critical parameters that deliver your value before you roll out a potentially overly complex solution to all your estate.
So the data scientists, working with the business can start to generate some cost:value trade offs to feed back up to the stakeholders. This then allows proper modelling of the benefits and investments required. From this point the actual roll out can be defined. In some cases you need a big analytics cluster to generate your insights, but in many you only need a few 'deep dives' to define a simpler system.
What you should take away from this is the following;
- IoT is not a single pass transformation. You need to go around the loop to find where your value is in a proof of value/research phase
- Do not give the technology and the data scientists free rein. Keep the tail under control
- You need the right expertise to steer through the minefield.
CGI is well placed to help you on all steps of this journey. Find out more visit www.cgi.com/uk/internet-of-things