New tech start-ups fail for many different reasons, but some of the most heartbreaking cases are when an entrepreneur correctly identifies a market need, creates a stunning solution to meet that need, but then runs out of money during the production process before entering the market. Matt Murphy, the Program Manager for MADE@ MassChanellege explains how costs can make or break start-ups.
“Costing for manufacturing is integral to any early-stage hardware startup. One of the most preventable failure points, especially when crowdfunding production, is not understanding the cash runway required to make it to production and then actually paying for manufacturing. At MassChallenge we ensure our companies have the appropriate mentorship to understand and plan for the costs of manufacturing.”
MassChallenge is the world’s largest Tech Start-up Incubator, and each year they provide 120 potential start-ups the classes, mentorships, office space and contacts to bring their tech product to the market. CAPINC is happy to be an in-kind sponsor of MADE@ MassChallenge 2015.
At CAPINC, we can’t help with the costs of regulation, personnel, or real estate (physical and virtual) needed to start a tech start-up, but we can help entrepreneurs quantify hidden technical costs they may not have considered.
Knowing these costs is important for calculating your runway, a key start-up metric which is your ratio of funds on hand to monthly expenses. If your start-up has $100,000 in the bank and costs $10,000/month to run, then you have 10 months of runway to hit it big or crash and burn. Below are some of the costs you can plan for (and CAPINC can help you avoid) if you don’t want to hit a sudden pothole just as your start-up is taking off. This also applies to those start-ups in “taxi” mode, slowly improving their product before they are ready to make that final, all-or-nothing push for public launch. They need to know what costs might be coming, too. Download the complete Start-Up Costs Guide below to better understand your cost options for creating and nurturing your tech start-up.
Let’s talk about the costs you’ll hit in order, starting with:
1. Prototyping Costs
Of course you’ll need to make a prototype to prove to yourself that your idea works, but here’s what we learned at a ENET meeting last year:
If you are bringing a physical product to market, no angel investor will invest in you unless you have a working prototype.
So while the prototype you make for yourself will be an alpha prototype (or a “works-like” prototype), the one you’ll want to show investors is at least a beta prototype (or a “looks-like & works-like”). So how do you budget for that? No matter what product you’re making, it can only ever have four classes of parts:
Obviously for Class A parts you get the prices from the suppliers, and for Class D parts you have to estimate the costs on your own. Let’s focus on Classes B and C.
The poster child for Class B is a custom printed circuit board (PCB). While your alpha prototype may get by with a couple Ardiuno/Raspberry Pi boards you strung together, you might want your beta prototype to be bit more optimized.
The major things that affect PCB manufacturing costs are:
Number of components on the board (which you can’t control much)
Method of insertion of components onto the board (automatic or manual) and,
Location of components.
To give a sense of scale, besides the raw cost of the component, having an automatic machine place the component on the board before soldering will run about $0.01/component. Having a human in the USA do the same task might cost ten times as much. And having a PCB with components on both the top AND bottom of the board will more than double your costs, since the PCB now has to be flipped halfway through the assembly process.
So to try and reduce your PCB manufacturing costs, you could either:
Try to redesign your circuit to use hundreds of fewer components, doing a lot of work for little monetary gain, or
Make sure you choose mostly standard components that automatic pick-and-place machines can handle and not oversized/weird ones that require a human to place, or
Try to design your enclosure so that all the components can fit on the same side of the board (small effort for big effect)
It’s obvious that the last choice offers the most bang for the buck, and that’s why CAPINC teaches our customers how to use the CircuitWorks add-on for SOLIDWORKS, which can take your 2D circuit board layout from Altium, Eagle or OrCAD and import it into your 3D model, so you can do a serious interference check to see if all the components will fit on just one side of the board:
Cellphone circuit board modeled in SOLIDWORKS through CircuitWorks. Note the components on the bottom of the board (in reflection), which will make manufacture of this board more expensive.)
A simpler circuit board that would be less expensive to manufacture if the lone bottom-side component (blue cylinder on the left) could be moved to the top side for easier placement during assembly.
Of course, it’s not always possible to move your PCB components to all one side, but since it usually takes less than 10 minutes for CircuitWorks to build a fully featured 3D model of your board as shown above, it’s worth doing that check at least once during your design process.
The poster child for Class C is your product’s outer case. It’s unlikely that out in the world there already exists some sort of shell that perfectly fits your components in the pleasing arrangement you want them in, and while making cases out of cardboard or wood are good enough for alpha prototypes, you don’t want to show that to investors.
Plastic is a natural fit for most cases, but injection molding only becomes cost effective at thousands of units. As we’ve discussed before, there are many types of plastics you can 3D print in (we even made a chart for you to compare all their properties) but what most folks don’t realize is, they don’t have to buy a 3D printer to get the benefits. In order from least expensive to most useful, you can:
Print parts on demand through a service (typically $4/cubic inch of your model volume + the service’s mark-up)
Lease a printer for a set amount of time (lets you cut the 1-5 day shipping times out of option 1, since the printer is now at your site)
Buy your own printer (from $2,000-$400,000, depending on the build envelope and features you choose)
The benefits of 3D printing your case versus using cardboard is that 3D printed cases will pass muster when looked at by potential investors. The benefit of printing in plastic versus cutting metal is that if you made a mistake in the design of case, you can just throw that part away and print a new one for a few dollars more in 1-2 days, versus another trip to a machine shop. 3D prints can even be painted or have screw threads inserted. With your prototyping costs estimated, you’re ready to move on to your second hidden area.
