Todays interenet of things market includes countless consumer gadgets such as routers, Internet-connected video cameras and smart TVs, as well as innumerable enterprise IoT devices for business, scientific and industrial applications.\nProducts in the latter category have traditionally been a "systems" sale, so they were expensive, pricing was secretive, consulting and installation services were usually part and parcel of a sale, and the real cost of a solution was dependent on how much you were willing to pay.\nBut enterprise IoT selling tactics are changing in response to the commoditization of IoT. Products are becoming cheaper, more easily sourced, transparently priced and much simpler to install. Unfortunately, not all players in the IoT world have responded to this trend, so buying some enterprise IoT products can be a lengthy and unnecessarily complex process if you're dealing with the "old school" vendors.\n\nI experienced this firsthand in a recent consulting engagement that\u00a0involved the analysis of enterprise IoT temperature monitoring technology as well as a hands-on evaluation of a selection of leading products. While temperature monitoring might seem a little specialized, that exploration yielded useful lessons that pertain to the broader enterprise IoT market, particularly remote sensing.\nBefore getting into the issues involved with selecting solutions, however, it's important to review what an enterprise IoT system looks like. You can break down the general enterprise IoT architecture into three main parts:\n Mark Gibbs\n\nEndpoints that include one or more sensors;\nGateways that aggregate data from multiple sensor endpoints and forward it onto the backend services; and\nBackend services where the sensor data are analyzed, stored, routed, and displayed.\n\nWith that in mind, here are 10 questions to consider when sourcing an enterprise IoT remote sensing solution.\n1. How much data will your sensors generate and how often?\nThe size of most IoT data payloads transferred from sensor endpoints to gateways is usually around 50 bytes. But aggregate the data transmissions of, say, 1,000 endpoints each reporting their measurements once every 20 minutes, and the local bandwidth usage becomes significant. Another concern is that most of the wireless technologies used for enterprise IoT systems are in the unlicensed ISM band, so interference from sources other than your own systems could become a future concern and underlines the importance of planning for both the short and long term.\n2. What sensor endpoint networking technologies are appropriate for your environment?\nCommunication can get tricky when you're connecting a cluster of sensor endpoints to a gateway inside a building. The premises' structure (walls, ductwork, electrical cabling, roofing) can cause all sorts of wireless communications problems including attenuation, reflections, and interference that collectively weaken and distort signals. Technologies such as Wi-Fi and Bluetooth only deliver short-range connections (300 feet maximum) and a poor radio environment can reduce that range considerably. Products using Bluetooth BLE, ZigBee, and Z-Wave mesh technologies look promising for building robust sensor networks, but currently they are less common in the enterprise IoT market. Currently, your best bet for enterprise IoT networking are the "chirp" spread spectrum technologies such as LoRaWAN and some of the proprietary 900MHz-based systems. These generally deliver excellent range (6 miles or more) and high reliability even in environments with lots of radio noise and radio-opaque or interfering structures.\n3. How are you going to connect the gateways to the backend services?\nConnections between gateways and backend services, which are either hosted locally or on a cloud service, usually use either in-house Wi-Fi or connect directly to an in-house Ethernet network. Alternatively, where sensors and gateways are used in a mobile environment (say, monitoring the temperature of vaccines being delivered from a laboratory to a hospital) or in locations where local premises network services aren't available (for example, monitoring soil moisture in a vineyard), gateway to backend services usually use cellular communications. A final consideration is that while some vendors offer backends that can be deployed in-house, these are generally less flexible than cloud-based offerings, require a lot more in-house IT support, and are less suitable for enterprises with multiple locations.\n4. What architectures, protocols, and standards do your potential suppliers use?\nWhen I first started on this project, I expected it would be relatively easy to find out how the various vendors had built their enterprise IoT offerings. I was wrong. The chaos that many vendors display when you start to dig deep is breathtaking. I talked with pre-sales people at leading enterprise IoT companies, as well as technical support specialists, who weren't adequately trained on their company's products and, in several instances, actually gave me wrong information. Before you make your final vendor selection, make sure that you really know exactly what's being offered.