When to Stop Prototyping on a Dev Board and Start Designing a Custom PCB

Knowing when to graduate from a dev board to a custom PCB design

Your prototype works. The LEDs blink, the sensor reads, the data hits your dashboard. You hand it to an investor and watch them squint at a Nordic DK strapped to a breadboard with three jumper wires hanging off the side like a bad haircut. They smile politely and don’t write the check.

Every hardware founder hits this moment: the demo that can’t leave the desk. Your firmware is solid and your logic is proven, but the thing in your hand can’t be shipped, can’t be certified, and won’t survive a pilot customer’s back office, let alone a wrist or a shipping container.

The real question isn’t whether to move from dev board to custom PCB. You should have started 3 weeks ago.

What Dev Boards Are Actually For

Dev boards exist to kill technical risk fast. You plug in a Raspberry Pi, an ESP32 DevKit, or a Nordic nRF52 DK because you want to know: can my firmware talk to this sensor? Will my BLE stack hold a connection? Does the power profile look sane?

They’re brilliant at that job. They’re terrible at being products.

A dev board carries hardware you’ll never ship: debug headers, USB-to-serial chips, current shunts, exposed GPIO, regulators sized for bench use. You’re paying $40 for $6 of useful silicon. The footprint is wrong, the antenna is generic, and the connectors point the wrong way for your enclosure.

The 7-Signal Checklist: Is It Time?

You should move from a dev board to a custom PCB when 3 or more of these are true:

1. Form factor Does your enclosure exist on paper but not in reality because no dev board fits inside it? For wearables, asset trackers, or anything pocket-sized, a 50mm x 70mm dev board with stacked headers will never fit. If yes, lean custom.

2. BOM cost Add up your dev board, breakouts, and wiring. Compare that to your target unit cost. If the dev kit alone is more than 30% of your retail price, the math never closes. A $35 ESP32 DevKit inside a $99 product is a business model on life support. If yes, lean custom.

3. Power budget Dev boards bleed current. There’s a USB-to-UART chip pulling power, an always-on regulator, and an LED you can’t desolder without lifting a pad. Together they eat 5-20mA of idle draw you’ll never see on a custom design. If you need 6-month battery life and you’re getting 6 days, no amount of firmware tuning will save you.

4. Connectivity Wireless is where dev boards lie to you most. The PCB antenna is tuned for an empty bench, not your plastic enclosure with a battery 4mm away. If you need cellular, BLE, LoRa, or satellite tuned to your actual product, you need your own board. (This is also where managed connectivity matters: if you’re targeting a network like Hubble’s, follow the transmission guidance and antenna constraints before you commit to a layout.)

5. Mechanical integration Mounting holes in the right place. Connectors on the right edge. A sensor that has to sit 2mm from the enclosure wall to read accurately. Dev boards put these where the dev board vendor wanted them, not where your product needs them.

6. Certifications FCC, CE, PTCRB, and the rest won’t certify a dev board as your product. They certify the exact hardware you ship, the exact serial number on the exact panel. If you’re heading toward a pilot that touches regulated radio (and almost everything does now), you need the production board months before you think you do. Test labs are booked out, and re-spins after EMC failures are routine.

7. Volume Past 50-100 units, dev boards become operationally painful. You’re hand-soldering jumpers, flashing one at a time, and explaining to your CM why the “product” arrives in an anti-static bag with a USB cable taped to it.

Score yourself. 0-2: keep iterating. 3-4: start designing now. 5+: you’re already late.

The Decision, Visually 

        ┌─────────────────────────────┐
        │   Working dev board demo    │
        └──────────────┬──────────────┘
                       │
                       ▼
        ┌─────────────────────────────┐
        │  Score the 7 signals (0-7)  │
        └──────────────┬──────────────┘
                       │
       ┌───────────────┼───────────────┐
       ▼               ▼               ▼
   ┌───────┐      ┌────────┐      ┌─────────┐
   │  0-2  │      │  3-4   │      │  5-7    │
   │ Stay  │      │ Start  │      │ You're  │
   │ on    │      │ design │      │ already │
   │ dev   │      │ now    │      │ late    │
   └───┬───┘      └───┬────┘      └────┬────┘
       │              │                │
       ▼              ▼                ▼
   Keep        Carrier board     Full custom
   iterating   w/ module         + certify

The Middle Path: Carrier Boards and Modules

Here’s the move most first-time founders miss: your first custom PCB shouldn’t be a fully integrated SoC design. It should be a carrier board wrapped around a pre-certified module.

