A new Android POS terminal takes about 26 weeks from a frozen specification to mass production. Roughly eight of those weeks are engineering. The other eighteen are queues: the tool shop, the certification lab, the panel supplier and the trial production line. Knowing which queue you are in is the difference between a project that lands in twenty-six weeks and one that lands in forty.
The calendar below is the one that holds for a 15.6-inch countertop terminal with a built-in 80 mm printer, CE and FCC, and an existing system-on-chip platform. A handheld with a new radio adds four to six weeks of certification. A machine built into an existing housing removes eight.
The calendar, week by week
| Week | Gate | What has to be true | Who owns it |
|---|---|---|---|
| W0–W2 | Specification frozen | Screen, SoC, memory, ports, printer width, battery, target certifications | Buyer and engineering together |
| W2–W4 | Industrial design | Shell, bezel, cable exits, mounting, service access | ID, with mechanical review |
| W4–W10 | Tooling | Steel cut, T1 samples, two reworks allowed | Tool shop |
| W6–W9 | EVT | The design works at all, on a bench, in a hand-made shell | Electrical and firmware |
| W9–W13 | DVT | It works at the limits: thermal, drop, ESD, printer duty | Reliability |
| W12–W19 | Certification | Lab slot booked, samples submitted, reports issued | Compliance, external lab |
| W18–W22 | PVT | The line can build it, with fixtures and a test program | Production engineering |
| W22–W26 | Mass production | First volume lot through all gates | Factory |
The overlaps are real and they are the whole point. Certification starts on DVT samples because waiting for PVT adds six weeks to the end of the project. Tooling starts before EVT finishes because T1 takes 25 days and cannot be compressed by working harder.
Weeks 0 to 4: the specification freeze is the only cheap decision
Every decision made before the tooling order is cheap. Every decision after it is a new mold. That is the entire economics of this phase, and it is why the specification review is worth two weeks of arguing.
Nine items have to be frozen, and they are not the nine a marketing brief usually covers. Display size and resolution, because both feed the mechanical design and the panel lead time. System-on-chip, because it determines the Android version and the firmware support window. Memory, because 2 GB of RAM is the difference between an application that runs and one that is killed by the low-memory handler during a print job. Port count and type, because they define the rear panel opening and the harness. Printer width, because 58 and 80 mm are different mechanisms with different power budgets. Battery presence, because it decides whether the housing needs a sealed compartment. Mounting pattern, because a VESA boss needs structure behind it. Certification markets, because they decide the radio configuration. And the service concept, because whether the unit opens with four screws or twelve changes the housing design.
The change that costs the most
Moving a connector after T1 is not a design change, it is a tooling change: the rear cover is re-cut, or the steel is welded and re-machined, and you lose three to five weeks. Moving a connector before T0 costs a redraw and one day. This is the single largest avoidable cost in a hardware project and it is always caused by a specification that was left open.
Weeks 4 to 10: tooling, and the 25 days nobody can shorten
A large injection mold for a 15.6-inch rear housing takes about 25 days from steel cutting to first trial. The steel is typically a pre-hardened 718H or a hardened S136 for a high-gloss part, and the mold is expected to run 300,000 to 500,000 shots before it needs work.
T1 is not a production part and is never treated as one. The first trial exists to reveal three things: sink marks over the internal bosses, warp across a large flat cover, and short shots at the far end of the flow path. Warp is the expensive one. A 15-inch rear cover that bows by more than about 1 mm will not seat on a counter flat, and the fix is either a gusset added to the rib pattern — which means re-cutting — or a slower, hotter process that adds cycle time to every part forever.
Budget three trials. T1 to T2 is typically seven to ten days because the tool is still in the shop and the change is small. T2 to T3 is longer if the change involves welding and re-polishing. A project that claims two trials has either a very simple part or a very tolerant definition of finished.
Weeks 8 to 14: EVT and DVT, and the rework nobody budgets
EVT answers one question: does the design work at all. The boards are hand-assembled, the housing is a machined or 3D-printed shell, and the firmware is a debug build. Twenty units is a normal EVT batch, and ten of them will be destroyed in testing.
DVT answers a harder question: does it work at the edges. This is where the antenna position gets moved because the metal back plate detunes it, where a thermal pad appears because the SoC throttles after forty minutes of printing, and where the speaker cavity gets a gasket because the printer motor buzzes through the housing at 2.4 kHz. Every one of those is a mechanical change, and every mechanical change is a tooling change if it is late.
The realistic budget is one to two mechanical revisions between EVT and DVT, and a further one before PVT. A project plan with zero revisions is a plan written by someone who has not shipped a product.
