The Moment You Lose Control of the Hardware

In the lab you control temperature, power, the operator, and the sequence. In the field you control none of it. The unit gets rained on, dropped off a tailgate, plugged into a generator with poor voltage regulation, operated by someone who did not read the instructions, and left in a truck bed overnight at 20 degrees F. That is the point: field testing exists to find the failures your lab cannot generate. It is also the most expensive testing you will do per data point, since a day in the field costs travel, a technician, a customer's time, and real risk to the relationship if the unit embarrasses itself. Preparation is the whole game.

Start With the Question, Not the Hardware

Write down, before anything is packed, what the test is supposed to answer. Not "see how it performs" but: does the sensor hold plus or minus 2 percent across 12 hours in an unconditioned warehouse; can an untrained operator finish setup in under 10 minutes without calling support; does the seal survive a washdown; does the battery finish a shift with 20 percent reserve.

Then define the pass criterion for each. A field test without written criteria always ends with everyone agreeing it "basically worked," which is how a product goes to tooling with an unresolved problem in it. Choose the site to match the question too: if humidity is the risk, go somewhere humid, and do not test dust ingress in a clean facility because it was easy to arrange. Recruiting the right sites and users is the same problem as building a good hardware beta program, just earlier and with fewer units.

Hardening: What Breaks First Outside

A lab prototype and a field prototype are different builds, and the changes are cheap before you leave and impossible three hours away.

  • Connectors and fasteners. Connectors are the number one field failure: replace headers and jumper wires with locking, keyed parts, because anything vibration can unplug it will. Add thread locker or nylon-insert nuts everywhere, since vibration backs out screws in hours.
  • Enclosure and sealing. The lab unit with an open panel needs a real cover, gasket, and cable glands. Set the target rating honestly using IP ratings explained, then verify it before you go with the methods in building a waterproof prototype. Do not learn about your seal from the weather.
  • Strain relief and temperature. Every cable exit needs relief, because cables get yanked. Verify operation at both temperature extremes before shipping: LCDs go sluggish in the cold, adhesives creep in the heat, lithium cells refuse to charge below freezing.
  • Handling. A padded case with foam cutouts. Most units that arrive broken were broken in transit, not in use.

Power and Electrical Reality

Field power is worse than bench power in every dimension. Generator output sags and spikes, vehicle 12 V rails see load dumps and cold-crank dropouts, extension cords add voltage drop, and grounding is inconsistent or absent. A prototype that has only ever run from a regulated bench supply has never been tested.

Before you go, run the unit from the actual power source it will see, add input protection and a brownout-tolerant reset, and confirm what happens when power is yanked mid-operation. A device that corrupts its configuration on abrupt power loss will do it in front of the customer. If it is battery powered, measure real consumption over a real duty cycle instead of trusting the calculated budget, and bring twice the capacity you think you need; the failure modes are covered in battery pack design.

Instrumentation: No Measurement, No Test

The most common field test failure is coming home with impressions instead of data: something went wrong at some point in the afternoon and nobody can reconstruct what.

Log everything to onboard storage with timestamps: sensor values, motor currents, supply voltage, internal temperature, state transitions, error codes, user actions. Buffer locally and sync when connectivity allows, because the cellular link will not exist in a basement. Add a physical event-marker button so an operator can flag the moment something felt wrong, which lets you find the interesting five seconds inside twelve hours of log. Bring independent instrumentation as a reference when you suspect the unit's own sensors, and take video, which is worth more than any written description of a failure.

Spares, Field Kit, and Safety

Bring at least two units if the budget allows, because one failure should not end the test. The field kit is spare connectors and pre-made cable assemblies, fuses, fasteners, thread locker, gaskets, tape, zip ties, a laptop with the toolchain and a known-good firmware image verified offline, a debug adapter, hand tools, and a multimeter. Assume no internet and download everything, including documentation you think you will not need.

On safety: an unfinished prototype in a customer environment needs a real risk assessment. Exposed voltages, pinch points, hot surfaces, and stored energy need guarding or clear labeling, someone on site needs authority to stop the test, and site access and insurance rules should be confirmed in writing before you arrive.

Prepare the People, Not Just the Equipment

The human side is half the result. Brief the site staff on what the unit is, what it is not, and what to do if it misbehaves, with a one-page quick reference rather than a manual. Tell them explicitly that finding problems is the goal, because polite users hide failures to be helpful and you lose the data. Set expectations with the customer in writing too: test unit, here is what it will not do yet, here is the duration and who to call.

Then watch quietly and do not coach. The instinct to help a user through a confusing step destroys the exact finding you came for. The observation discipline is the same as in user testing with a prototype, and harder in the field because you are standing right there.

Debrief the Same Day

Before you drive home, write down every observation, intervention, workaround, and question the user asked. Memory degrades fast and the small details are the valuable ones. Note not just what failed but what the operator did instead, because their workaround usually points at the real design fix. Then classify: design changes, documentation or training issues, prototype build artifacts production would not have, and things genuinely out of scope. That triage converts a chaotic day into a work list.

From Field Test to Pilot

A field test with two units and a supervising engineer is not a pilot. The pilot comes after the findings are closed out: more units, longer duration, less supervision, and metrics like uptime and interventions per hundred hours. It feeds directly into the build described in a pilot production run. Expect at least two field rounds; the first finds the obvious things and the second finds what only appears once the obvious ones are gone.

Plan the Field Test Before You Build the Unit

Projects House builds field-capable prototypes with hardening, logging, and a written test plan, and supports the trials themselves. Tell us the environment, the users, and the questions you need answered through our contact form and we will scope the build and the test together.