Why Doesn't Your Prototype Battery Pack Lead Time Match Mass Production?
A prototype battery pack lead time and a mass production lead time are not the same number measured on two scales. They are two separate clocks, started by different events, counting different work. The prototype clock is short because a prototype is a handful of packs built by people who already know how to build them, on fixtures they made by hand, with nothing to prove and nobody asking for evidence. The production clock is long because only a small slice of it is spent building anything. The rest is design freeze, tooling release, process qualification, end-of-line test setup, traceability, and then waiting in a queue for a slot on a line that is already committed to other programs. Assume the prototype timeline simply scales up and the launch plan will miss by months, not days.
Walk with me to the end of a sample bench. It is a corner of the floor the supplier did not have to reserve for you. An ESD mat, a hand-fed spot welder, a stack of cells in a grey bin, a BMS board sitting in a plastic tray, a fixture somebody machined last week out of scrap aluminum because the real one does not exist yet. One technician, one torque driver, a printed traveler on a clipboard. The supervisor tells you twenty units by Friday and you believe them, because you have watched this exact scene play out the same way at a handful of other suppliers.
Now walk forty feet, through the fire door, to the other side of the wall. Same building. A conveyor, a nest that will hold the pack at a fixed angle, a weld head that moves on a servo, an end-of-line rack with coiled test cables and a power supply that is not switched on. Taped to the machine is a commissioning checklist with three initials on it. That line is going to build your production units, and right now it cannot build anything.
The distance between those two rooms is not a factor of volume. It is a list of gates, and each one owns its own stretch of calendar.
Why the prototype clock is short
It is short for reasons that evaporate the moment you need repeatability.
Prototype labor is engineering labor. A senior technician compensates in real time for a cell that sits slightly off location, a weld that looks cold, a connector that needs a shim. Nothing has to be written down, because the person who knows is standing right there.
Prototype tooling is soft tooling or no tooling. Machined fixtures, printed jigs, hand-loaded nests. They exist to make ten units, not to hold tolerance across thousands.
Prototype units do not queue. A handful of packs walk onto a bench the same week they are asked for. A few thousand packs need a slot in a production schedule, and that schedule was filled months ago by other customers.
And prototype quality is judged by whether the pack works, not by whether the process that made it works. That distinction sounds academic until you are the one who has to prove the second one.
I have watched teams treat a fast prototype as evidence that the supplier is fast. It is not evidence of anything except that the supplier can build a small number of things by hand. The useful read is narrower and more honest: the prototype tells you the design is buildable. It says nothing about whether it is buildable at rate.
The gates that actually eat the calendar
These are the places where weeks disappear, and they disappear in roughly the same order every time.
Design freeze. Everything downstream is quoted, ordered, and scheduled against a frozen design. A change after tooling release does not just cost money; it resets a clock. If you are still iterating on an enclosure boss or busbar geometry while the tool shop is cutting steel, you have already spent the tooling lead time twice.
Tooling release and first article. Injection molds for the housing, weld fixtures, potting fixtures, the nest that holds the pack in the line. These get ordered after freeze, not before, and they are long-lead. When first article comes back it gets measured against the drawing, and if it is out, the correction loop starts over.
Cell allocation. Prototype cells come out of a distributor's stock or a sample program. Production cells come off a plan. If the cell you designed around is on allocation, your production start date is not yours to set.
Process qualification and the approval gate. Somewhere in here sits the formal moment where the supplier demonstrates that the process, not just the pack, meets requirements. Call it PPAP or call it whatever your supplier calls it. The calendar cost is the same, because this is a milestone with evidence attached, and evidence takes time to accumulate.
End-of-line test setup. This one gets underestimated constantly. You need a test fixture. You need limits derived from engineering data rather than from guesses. You need the fixture correlated against the bench measurements you trusted during prototyping. And you need a system that stores the result against a serial number. Building the test station is one project. Writing and proving the limits is a second one.
Production records and traceability. Serialization scheme, traveler, lot genealogy for cells and boards, retention rules. Somebody has to design it, and it has to be running before the first shippable unit, not after.
Regulatory and transport testing. UN 38.3 and whatever else the destination requires runs on its own timetable, and it is gated on having representative production cells. You cannot start it early with prototype cells and have it count.
None of these are manufacturing. All of them are on the critical path.
What changes when the bench becomes a line
Here is the counterintuitive part, and the one I have had to explain most often to program managers: the first automated run is slower than the manual build it replaces.
Manual assembly has a human absorbing variation. Automation has a fixture that either locates the part or does not. Cycle time is not the constraint at first. Station balance is, fixture repeatability is, the weld window is, and yield is. You will spend weeks finding the window where the weld is hot enough to be reliable and cool enough not to damage the cell, and that window may turn out narrower than the manual process ever revealed, because the manual process was quietly absorbing the spread.
Most teams I have worked with do not jump straight to full automation. They go semi-automated: automated welding or automated test, manual load and unload. That is usually the right call, and it still does not remove the calendar cost. It moves it. You trade fixture complexity for operator-dependent variation, and you still have to prove the whole line.
Getting a production number out of a pilot build
A pilot build is the most useful data you will get before production, and it is routinely misread.
The pilot rate is a measurement, not a multiplier. Seeing a couple dozen packs a week come off a pilot cell does not mean a much larger weekly volume is a proportional number of weeks away. What the pilot actually gives you is a parts list of truth: the real process steps, the real station count, how long end-of-line test actually takes per unit, how much rework actually happens.
From there, the honest production estimate comes from the slowest step, not the average step. If end-of-line test takes a fixed number of minutes per pack and you have one test station, that station sets your ceiling no matter how fast everything else runs. Add the queue time in front of it and you have the real constraint.
Put a range in front of people, not a point, and attach the assumption to it out loud. "Production start is four to six weeks after tooling first article is approved and the test fixture is correlated" is a sentence you can defend. "We will be shipping in Q3" is a sentence that will be quoted back to you.
Saying it out loud to people who want one date
The habit that has saved me the most grief is this: never give a phase-flat date. Give a gate, a duration, and a trigger.
Prototype date: driven by design maturity, quoted in weeks from the day the design stops moving.
Pilot build: driven by whether soft tooling and fixtures are ready, and by whether the cell sample order actually landed.
Production: driven by design freeze plus tooling plus qualification plus test setup, and it usually is not compressible by adding people.
Keep two production dates separate, because they are different milestones. First shippable unit is one. Stable rate, where the line is running consistently and yield has settled, is another. Everyone remembers the first one and plans revenue against the second one. Naming both of them early is the whole game.
Some things genuinely do overlap. You can duplicate a fixture and qualify two stations at once. You can run pilot builds while tooling is being cut, if the tooling is not on the critical dimension. You can prepare the traceability system in parallel with everything else, because it is software and paperwork. What you cannot compress is cell allocation, long-lead tooling, the first-article correction loop, and the time it takes to accumulate evidence that a process is repeatable. Those four set your floor, and I have never seen a schedule meeting move any of them.
The reason the prototype number feels like a promise is that it is the only number anyone has early. It is not a preview. It is a different measurement of a different thing, and the production date does not start when the prototype ships. It starts at design freeze.