Weeks 12 to 19: certification is a queue, not a task
The testing itself takes days. Getting a slot takes weeks, and the report takes another week after that. In the fourth quarter, when every brand is trying to ship for the holiday season, lab slots are booked four to six weeks out, and no amount of money moves you to the front.
| Approval | Standard | Typical lab time | What it covers |
|---|---|---|---|
| CE, EMC | EN 55032, EN 55035 | 2–3 weeks | Emissions and immunity |
| CE, safety | EN 62368-1 | 2–3 weeks | Mains isolation, thermal, fire enclosure |
| CE, radio | EN 300 328, EN 301 489 | 3–4 weeks | 2.4 and 5 GHz Wi-Fi and Bluetooth |
| FCC | Part 15 B and C | 2 weeks | US emissions and radio |
| UN38.3 | UN Manual of Tests and Criteria | 3 weeks | Lithium pack transport, if a battery ships |
| RoHS and REACH | EN IEC 63000 documentation | 1 week | Material declarations, not a test |
Two things about this table surprise buyers. The first is that CE is a self-declaration: the lab produces a test report, and the legal artifact is the Declaration of Conformity plus a technical file that you sign. A folder containing only a lab report is not compliance. The second is that a change made after certification — a different panel, a different power adapter, a moved antenna — can void the report and require a retest of the affected item, which is why the design has to stop moving at W13 rather than at W19.
Weeks 18 to 22: the three trial rounds
PVT is where the factory, not the design, is tested. The unit is already known to work; the question is whether a line with real operators, real fixtures and a real test program can produce it repeatably.
Three rounds is standard, and each has a different purpose. Fifty units proves the process exists and shakes out the assembly sequence. Two hundred units proves the test fixtures and the burn-in rack have the capacity, and produces enough units for a packaging drop test and a few customer samples. Five hundred units is the dress rehearsal: same operators, same fixtures, same shift pattern as volume, with the first-pass yield measured and reported. If first-pass yield at 500 units is below about 95%, the line is not ready and the honest answer is another round.
The four things that actually slip
Ask anyone who has run this schedule where it went wrong, and the answer is almost always one of four items — none of which is engineering.
| Slip | Why it happens | What it costs | How to contain it |
|---|---|---|---|
| Panel lead time | A 15.6-inch 1080p panel is 45 days from order, often with a 500-unit MOQ | 4–6 weeks | Order EVT panels at W4, before the design is frozen |
| SoC allocation | Mid-range chips are allocated, not stocked, in a tight market | 6–10 weeks | Confirm allocation in writing with the module supplier, not the factory |
| Certification retest | A late change invalidates an EMC report | 3–4 weeks per cycle | Freeze the design at W13; batch all changes before it |
| Mold rework | Warp or a short shot needs steel work | 2–3 weeks per loop | Review the rib pattern and draft angle before the steel is cut |
The first two are procurement, and they are the reason a serious project places long-lead orders before the design review is signed. The second two are discipline, and they are the reason the design freeze date is a commercial commitment rather than an engineering aspiration.
Where a buyer can legitimately compress the calendar
Three shortcuts are real and do not add risk. Reusing an existing housing and tooling removes about eight weeks and the whole tooling budget, at the cost of the industrial design being someone else’s. Using a pre-certified radio module removes three to four weeks of radio testing, because the module carries its own approval. And submitting EVT-level samples for certification, rather than waiting for PVT, overlaps six weeks that would otherwise be sequential — provided the design genuinely stops moving.
The shortcut that does not work is asking a factory to run the phases in parallel that cannot be parallel. Tooling cannot start before the mechanical design is released. Certification cannot start before there is a stable firmware build. A schedule that shows them overlapping by more than the natural overlap is a schedule drawn to win the order.
FAQ
How long does a new POS terminal really take?
Twenty-six weeks is realistic for a countertop terminal on an existing platform with experienced tooling and certification partners. A handheld with a new radio and a new battery is thirty to thirty-four. A first product from a factory that has not built one before is longer, because the queues are unfamiliar.
Can the tooling phase be skipped for a small order?
Yes, by using an existing enclosure and changing the badge, the colour and the software. That is a private-label build rather than a new product, it is a legitimate way to start, and it removes the eight weeks and the mold cost entirely.
Who owns the mold?
Whoever paid for it, and this needs to be in writing. If the factory pays, the tool is theirs and moving to another supplier means paying again. If you pay, ask for the steel certificate and a clause that the tool is released to you on request.
Where to go next
The calendar is a queueing problem, and the queue you are in decides the delivery date. For the platform decisions that set the rest of the schedule:
DCA-155A, the 15.6-inch reference platform POS-116A 11.6-inch with printer Android POS systems who pays for the tooling, and who owns it what the display assembly goes through the benches DVT runs on · countertop, handheld, kiosk or all-in-one
Specifying the hardware? The hardware side is covered in The five lines on a POS bill — which of them are hardware and which are not.