Choosing an IoT Platform for Your Plant: 8 Technical Criteria

Failed IoT projects rarely fail in year one. Dashboards light up, sensors report, everyone is satisfied. The trouble arrives around month 18: you want to add a different sensor brand and it isn't supported, you want to pull three years of history but retention is 90 days, or you want to change one widget and are asked to raise a new purchase order. By then, migrating to another platform usually costs more than the entire original project.

The eight points below are what actually decide the outcome — and every one of them can be tested before you sign.

1. Ingest protocols: open or closed?

This comes first because its consequences are the most permanent. A platform that only accepts its own hardware locks you to one price list indefinitely. Three paths need to exist:

• MQTT — the de facto standard for modern IoT devices; lightweight and tolerant of poor networks.
• HTTP/REST API — for pushing data from other systems, PLC gateways, or in-house scripts.
• Modbus RTU/TCP — the path that keeps existing power meters, inverters, and legacy PLCs usable.

For cameras, ask about ONVIF. If the answer is "we'll build a custom integration", every new device becomes a billable project.

2. Who is allowed to change a dashboard?

Operational needs shift every quarter. If every added chart must go through the vendor, two things happen: change becomes slow, and the floor team stops asking — then quietly returns to Excel. Ask for a live demo: build one dashboard from scratch, in front of you, by someone non-technical. If it takes more than ten minutes, you have your answer.

3. Retention: how long is raw data kept?

Many platforms store aggregates (hourly averages) and discard raw data after a few weeks. For operational charts that is fine. For root-cause analysis and predictive models it is fatal — most failure patterns are only visible at second or minute resolution. Ask for two separate numbers: raw retention, and aggregate retention.

4. Automation: can the platform act?

Charts save nothing; actions do. Test whether the platform can run a rule such as "if pressure drops below X for 5 minutes, notify the technician group on WhatsApp and log the event". Check for delays and hysteresis too — without them, one sensor chattering at a threshold sends 200 notifications overnight and the whole system gets muted.

5. Intelligence on top of the data

Fixed thresholds only catch problems you already anticipated. Anomaly detection catches the ones you didn't — a combination of vibration, temperature, and current where each value is individually normal but the combination is not. This is what separates a plain IoT platform from an AIoT platform; the distinction is covered in full in our guide to IoT & AIoT platforms in Indonesia.

6. Data isolation, permissions, and audit trail

Three concrete questions: is data genuinely separated between organisations; can permissions be scoped per site or per line rather than just "admin" and "viewer"; and is every configuration change recorded with who and when. The third is routinely ignored until an incident occurs and nobody can say when an alarm threshold was changed.

7. The on-premise option

Some sectors cannot place operational data in a public cloud. Even if you can today, policy can change. A platform with no on-premise path means a policy change equals a full migration.

8. Local support and billing currency

Two things absent from any spec sheet, yet most felt in daily operation. A 2 a.m. incident needs an engineer answering in the same language, in the same time zone. And dollar invoices make operating budgets move with the exchange rate — a real problem for teams that set annual plans a year ahead.

Test it, don't just ask about it

Verbal answers are cheap. Ask for proof during a two-week trial with a small but genuine scope:

• Connect one third-party device you already own — not the vendor's demo kit.
• Have your own operations staff, not a consultant, build one dashboard.
• Create one automation rule and deliberately trigger its condition.
• Export a week of raw data and open it in Excel. If exporting is hard, the data isn't really yours.

Those four steps reveal more than a three-hour presentation.

How INCLUDE answers these eight points

IncludeApps — the core platform of INCLUDE Smart Industry — ingests data over MQTT and an HTTP OpenAPI, plus ONVIF for cameras, so cross-vendor equipment already installed does not need replacing. Dashboards are assembled drag-and-drop by your own team, and device quotas and retention are stated openly on the pricing page. For legacy serial equipment, IncludeGateways bridges Modbus into the platform. The Enterprise tier offers on-premise deployment for organisations required to keep data on their own infrastructure.