Why Almost Every Hardware Schedule Breaks

Ask ten hardware founders how their first schedule performed and you will hear the same shape of answer: the plan said nine months, the product shipped in nineteen. The overrun is rarely caused by one catastrophic event. It is caused by a planning method borrowed from software, where a task that takes three days takes three days, and applied to a domain where a task that takes three days has to happen four times because physical parts do not compile.

A hardware schedule that holds is built on four principles. None of them are exotic. All of them are routinely skipped.

Principle 1: Plan by Iterations, Not by Tasks

The typical bad Gantt chart contains a bar labeled "mechanical design, 6 weeks" followed by "prototype, 3 weeks" followed by "test, 2 weeks." That plan assumes the test passes. It never does the first time.

Plan instead in loops. One mechanical loop is design, fabricate, assemble, test, and redesign. For a 3D printed enclosure that loop is roughly two to three weeks. For a CNC machined part it is three to four. For a printed circuit board with assembly it is four to six weeks door to door, sometimes three if you pay for expedited fabrication. Then ask the real question: how many loops does this subsystem need? A well-understood bracket might need one. A new sensor front end, a sealing design, or a mechanism with a novel motion path will need three to five. Multiply.

Two loops of electronics and three of mechanics on a moderately complex product is already five to six months of calendar before you have a design worth freezing. Writing it that way in front of a board or an investor is uncomfortable once and accurate forever. The stage-gate vocabulary in EVT, DVT, and PVT explained gives each loop a name everyone recognizes, and the broader duration ranges are laid out in how long product development takes.

Principle 2: Identify Long-Lead Items on Day One

The critical path in hardware is almost never the design work. It is the items with fixed external lead times, and they are knowable at the start of the project if anyone bothers to look.

  • Injection mold fabrication: six to fourteen weeks depending on complexity and tool steel, plus two to four weeks of tryout and sampling. The variables are broken down in how long it takes to build an injection mold.
  • Semiconductors and connectors: anywhere from stock to 40-plus weeks. One 12-cent part on backorder stops a $200 product. Plan around it using component shortage planning.
  • Certification testing: lab scheduling alone is often four to eight weeks out, before any test time. A failed emissions run costs a full requalification cycle.
  • Custom items: LCD modules, custom cables, molded silicone, printed batteries, custom LiPo cells with their own transport testing.
  • Factory shutdowns: the annual lunar new year closure removes three to five weeks of usable production capacity and distorts the schedule for weeks either side, as described in the Chinese New Year factory shutdown.

Make a long-lead list in week one, put the quoted lead time next to each item, and schedule backward from the date you need it. If a decision has to be made before that item can be ordered, that decision is now on the critical path and deserves the attention the design work is getting.

Principle 3: Size Buffers by Uncertainty, Not by a Flat Percentage

Adding 20 percent to everything is a superstition. It over-pads the tasks you understand and under-pads the ones that will hurt you. Sort every work package into three bins.

Known work is something your team has done before with the same tools and vendors: a second revision of a familiar board, a bracket redesign. Buffer 10 to 15 percent.

Estimated work is familiar in kind but new in specifics: a new microcontroller family, a first enclosure with a gasket, a new supplier. Buffer 30 to 50 percent.

Genuinely uncertain work is anything where you cannot state the acceptance criterion yet: a novel sensing approach, a mechanism nobody has built, a material that has to survive an unproven environment. Do not buffer it as a task at all. Schedule it as a time-boxed investigation with a decision at the end: it works, it needs another box, or we take the fallback. Six weeks and a defined go/no-go is a plan. "Eight weeks to solve the hard problem" is a wish.

Hold buffer at the project level rather than hiding it inside each task, because task-level padding gets consumed automatically and teaches you nothing. A visible project buffer that shrinks tells you the truth about your progress. This is the same discipline described in product development project management.

Principle 4: Synchronize the Parallel Tracks

Mechanical, electronics, firmware, app, industrial design, packaging, and certification all run at once, and they exchange information at specific moments. Those handoff points, not the tracks themselves, are where schedules quietly break.

Name each dependency explicitly and put a date on it: board outline and connector positions to mechanical; final enclosure volume back to electronics for thermal and antenna; a working board to firmware; a firmware build that can run a functional test to production engineering; a final BOM to purchasing; final artwork to packaging. When mechanical slips two weeks, you can then see immediately that certification slips two weeks, because the plastic housing affects the wireless test.

The practical rule: no track should be waiting on a deliverable whose date is not written down and owned by a named person. And no track should be idle waiting for perfection. Firmware can develop against a development board while the custom hardware is fabricated; app work can proceed against a simulated device.

What to Do When the Plan Moves Anyway

It will. The question is what you do in the first week after you know, not in the fourth.

Reforecast honestly and immediately. A three-week slip disclosed in week ten is a manageable conversation; the same slip disclosed in week twenty-two, when it has grown to nine weeks, is a credibility event. Decide explicitly whether you are protecting scope, cost, or date, because you cannot protect all three, and the cheapest lever is almost always scope. Cutting a secondary feature out of version one preserves both the launch window and the budget, which is the argument behind running a pilot production run before committing to full volume.

Finally, treat late design changes as the expensive events they are. A change during design costs engineering hours. The same change after tooling costs a tool modification and weeks of calendar, which is why a formal engineering change order process is worth the overhead as soon as your first mold is cut.

Get a Schedule You Can Show a Board

Projects House builds hardware development schedules with iteration counts, long-lead ordering points, and gate criteria that survive contact with a real factory. Send us your product scope and target launch through our contact form and we will tell you what is achievable.