A module is a small, shielded, FCC/CE-pre-certified subassembly: Nordic nRF series modules, u-blox cellular modules, ESP32-WROOM, Quectel BG95, and so on. They cost a few dollars more per unit than the bare chip. They save you 6-12 months and tens of thousands of dollars in RF layout, antenna tuning, and certification work.

Your carrier board does the boring stuff: power management, sensor interfaces, connectors, mechanical fit. The module does the hard radio work. You inherit the module’s certifications (with caveats, so read the modular approval docs).

DEV BOARD ──► CARRIER + MODULE ──► FULL CUSTOM PCB
  Fast            Balanced              Optimized
  Bulky           Production-ready      Smallest/cheapest
  $$$/unit        $$/unit               $/unit at volume
  No cert         Pre-certified module  Full cert required

For 90% of hardware startups, the carrier board is the right first PCB. You can always integrate the radio onto your second-generation board once you’ve shipped 10,000 units and need to claw back $2 of BOM.

What the First Custom PCB Actually Costs

Honest numbers, because nobody tells you this until you’re 6 weeks in:

  • Design: $8K-$40K. An experienced EE contractor runs $100-$200/hour. A 4-layer carrier board with a wireless module, power management, and a handful of sensors runs 60-200 hours. In-house EE costs more upfront but pays back across spins.
  • Prototype fab and assembly: $1K-$5K for 5-10 units. Faster turn = more money. JLCPCB and PCBWay are cheap; quick-turn US fabs aren’t.
  • Spins: budget for 2-3 revisions before you have a board that’s manufacturable and passes pre-compliance. Anyone who tells you they nailed it on rev 1 is either lying or hasn’t tried to build 1,000 of them yet.
  • Timeline: 8-16 weeks to first working revision. Add 4-8 weeks per spin.
  • Certification: $15K-$50K for FCC + CE on an intentional radiator, less if you’re using a pre-certified module under modular approval.

Total for a first custom PCB program from kickoff to certified, manufacturable hardware: $30K-$120K and 4-9 months.

That sounds like a lot. The cost of not moving is worse: blown pilots, broken unit economics, fundraising rounds that stall because you can’t show real hardware.

Common Mistakes on the First PCB

  • Designing custom silicon on rev 1. Use a module. You’re not Apple. You don’t have an RF lab.
  • Skipping the carrier board step. Going dev-board-to-full-custom triples your risk and your timeline.
  • Underestimating firmware porting. Your code ran on a dev board with a bootloader, debug output, and known-good clocks. Plan a few weeks to bring it up on bare hardware. The Hubble device SDK and reference apps are worth a look here if you’re using a supported chipset, since most of the porting pain is in BSP and pin mapping.
  • Not budgeting for spins. If your plan assumes rev 1 ships, your plan is wrong.
  • Hiring a PCB designer who’s never worked with a startup. Defense and medical EEs design beautifully and slowly. You need someone who’s shipped consumer hardware in under 12 months.

Score Yourself and Make Calls This Week

Score your prototype against the 7 signals. Be honest. If you’re at 3 or more, start interviewing EE contractors and pricing pre-certified modules that match your connectivity needs. Begin with a carrier board, not a full custom design. Budget for 2-3 spins and 6 months.

If you’re at 5 or more, stop reading articles and start making calls. You’re behind, and every week you spend hand-soldering jumper wires onto a dev kit is a week your competitors are spending in a test lab.

The dev board did its job. Let it retire to the parts drawer where it belongs.

Hubble Network provides global Bluetooth connectivity directly from your custom PCB, no cellular modem or WiFi provisioning required. See how it works →