\nWhat will make your research more complicated is the confusion caused by marketing in many of these companies. Some of the larger vendors in the enterprise IoT market have grown by acquisition and have multiple brands they're trying to consolidate under a single marketing strategy. One major player in the market has five different brands they're trying to reconcile. When I started testing their equipment, I had to get on the phone with an "onboarding" specialist to provide the device IDs so he could register them with the backend services. Even though I was only configuring one sensor endpoint and one gateway, this onboarding process was time consuming. I asked the onboarding specialist how this would work for a thousand devices, and he said he had no idea.\u00a0\nThe takeaway here is be prepared to dig deep.\u00a0\n5. What level of security should your enterprise IoT systems have?\nIt's already been definitively and dramatically shown that unsecured IoT devices are really dangerous. It's essential that every device in your enterprise IoT deployment can be secured, can be update over-the-air, and that you actually keep them up to date. There can be no default accounts or passwords, communications must be robustly encrypted, and all exceptions (devices going off-line or behaving unexpectedly in any way) must be detected, investigated, and where there's legal liability, actions and resolutions logged.\n6. How long will the endpoint batteries last?\nWhen you have a constellation of any significant number of endpoints, you don't want to be replacing batteries all the time. Many of the leading enterprise IoT sensor endpoints have battery lives of five years or more (Monnit claims over 33 years for their Alta temperature sensor using two AA Lithium batteries) but even then, a proactive strategy is required unless you want to have to change, for example, 1,000 batteries over a short period. Endpoints should report their battery status, and there has to be the ability to raise an alert when a battery is too low (how low is situation dependent; a remote cold storage facility that's a day's drive away obviously needs to be serviced sooner than a walk-in cooler down the hall).\n7. Can you afford to lose measurements?\nIn any large installation over enough time, things happen; power outages occur, devices get unplugged, fail, get broken, or get stolen. Where there are legal and or health consequences should your organization fail to properly maintain, say, refrigerated medical supplies or foodstuffs, you'll want sensors with the ability to store measurements should your gateway go down. Then there's the possibility of your Internet connection going down so you'll want the gateway to have both backup power and adequate sensor measurement storage.\n8. What kind of sensor platform enclosures do you need?\nAn important consideration is that the packaging of the endpoint needs to be appropriate to the environmental conditions. For example, if the endpoint will be used in what's termed a "wash down" environment where cleaning is performed using high pressure water or steam, then there's an obvious need for enclosures with a high Ingress Protection rating. I also found that some vendors' sensor endpoints enclosures are surprisingly poorly designed making them difficult to reliably affix to, for example, wire racks or walls because they had no mounting tabs.\n9. Is the backend service stable and appropriate for your needs?\nMany of the vendors in the enterprise IoT market are evolving their products, and I've spotted a significant level of "bugginess" in many backend services. Two examples are corrupted measurement data with no explanation, and status reports that show gateways online when, in fact, they were powered down.\nAn important issue in regulated industries is whether the collected measurements are immutable; that is, whether they can be verified to be the original, unaltered data. As of writing, only a very few vendors have got to grips with this issue (expect to see blockchain putting in a big appearance in this area in the next few months).\nA feature that you'll definitely want to look for \u2013 unless the backend provides links to third party services\u00a0\u2013 is built-in analytics beyond the basics of average, maximum, and minimum values. When dealing with even moderately-sized enterprise IoT installations, the sheer volume of data makes it hard to spot trends and unexpected conditions, so you'll want more sophisticated event detection than just "send an alert when the temperature is greater than 40 degrees Fahrenheit." For example, statistical analysis of temperature fluctuations will reveal failing compressors, which in turn allows for proactive maintenance rather than having a potentially far more expensive catastrophic failure.\n10. Is there an API?\nFinally, the availability of an application program interface (API) in the backend services isn't just a nice idea, it's crucial to integrating your enterprise IoT infrastructure with other enterprise analytics and monitoring systems; skip this feature at your peril.\nThe bottom line\nSelecting e-IoT solutions is currently a surprisingly complicated and tortuous process, so if you're embarking on a large-scale enterprise IoT monitoring project, make sure you allocate enough time to dig deep, exhaustively test your top picks, and ensure that your solution doesn't lock you into a vendor path for longer than planned.