Predictive analytics business case: let's talk about planes for once

Everybody is talking about predictive analytics these days, but you’re not really sure what it really means, or what asset it could be to your business. Well, let’s be a little more down to Earth then, and talk about a real business case.

Let’s say that you work with airplanes. Why airplanes? Well I don’t really know, apart from the fact that it’s pretty cool that they fly. Hum… As I was saying, you’re working with airplanes, and with your team, you’re in charge of supervising the maintenance planning of aircraft engines from your commercial fleet. Your boss told you to improve your main KPI, the Holy Grail: the overall cost.

Incidentally, during the late session of the last day of the international conference on prognostics and health management that you went to, half sleeping, you heard about the promise for considerable costs savings, by reliably estimating from sensor data the remaining life expectancy of engines.

That was quite a revelation. 

The day after, fired up and freshly dressed, you thought: « Indeed! What if I could be warned early on, and automatically of engine parts deterioration? I could avoid unscheduled maintenance, or increase equipment usage by changing operational characteristics (such as load)! ». You thus decided that you wanted to receive every day the updated list of engines who will very probably fail between 15 days and 30 days. So that you could organize your staff and stuff. For example, you know that your equipment supplier needs four working days before any delivery, so you’ll have to take this into account. And you also know that every other week, your team will lose half of its technicians to the space rocket project (damn space junkies!).

But unfortunately, you’re stuck. You cannot read sensor data like a fortuneteller reads palms. That’s where predictive analytics comes in, flying through the skies to save the day. (Disclaimer: we are not fortunetellers, or Superman, we are just data citizens. But we’ll do the job as good as any of them.)

Predictive analytics promise: you will know everything. Beforehand.

So how does it work? Predictive analytics provide you with a model. « What’s a model? », you say. Well, simply stated, a model is a piece of (magical, mathematical, esoteric) code easily embeddable in your existing process. You can use it whenever you want to and you can apply it to whichever engine you want to. You simply feed it with the freshest data from the sensors of each engine, and that’s all. The answer from the model is the probability that the engine will have to be declared ‘out-of-order’ within a specific timeframe, for instance: during the next 14 days, within the next 15 to 30 days, beyond 30 days.

Providers of predictive products, like PredicSis, known as one of the best on the market, build models from your past sensors data of the now out-of-order engines. According to the evaluation on past data, the model is able to deliver a prediction which is 90% accurate in this case. This performance is enough for you to modify your business process and act upon the engine that the prediction declares to be out-of-order within the next 14 to 30 days.

How does the magic works?

First, we need you! Erm your data! Great, seems like you have plenty of them, presented like this:

Credits: Excerpt of tabs provided by NASA.

Credits: Excerpt of tabs provided by NASA.

Well, considering that you’re a wise business manager, you don’t easily hand over your data and you prefer to give access to the data publicly available here. Then a data citizen of your team prepares the data. To be totally honest, she found that Microsoft already produced a script to prepare the data, so for now, let’s just re-use the R script. 

We’ll show you our product for your team to efficiently prepare the data in another post. ;)

Anyway. As the data are now prepared, the data citizen can spend hours building tons of models in order to select the best one.

Well, this was the old times, before PredicSis.ai You can see in the video below the new way of creating accurate models. 

Please note that we could have built the model with PredicSis' Predictive API. That is equally efficient. If you want to give it a try and tackle your problem using your data, have a look at our ML Studio. Please note that ML Studio lives on top of the predictive API, allowing your data citizens to easily embed the model in your business process.

Conclusion

This is the way how fast and accurate predictive products, easily embedded into your business process, can help you proactively improve your KPI. Our next blog post will give you more insights on the core features developed by PredicSis, making the process data-preparation free.


Intrigued? If you want to get a sense of PredicSis.ai, we made it easy for you through our demo page