When customers ask us how to accelerate IoT product development, we usually start with an uncomfortable observation. The average OEM now takes 41 months to bring a connected product to market, up from 23 months four years earlier. That is an 80% increase in time-to-market for an industry that, on paper, has more low-code tools, more cloud services, and more reference architectures than ever before.
In our deployments across system integrators and product teams, we have seen the same pattern repeatedly. Teams adopt better tools, vendor pitches promise 50-90% reduction in development time, and somehow the calendar still expands. The question worth asking is not how to fix the tools. It is whether “faster” was ever the right target.
The Time-to-Market We Were Sold
The vendor narrative is consistent and supported by named research. Forrester reports a 50-90% reduction in initial development time when teams move to low-code platforms. Gartner forecasts that 70% of new enterprise applications will use low-code or no-code by the end of 2026. Vendor case studies, such as OutSystems’ composite organization analysis, cite payback periods under six months and three-year ROI figures above 300%.
These numbers are not wrong. They are also not the full picture.
The IoT Analytics survey of 100 senior OEM executives breaks the 41-month average into two stages. Eighteen and a half months go from project kick-off to a working proof of concept. Another 22.8 months pass before the first paying customer. Even teams using mature platforms now spend nearly two years between PoC and paid revenue.
Recap: the development phase got faster. The product cycle got longer.
Why “Faster” Is the Wrong Target
Four forces have stretched IoT timelines simultaneously, and none of them respond to faster IDEs.
The first is project complexity itself. Modern IoT products integrate edge processing, AI inference, legacy industrial systems, and consumer-grade UI in a single deployment. The components multiplied. The integration surface grew with them.
The second is regulation. NIS2 in the EU, the Digital Product Passport, the US Cyber Trust Mark, and the EU Data Act all introduced compliance work that did not exist five years ago. Each adds testing cycles, documentation requirements, and architecture review checkpoints.
The third is security scrutiny. After the Verkada breach in 2021 and the earlier Ring incidents, enterprise buyers tightened their security audit standards. As one senior automotive manufacturing manager put it, “the major challenge was aligning our IT security standards with the security provided by vendors.” That alignment work is now part of every meaningful deployment, not a final-mile checklist.
The fourth is the use cases themselves. Customers no longer accept dashboards as the deliverable. They want workflow optimization, predictive logic, and operational integration. IoT Analytics found that 61% of companies with successful connected products, defined by amortization periods under 24 months, cite workflow optimization as crucial customer value. Only 21% of less successful companies say the same.
The pattern is clear. Extended product timelines do not signal failure. They signal that the work changed. Treating “shorter time-to-market” as the success metric for an IoT product in 2026 is like measuring a software team by lines of code.
When teams come to us asking how to avoid building a platform from scratch, we walk them through the same framework IoT Analytics documented across hundreds of organizations. There are not two options. There are three, and the one most teams default to is statistically the worst on time-to-value.
Build from scratch means the customer constructs the technology stack internally, sometimes with external services, sometimes purchasing infrastructure components. Forty-seven percent of surveyed projects took this path. The median time to develop the business case alone is nine months.
Buy a complete solution is the rarest of the three approaches, taken by 30% of projects. The median business case develops in six months. Median break-even arrives at twelve months, almost twice as fast as build.
Buy-and-Integrate sits in the middle. The team purchases pre-built components or software that requires moderate modification. Thirty-eight percent of projects take this path. It looks like the best of both worlds: vendor leverage on the boring parts, custom logic on the differentiated parts.
Here is the counter-intuitive finding from the same survey. Buy-and-integrate has the longest break-even of the three approaches despite the fast initial implementation. The reason is consistent across our deployments. Thirty percent of “ready” platforms turn out to miss capabilities the team needs in production. The integration work to close those gaps takes longer than the original build estimate, and it produces a system that no one fully owns.
A US pharmaceuticals chief experience officer summarized the trade-off honestly. “We did not have adequate in-house knowledge in IT or operations to consider building from scratch. We did not want to be entirely dependent on third parties for maintenance. A tailored solution was essential to ensure operator confidence, so a hybrid model was best.”
This works when the capability gap is small and the team is honest about its boundary. It fails when teams enter a hybrid hoping the platform covers more than it does.
Rule of thumb: if more than 30% of your top user stories require custom development on top of the platform, you are not buying a platform. You are building one with extra steps.
When Building From Scratch IS the Right Call
We do not pretend platforms always win. Building IS justified in three specific scenarios.
First, when the product covers a very narrow vertical with a small set of device types. Buildings industry teams chose build approaches in 89% of IoT Analytics survey responses. Highly customized facility-specific requirements make standard platforms a poor fit.
Second, when the team has a large in-house engineering organization willing to maintain protocols, security, and scaling permanently. The Hacker News community has been clear about this trade-off. One developer wrote in the Google IoT Core discontinuation thread: “I would not have depended on any cloud IoT product. Use VMs, blob storage, containers instead.” That is a rational position for a team with the engineering depth to operate the substrate itself.
Third, when the platform IS the differentiation. This is rare. Most IoT products differentiate on domain logic, UX, and industry-specific integrations, not on having a unique device-management substrate.
The antipattern we see most often is teams that pick build for the wrong reason: “we are an agile team, we will refactor later if we need to.” Later does not come. After MVP, paying customers appear. With them come contracts, uptime expectations, and the fear of touching production systems. Refactoring an IoT product stack is not a sprint. It is a multi-month risk exercise that surprises management with costs no one budgeted.
