Skip to content
See packages

Delivery

Hardware pilot projects: scope one to buy a decision, not a demo

Scope a pilot backward from the go or no-go it must produce. A demo that works under supervision settles nothing about the rollout.

Computing America Hardware. Custom hardware for physical automation and monitoring.

Describe the decision
4 min readComputing America
One machine with a neatly installed sensor in front of a long row of machines with none.

In short

  • If the scope can't say what a no looks like, it isn't a pilot. A phase commissioned to carry a budget request is a legitimate purchase, but a different one.
  • Four common unknowns stop rollouts: measurement validity, environment, connectivity and support load. Test the one or two that block yours.
  • Run it where it's hard. A pilot in the clean end of the plant tells you about the clean end of the plant.
  • Success criteria are numbers fixed before the phase starts: a bound on agreement with the reference, and an availability figure with "unavailable" defined.
  • Budget the removal. A pilot with no plan for taking the hardware out becomes production by default, on hardware chosen for a trial.

Name the number that would change your mind about the rollout, then install only the hardware that measures it. Most of the scoping work is choosing the one or two unknowns that block the rollout and leaving the rest for later.

A pilot can succeed and still decide nothing. Devices get installed, data appears, someone demonstrates a screen in a meeting, and the phase is written up as a win. Then the rollout waits for a second pilot, and the team can't say what changed in between. The pilot answered a question, and it wasn't the one blocking the decision.

A pilot proves feasibility only in the conditions of the pilot. The organization needs to know whether the thing works at three hundred units, unattended.

Write the decision first

Write one sentence naming the decision the phase exists to enable: at the end of this we will commit to X, or we will not, and the thing that decides it is Y.

Writing it makes a no explicit, which is why it gets skipped. A pilot commissioned by someone who has already decided, to carry a budget request, is a legitimate use of a phase, but it's a different thing to buy and it changes what should be built. If the scope can't say what a no looks like, it isn't a pilot.

Four unknowns worth a phase

UnknownThe question underneathWhat a phase must do to settle it
Measurement validityDoes the sensing see the phenomenon well enough to support the decision?Run against an independent reference long enough to characterize the disagreement, including during the events the system exists to catch.
EnvironmentDoes the hardware survive the place, in the state the place is actually in?Install in the worst location in scope, through one full cycle of whatever the site does to things: a washdown, a seasonal swing, a shutdown and restart.
ConnectivityDoes the data get out from where the devices have to be?Survey signal on site rather than from the drawing, and measure at the far unit in the far building, on the network the operation will use.
Support loadWhat will owning three hundred of these cost per year?Count every intervention, including the casual ones and your own team's: the power-cycle, the re-seated connector, the restarted service that took four minutes because the person was already there. At fleet scale each becomes a dispatch, so that number, scaled, is the operating cost.
What commonly stops a rollout, and what settles each one

Run it where it's hard

Pressure to run where it's easy is constant and reasonable: the accessible machine, the new building, the shift with the engaged supervisor. It makes the phase cheaper and the result worth less, because the rollout will reach the far end of the yard, the machine from 1994 with no spare terminals in the panel, and the shift that wasn't consulted.

Criteria are numbers, agreed before

After the fact, every pilot can be narrated as promising: the readings mostly tracked, the outages were mostly explicable, the operators mostly liked it. None of that supports a decision. Fix the criteria in advance:

  • Agreement with the reference as a bound: within this much, this often, including during the events that matter. A correlation isn't enough. Illustrative example: within 2% of the reference on 95% of readings, and on every reading during an event over the alarm threshold.
  • Availability as a fraction of the period, with "unavailable" defined. A device reporting a stale reading isn't available and will otherwise be counted as up. Illustrative example: at least 98% of hours in the pilot, where a reading with a timestamp more than 15 minutes old counts as unavailable.
  • Interventions per device per month, with no exemption for whoever happened to be passing. Illustrative example: no more than 0.5 interventions per device per month.
  • Time to recover a failed unit without a site visit, measured on a unit you failed on purpose. Illustrative example: remote recovery within 30 minutes.
  • One human criterion, measured by watching people work: did the work that was meant to move to the system move there, or does the old process survive beside it?

Technical specifications tend to leave the last one out. A deployment where the old record survives beside the new system has failed whatever the uptime figure says, and you learn that by standing on the floor at shift change.

Fig. 1

interventions a month

  • One device0.5
  • The first ten units5
  • Three hundred units150

An intervention rate of 0.5 per device per month sounds small, and across the three hundred units a rollout reaches it is 150 dispatches a month.

SourceThe illustrative criterion in this article, 0.5 interventions per device per month, scaled to one device, ten units and three hundred units.

Budget the removal

Someone has to own taking the hardware out, restoring what it displaced, and deciding what happens to the data, because a build chosen for a trial has no support plan.

When a pilot is the wrong tool

A phase that tries to settle every unknown at once is a rollout with a smaller budget, and the deliberation has a cost: while criteria are negotiated, the plant still runs with the problem. We'd keep a phase to two unknowns, and choosing which two is the work. If comparable hardware already runs elsewhere in your estate, in the same conditions, skip the pilot and roll out with the first ten units watched closely.

Next step

Tell us what you would have to know to say yes.

Describe the rollout you are considering and the thing that is stopping you committing to it. We will tell you what a first phase would have to measure to settle that specific question, and roughly what shape and size it would be.

Reply
A person replies, not a sequence: within one business day, from someone who would be on the engagement.

Packages

Start with a package

Four packages, each with its price on the page. Pick one and buy it online.

Billing