In this guide 6 sections
Pick one chamber, pump house or machine. Write the pass criteria before you buy anything. Run the sensors beside your paper log across shifts, compare each reading with the gauge on the asset, and roll out only when the shift team agrees.
Pick the one asset that will teach you most
Choose a single walk-in chamber, one borewell pump house or one injection moulding machine, and leave the rest of the floor alone for now. Five sites blur every fault. Keep it small.
Look for an asset where a bad reading already costs you something, like a chamber that drifts overnight, a pump that trips unannounced or a press that stalls unrecorded. The pilot should answer a live question.
Walk the asset with the supervisor who owns it and note the existing gauge, the clipboard log, who writes in it and how often an entry gets skipped. Visit during a running shift, because a quiet weekend hides the skipped entries, the gauge nobody trusts and the habits that decide whether the pilot data means anything.
- One asset, one owner, one shift lead
- An existing gauge or meter to compare against
- Mains power and a signal path within reach
- A problem the team already complains about
Write the pass criteria before you buy anything
Put the test on one page and have the plant head sign it before any probe, gateway or cable is ordered, so nobody can redefine success later. Otherwise the pilot ends in argument. Nobody agrees what worked.
Criteria must be measurable. The sensor reading stays within a stated tolerance of the reference gauge, and the dashboard shows no data gap beyond a set interval of minutes. Alerts reach the named person in time.
Agree what counts as a fail too. A pilot that cannot fail teaches nothing. If a probe drops out twice in one week, record it as a finding, not a footnote, then have the electrician recheck the mounting and cable run. Add one line for the people side: the night supervisor opens the right screen unaided, in under a minute, while the day team watches and takes notes. That test costs nothing. It often decides adoption.
Run the system beside the manual log
Keep the clipboard. Do not retire it on day one. The paper register carries on exactly as before while the new probes, the gateway and the screen run in parallel beside it.
Now you hold two records. When they disagree, you learn something useful: the operator rounds the figure, the probe sits beside the door or somebody filled the log from memory at shift end.
Run it across several shifts, including nights, a changeover from one product to the next and one stretch of planned maintenance, so the data covers the awkward hours. A quiet Tuesday proves little.
- Paper register continues unchanged
- Probe, gateway and dashboard live from day one
- One person checks gaps in data each morning
- Night shift and changeover both included
Compare every reading with the existing gauge
Once a shift, stand at the asset, read the analogue dial or handheld thermometer, and write the value beside the dashboard reading for that same minute. A week gives you a paired table. Short, and honest.
Study the difference, not the average. A probe that reads steady but sits two degrees from the reference needs an offset, while a probe that wanders needs a better mounting position or a screened cable. Log which fix you chose.
Also test the failure paths. Pull the network cable. Does the gateway keep recording? Then leave a door open and watch who receives the alert, how quickly it arrives and whether anyone acknowledges it before the shift ends.
Decide on rollout from the evidence
Take the pass page and tick it line by line, giving every failed line a cause, a fix and a second short run on the same asset. Widen nothing until it passes.
If it passes, list what the next five assets need: probe types, cable lengths, gateway positions and alert recipients, so quotes rest on facts, not guesses.
Your next step: book thirty minutes with the plant head this week and fill in the pass criteria page for one named asset.
- Criteria ticked, failures explained
- Operators comfortable with the screen
- Alert routing confirmed by a real test
- A parts and site list for the next assets
IoT pilot pass sheet
One page, signed before the pilot starts. The top block names the asset, its owner, the reference gauge and the shift lead who checks the data gap list each morning. The middle grid lists each criterion with its target, the method of measuring it, the result and a pass or fail tick. The bottom block logs paired readings by date, shift, gauge value, dashboard value and the difference between them, plus a signature line for the plant head.
Open a blank worksheet to printQuestions owners ask
How long should an IoT pilot run?
Long enough to cover every shift and at least one changeover or full production cycle, which for most plants means a few weeks, not a few days. The pass sheet fixes the minimum in advance, so early good numbers cannot shorten the test, whatever the enthusiasm of the sponsor.
Which asset is best for a first pilot?
Choose one with a known problem, a reference gauge to compare against and an owner who wants the answer, since that combination produces a clear verdict inside weeks. A drifting chamber teaches more.
Should the paper log stop during the pilot?
No. Keep it running until the pilot passes. The paper record is your comparison baseline, and it protects the plant on any shift when the new system loses its network link or a probe drops out.
What if the pilot fails?
Write down why. A failure traced to probe position, signal path or alert routing is a useful result, because it tells you what a wider rollout would have repeated. Fix the cause. Repeat on the same asset.