The Question That Survives Years – Will the Platform?
Even if the build-versus-platform decision goes in favor of a platform, the next question is rarely asked early enough. Will this platform exist in five years?
The recent record is not encouraging. Google IoT Core was discontinued in August 2023. IBM Watson IoT shut down in December of the same year. Microsoft has restructured Azure IoT Central in a platform rethink, signaling a shift in its IoT-as-a-service commitment. AWS IoT Core remains active, but the wider hyperscaler track record on IoT-as-a-service is now a documented risk factor for any team selecting a platform expected to outlive a decade.
A Hacker News thread on Google IoT Core surfaced a comment that resonated with 215 upvotes, citing Steve Yegge’s framing: deprecation means “we are breaking our commitments to you.” For teams that built on the service, the operational consequences were concrete. The PrintNanny.ai team migrated infrastructure in 20 hours after spending 1.5 years on IoT Core. They reported “hundreds of hours implementing and debugging glue between GCP’s Pub/Sub product, websocket-based subscribers, and MQTT subscribers” before the shutdown forced the move. That was for a small fleet. A 50,000-device migration runs into six or seven figures of unplanned engineering work.
Flexera’s State of the Cloud 2026 survey shows that 89% of enterprise organizations now run multi-cloud strategies, and 42% cite vendor lock-in prevention as the primary reason. The market has internalized the lesson. Teams selecting an IoT platform should ask three questions before commercial discussions begin.
Will this platform exist in five years? This is not a wish. It requires evidence. The platform must be the vendor’s core business, not a side bet. Look for seven or more years of production deployments, dozens of customer references across multiple industries, and platform releases that are substantive rather than cosmetic.
Is there a documented exit path? Software escrow with a trusted third party, contractual transfer of source code on insolvency or discontinuation, and clear protocols for migration assistance. These belong in the contract before signature, not in a polite vendor email after a shutdown announcement.
Can you export your code as machine-readable artifacts? If your business logic lives only inside visual editors that have no XML or JSON export, you do not own your code. You own a license to view it. The only intellectual property that survives a vendor change is exportable, version-controllable configuration.
What a Reusable IoT Platform Must Actually Deliver
Once survivability is established, the next filter is fit. Steven Hilton, co-founder and president of MachNation, has documented four enterprise requirements for any IoT application enablement platform: developer usability, flexible deployment models, operational sophistication, and a partner ecosystem. Each is a yes or no question, not a sliding scale.
Developer usability means the platform was designed for the IoT developer’s workflow. It is testable. Take your top three user stories and prototype them in a one-week vendor trial. If your engineers struggle with the basics, no amount of training will fix it.
Flexible deployment means cloud, on-premise, and edge. Not one. Not two. Industrial, telco, and government deployments routinely require all three modes in the same product. Platforms that support only cloud-native workflows will lose enterprise opportunities.
Operational sophistication is platform management plus device management plus monitoring. The MachNation IoT Test Environment found that platforms range from 10 to 181 minutes for identical operational tasks. The variance is not theoretical. It compounds across thousands of devices and hundreds of operations per day.
The partner ecosystem matters because no single vendor implements every protocol, every integration, every regional requirement. A working partner network multiplies your reach. An empty one constrains it.
A fifth requirement has emerged in 2026 and is becoming non-negotiable: AI-readiness. AI agents and edge inference are now part of every serious IoT roadmap, and a reusable platform should expose an MCP server or equivalent agent-friendly interface – not as a buzzword, but as a structural requirement for the next product cycle. Platforms that treat AI as a marketing layer rather than an architectural one will fall behind quickly.
Beyond these five, a serious reusable platform must also provide multi-tenancy designed from day one rather than retrofitted, a unified data model that normalizes inputs across protocols, and machine-readable code export for every artifact your team creates.
For a 20-point evaluation framework that goes deeper into protocol coverage, scaling architecture, and partner program quality, see the [IoT platform selection checklist we maintain for system integrators and engineering companies.
A Decision Framework Without the Vendor Pitch
We end most platform conversations with the same four-step framework. It avoids the trap of comparing vendor brochures and forces the team to think in long-term product economics.
Step one: draw the absorb-versus-differentiate line. List the capabilities your product needs, then sort each into “core differentiator” or “infrastructure.” Multi-tenancy is infrastructure. Role-based access control is infrastructure. Time-series ingestion is infrastructure. Your industry-specific workflow logic is differentiation. If 70% of your list is infrastructure, building from scratch means your team will spend most of its time on undifferentiated heavy lifting.
Step two: apply the three survivability questions to every shortlisted platform. If any one returns no, the platform is out. Vendor abandonment risk is not negotiable when your product runs for ten years.
Step three: run the 30% capability test. Take your top thirty user stories. Count how many the platform delivers out of the box. If less than 70% pass, you are entering the hybrid trap. Either pick a more capable platform, or accept that you are building, just with extra dependencies.
Step four: demand a contractual exit path. Source escrow with a defined release condition. Machine-readable export for all configurations. Contractual migration assistance if the vendor changes ownership. None of these are optional clauses for an IoT product expected to live a decade.
Iotellect is one platform designed around exactly this set of survivability and exit criteria, offering a practical alternative to building from scratch for system integrators and product teams who think in long lifecycles.
Faster is the wrong target for IoT products. Absorbing complexity, and surviving the platforms that promise to absorb it for you, is the real one.
Discover more from WikiTechLibrary
Subscribe to get the latest posts sent to your email.
