Predictive maintenance is, arguably, the most hyped application of IoT technology currently available to the enterprise user, and it\u2019s easy to understand why: Getting greater insight into industrial machinery, fleets of vehicles or anything else that can be digitally instrumented seems to offer a fairly direct path to savings through lower maintenance costs and less downtime.\nBut it\u2019s not as simple as just grafting sensors onto existing equipment, according to experts, and reaping the benefits of predictive maintenance isn\u2019t an automatic win for the asset-heavy businesses that can profit most from this IoT implementation.\n\nThe challenges, according to ABI Research, can be seen clearly in the track record of IoT usage in the oil-and-gas industry. Offshore oil spillage is still relatively common, despite the widespread use of IoT services, and a big reason for that is that the AI\/ML piece of IoT just isn\u2019t that well implemented as yet.\n\u201cWhile top oil players market themselves as pro-tech, with predictive analytics being the key to their investment,\u201d ABI analyst Kateryna Dubrova wrote last month, \u201cconsulting firms and the hiring of a few experts is not making the technology work and subsequently not making a difference in preventive measures.\u201d\nNot having a top-to-bottom plan for getting real value out of the oceans of data an IoT project can generate is the biggest reason that companies don\u2019t see measurable results from predictive maintenance, said Forrester analyst Frank Gillett. Businesses sometimes get excited, place sensors everywhere they can, and then expect the payoff to develop on its own.\n\u201cThere\u2019s lots of examples of people looking at sensor data and then trying to build a business case, rather than trying to build a business case first,\u201d he said. \u201cIt\u2019s like walking around with a hammer and not finding any nails.\u201d\nMuch of that has to do with the fact that making AI and machine learning work correctly is difficult. Companies need plenty of data science expertise \u2013 whether in-house or from their vendors \u2013 to make sure that training data is teaching the model the correct lessons. Moreover, moving data around freely is tricky in certain industries, where companies might be reluctant to hand over operational information to a third party. For example, a manufacturer might not want to release performance info on factory equipment if that info could provide outsiders with an insight into confidential processes at work.\nUsers also need a much more holistic understanding of how predictive maintenance actually drives business value, according to Cambashi principal consultant Alan Griffiths, who also noted that institutional expertise in IoT is invaluable to make everything work.\n\u201cWhen you look at the technology required, it\u2019s quite sophisticated,\u201d he said. \u201cEach [component of IoT] is fairly well understood, but it can be complicated to implement, especially with old-fashioned IT departments.\u201d\nYet it\u2019s easy to understand why companies are in such a hurry to adopt the technology \u2013 there are simply too many potential benefits to ignore. Tracking maintenance information offers businesses additional surety that the money they\u2019re spending on replacements and repairs is being spent in the right places, and lets them cut down on unnecessary outlay.\n451 Research vice president Christian Renaud said that the possible upsides are hard to overstate.\n\u201cThere\u2019s a bunch of different ROI things, production, asset value, worker safety, and then all the fluffy benefits where you collect all the data from these things and glean insight from historical trends,\u201d he said. \u201cThis is something that has been the ultimate use case, long before we started calling this IoT.\u201d\nAnd, despite some hiccups and a knowledge gap where the data analysis piece of the puzzle is concerned, there are plenty of users out there making predictive maintenance work for them. A recent survey from 451 shows that predictive maintenance is the most-used application of IoT among operational technology companies, and that, of those, the vast majority report at least \u201csomewhat positive\u201d ROI.\n\u201cThere are so many benefits to getting telemetry off these machines,\u201d said Renaud.