skip to Main Content

Essential questions to ask before choosing a predictive maintenance supplier

With an increasingly wide range of predictive maintenance solutions now available in the market, each with a subtly different approach to tackling the challenge of optimising maintenance programmes and bringing down OPEX costs, it has never been more important to understand the capabilities – and limitations – of what’s on offer.

In order to help you understand exactly what it is you are buying and how it can meet your specific needs, below we outline 13 essential questions that you need to ask before choosing a predictive maintenance supplier. 

1. What is the scope of your offering?

“As a supplier, can you provide us end-to-end products and services (e.g. data collection/hardware, monitoring, predictive analytics and maintenance/inspection software)?  Do I have to take the whole solution, or can I get just the components that I want?  Do you provide the monitoring, or can we do it ourselves?”

Obviously, using a provider that has a full suite of solutions creates efficiencies by enabling you to get your products and services from one supplier. It also ensures that you have the flexibility to grow, by taking the parts of a solution as and when you need them – and develop your own strategy by deciding what you want to do in-house, and what you want to be done for you.

2. Specifically, what type of data analytics and/or predictive analytics do you do?

“Does your solution:

  1. Use historical data or real-time/semi-real time anomaly detection;
  2. Perform basic performance analysis only (e.g. SCADA data);
  3. Use basic vibration analysis (simply identifying that something is different/wrong?);
  4. Provide advanced vibration and performance analysis together (to tell you when and what is happening)?”

It is often difficult to tell the difference between predictive analytics suppliers when it seems they are all saying the same thing. You should ask these questions and think about how the system provides actionable information, which in turn can help you reduce your maintenance costs and increase your output.

Ideally, the solution will not just provide anomaly detection (i.e. “something is wrong”) or basic performance data (i.e. “you are not performing optimally in some area”). The best solutions tell you not only that a failure is happening, but rather what it is and at what rate it is failing.

3. Does your predictive analytics take into account my assets’ specific physical models?

When a supplier understands the physics and engineering of an asset (e.g. a specific wind turbine make and model) it can significantly increase how accurately their solution can determine the exact type of failure, and the failure rate.

It also means that you can get up and running faster, with less burden rate for implementation, because the solution does not require you to provide historical data to enable the system to learn.

The best predictive analytics solutions have models of your specific assets already built into the software. This means that they give you immediate value – and you can cut out the time spent building models.

4. How reliable are the alerts you generate?

“What is the typical ratio of true versus false positives your solution generates on a given set of assets? What about our specific assets? How does your solution improve these going forward?”

“False positives” are one of the most common reasons why predictive maintenance programs fail. It does not take many instances where maintenance personnel are sent out to respond to identified failures where nothing is wrong – or, even worse, tear down a machine to find out that it is actually ok – before people lose confidence in the solution. It is imperative that your potential supplier can demonstrate that its true/false positive ratio will not only be very good upon implementation, but can and will improve over time.

5. How does your solution adapt and improve over time?

You want to make sure that the solution you choose is fit for the future. Not only from the standpoint of continually improving the information you get, but also that it is flexible enough to meet your needs into the future as technologies develop.

6. How will you help us take a best practice approach?

“How do your solutions help us embed and improve best practice engineering and maintenance processes? Do your products have built-in engineering inspection processes, failure mode identification and rating information – or do we need to put it in place ourselves? Can we tailor it to our specific operations?”

Many predictive maintenance products only provide a framework or minimal content, requiring end users to configure them and populate them with their own data and information. The ideal solutions have best practices, processes and expert content embedded within them, from which you can generate results and improvements right away. Furthermore, they should give you the flexibility to input best practices that you already have, and build upon both, as you go forward.

7. How do your products help digitalize my O&M practices?

Going paperless is a step forward in improving any process, but with the right solution in place it is only one of a number of benefits.

The best solutions 1) enable you to go paperless, 2) provide instant access to all relevant information for any given task, and 3) eliminate redundant data entry.  They also allow for a more connected environment, eliminating silos in an organization between systems and people.

Perhaps the biggest potential benefit, however, is the ability to cut out time-consuming data cleansing, so that you can immediately start to analyse, interrogate and use your data to inform more sophisticated machine learning and AI.

8. Does your solution work with my existing hardware?

“Are your systems hardware-independent? If we do not have existing hardware, can you provide it?”

If you already have monitoring hardware in place, you do not want to have the expense of replacing it, so it only makes sense for any software you choose to work with what you have.  Furthermore, as you move forward, it is advantageous not to be tied to any one supplier of hardware or sensors. Hence, ensuring that your software is hardware independent makes good sense both in the present and the future – but your predictive maintenance supplier should nonetheless be able to advise on the right tool for the job.

9. Does your product interface with my existing (ERP/CMMS/SCADA) systems?

As the Industrial Internet of Things (IIoT) and Industry 4.0 continues to advance, it becomes increasingly important that you have the ability to bring all of your data channels together effectively – whether it is a channel of performance or monitoring data feeding in to your predictive analytics software, or a data stream from your predictive maintenance software to other systems. This connectivity ensures that you always have control and can truly leverage the power of your data.

10. How easy is your solution to implement?

“How much time and effort is needed to implement your solution?  What is required of resources in my business?”

No matter how good a solution, or its price, do not overlook implementation.  Implementation costs and time can be significant – sometimes more than the solution itself!  Additionally, there may be significant resources required from you, whether IT, engineering, operations or maintenance.  It is a good idea to have this clearly scoped before you make your decision.

11. How strong is your track record?

“What is your track record advising on equipment health and maintenance?  How many assets do you monitor across how many operations?  Has the hardware and/or software solution been implemented on my equipment types? Can you share specific instances where you have identified impending failures and averted/saved clients significant money?  Do you have specific references from clients who use and recommend your products and services?”

A healthy track record assures you that your provider’s products and services are tried and tested.  Knowing that the solution is used across multiple clients and on assets like yours also allows you to request specific proof points, case studies and references, which is critical to ensuring that their products work as advertised and that you can expect to get what they are promising.

12. How do you deliver your products?

“Can you do it for me (e.g. you provide the resources to monitor my assets)? Can I do it myself (e.g. using your software and/or hardware and my personnel)?  Are you willing to do the monitoring for me now and train my team to do it themselves going forward into the future?”

Knowing that a provider will allow you to get what you want, the way you want, shows that they have your best interests in mind. All businesses are different, at different stages of maturity in the different aspects of their operations, so having a supplier that can provide it “your way” gives you flexibility in accessing the services you need right now and the ability to build on that if you wish. In the end, this allows you to get the best return on investment sooner and at the lowest cost.

13. What is your approach to data access?

“What level of access to data do you give to your customer?  Processed data? Raw data? Who owns and controls the data? Is it me, as the owner/operator of the assets, or the product/service provider, whether it is an OEM, or independent hardware, software or solutions provider?”

Data ownership and access is a very hot topic especially in this age of digitalization and IIoT.  Businesses increasingly recognize the value of their data and how it can help their competitiveness today and relevance tomorrow.  This makes it even more important that your suppliers recognize that you own your data and that any raw data they collect is only for the purposes of providing you that particular service.  If they are sensor companies, then that sensor data generated is yours as well.  It may be that “how” a provider processes your data is their intellectual property – but you as an owner/operator should have full and unfettered ownership to both the raw data input and the processed output.

Back To Top