When and How to Build Prototypes: Expert Tips for Success

Let’s talk prototypes! How long should it take to build them? How much should it cost? How many should you make? Who should build them? What about when the customer wants them?

Prototype electronic devices are the vessels to put all the blood, sweat, and tears up to that point into your project. They are a critical path for the project to move forward, and it’s common for the hardware team to stall out when it’s time to make them. You can have the hardware folks work on other projects or knock out documentation, like say the DVT plan for instance if you don’t already have it.

But time is money. And there are some key factors that may not be obvious here.

1. Build enough prototypes to avoid costly delays. If you don’t build enough units, you will spend extra time and money fixing units that get damaged along the way. Always build two units for every engineer on the project plus two for the customer and two general spares. The incremental cost for a few more units is trivial compared to the labor costs of having a hardware engineer debug and fix issues on a board because there aren’t enough units to work with. If a unit is damaged in development or there is build fallout, what you WANT to happen is the engineer to put that unit aside, grab another good one and keep making progress on the project.
2. Partner with contract manufacturers (CMs) for prototypes. Build your prototypes with a CM who specializes in building protos. Do NOT try to build protos on a regular production line. At F3, we have a short list of trusted CMs to do this. If we give them 14 units worth of parts, we get 14 working units. If there is build fallout, they’ll work with us to fix it. Nothing is more expensive in budget or schedule than having to fix prototypes.
3. Optimize prototype timelines through planning and local resources. If you get parts, docs, etc. to your CM in advance and time everything around getting the PCBs, you can have the boards come in, get assembled, and start doing initial bring-up all in a day or so. This requires good planning, good communication with your CM, and PCB source and it really helps if the CM is local so you’re not burning a day or more in shipping time. Some devices are more complicated and take multiple assembly processes, so ya can’t always do this. Calendar days in a project schedule can’t be recovered, so anything you can do to speed things up is generally worth it. In situations where you have to ship something like this, consider sending them counter to counter at the airport. Yea, it’s more than overnight shipping but, depending on time zones, you can buy back a day for only a few hundred bucks.
4. The true cost of prototypes: speed and accuracy matter. When you build protos, the value is in speed and accuracy. Your labor costs will be much higher than any production build and that’s expected. Building 14 units of a single PCB device with parts on both sides, just the PCB assembly not mechanical or box build, as of August 2024 we see numbers in the $6k range, assuming we’re supplying the parts kit and PCB. That may seem expensive, and it is compared to regular production costs but you’re paying for very skilled people to make you the first 14 of something that has never existed before, that they’ve never seen before, and to do it in a day or three without any mistakes. This is some of the best spent money in your whole project as long as you get the speed and accuracy you paid for.
5. Set expectations for alpha prototypes. At F3, we have a general policy that we do not give alpha prototypes to customers. Yea, they paid for it and yea, sometimes we have to do it anyway, but it’s important to set expectations with customers that these protos are extremely valuable to the development, but they don’t actually DO anything yet that the customer can experience. There is also the matter of unit testing. There is no automated production tester yet, so any board sent to a customer needs to be tested by engineers on the lab bench. Normally engineers will only test the parts of the design they’re responsible for. Thus, every unit doesn’t get every aspect tested, as that would be a waste of time and money.

Units sent to a customer, however, must have every aspect tested. You can’t send untested devices to a customer, because whatever you didn’t test will fail as soon as the customer tries to use it. This is typically 4-6 hours of lab testing by an engineer per unit. If a customer demands, say 40 alpha protos, well, that’s 160 hours, that’s a whole labor-month of engineering time. That’s not a good use of the customer’s money AND it assumes that everything will be perfect on the first board spin. It won’t. Great engineers knock out a product in two proto spins plus a pilot production run. We often see customers do more like 4-6 spins for more complicated devices. When we do our rescue mission work, where we jump in and help customers who are struggling to get to market, we often see they’ve done 10 or more proto spins before they call us. If each of those took only 2-3 weeks, that’s on the order of six months of basically zero progress. You can imagine how expensive that is.

We have staff whose whole job is to quote, plan, and manage prototype builds. We also do consulting where we can review your processes and make recommendations to lower total project cost and schedule time. If you’re experiencing more than two weeks or so to get prototypes built, F3 can help speed that up.