2. Design Costs
Why did Design come after Prototyping- isn’t that backwards? Not really, since most smart start-ups in the tech field use the “Build-Test-Learn” cycle made famous by Eric Ries’ book The Lean Start-Up, an unmatched resource on how to build the right product in the least time, from someone who’s been there.
So after each Build-Test-Learn cycle, you’ll be doing design changes based on what you’ve learned. We all know that changes made early in a design are 10x to 100x cheaper than changes made late or after production. And no company wants the horrible PR that comes with a string of possibly predictable failures after a product has been released to the public (See Potato chips with Olestra in them).
So why don’t companies test for every major failure mode during every design iteration? Because of that end of the runway inexorably creeping closer: they don’t have time. Or at least, they don’t have the time for physical tests every iteration. Finite Element Analysis (FEA) tests done in software only aren’t as good as the physical tests they simulate, but take 100x less time and can catch the same gross failure modes you want to avoid.
And all modern FEA systems, SOLIDWORKS Simulation included, can take a stress study done in one variation of a product and easily apply it to the next iteration of a product with minimal effort. That means a test you set up for Build-Test-Learn cycle 1 can be applied to cycle 2, 3, 4, etc., which is useful to spot trends or compare a not-yet-built model to something you’ve already broken.
Check out a past article I wrote to learn what every user should do in FEA first. If you want to know what to do last, right before you show your stress/strain results to anyone, we got you covered there as well. At a minimum, all mechanical parts should be tested for overload and fatigue failure, and electrical products should be tested for overheating. We’ve also covered how to use CFD to test for electrical over heating failures.
And anything hand-held should be drop tested in FEA and real life, since not meeting customer expectations in that area could lead to that failure mode becoming the most searched Google term for your product for a while:
All four of these tests – overload, fatigue, overheating and drop test – can be done on FEA programs that can cost $5,000 and up. And while CFD programs can start at $20,000, which is a big deal for start-ups, the real killer is the time it takes your engineer to get good enough with simulation to produce trustworthy results. This can be 1-3 weeks for an engineer new to FEA, which isn’t a lot of time compared to an engineer’s whole career, but that’s a lot of runway to chew up.
Especially when you consider that the carrying cost of most design engineers in the New England area is around $100,000 (counting salary, insurance, office overhead, etc), which works out to around $1,900/week. This is why CAPINC offers FEA and CFD consulting for $2,000/day, with the idea that each day of consulting saves a company 1 man-week of engineer time, letting you focus on your core strengths and not have to learn a new skill you’ll only use 3 times a year.
Simple questions like “Will it break?” or “Will it overheat?” can usually be answered in 1-2 days, with deeper questions like “Which is the best design direction to avoid future failures like this?” taking longer. Once you’ve got an estimate of the costs for each design cycle, you can move on to your biggest cost area of all.
3. Production Costs
Per-unit production costs can make or break your tech start-up and are intimately tied to decisions you made in the design phase. Tish Scolnik, CEO of GRIT, is a previous MassChallenge winder, bringing all-terrain capable wheelchairs to developing countries, and now to the United States. Tish elaborates on how much final per-unit cost of manufacture affected the success of their all-terrain wheelchairs designed for developing countries:
“We had a price benchmark we were targeting from the start, so it was very important that our per-unit cost of manufacture came in reasonably below that. That being said, we were focused on getting the technical aspects right. We did as much Design-For-Manufacture (DFM) as we could [early] in the process, and realize that we will be doing a lot more DFM down the line, especially as our production volume increases and we can take advantage of other high-volume manufacturing techniques.”
Consider something like this bracket assembly:
That’s a simple enough Class C type supporting bracket that gets zipped off early in the prototype stage when you just need some way to hold your special sauce parts in the right position and move on. But if you’re now going to make thousands of these, that means someone will have to line those two parts up and insert those two bolts every single time. Why not make it one part?
In general, reducing part count always reduces costs. First, it’s fewer part numbers, and studies have shown that each new part number needed to be assigned, tracked and carried in your business costs around $5000 initially + $100/year. Secondly, assembly isn’t free. Numbers always vary, but you can estimate that hand assembly costs in the USA and overseas vary greatly.
Looking back at the single bracket, the corners are sharp; this part is meant to be cut or milled into shape, not bent. In days past, CNC machining a part like that was the quicker way to manufacture it, since setting up tools and dies to bend it out of sheet metal took longer, since every part had to have part-specific tooling.
But modern sheet metal shops have overcome some of these limits by using machines and dies that adjust to the part they are making, eliminating the more expensive need to stop and switch tooling between parts.
Using SOLIDWORKS “Convert to Sheet Metal” function, I can make the above CAD model into its sheet metal version in about 30 seconds.
Now those sharp corners are standard round bends that sheet metal shops can make in their sleep, and the benefit of doing this conversion in CAD first is that you can now check for interferences from the newly changed shape, instead of finding out after 1,000 parts arrive.
For example, I would like to notice early on in my Design-For-Manufacture phase that the three small holes cut through the upper tab of my bracket are coming perilously close to the bend used to make that upper tab.
Experienced sheet metal houses will give you guidelines for how close your important features can come to their bend radii without causing problems, but our eye warns us it’s probably too close already.
And that’s only where the benefits of using CAD in DFM start. Now that you’ve got a designed shape, how would you estimate the cost of a Class C part like that? You could guess. Is $5/part too high? Too low? You could use the raw material cost, which is about $7 for a part like that made from Aluminum.
You could calculate the rough perimeter needed to be cut from a sheet of raw metal, figure out how long a water jet cutter would take to travel that distance, multiply by machine rental cost, then painstakingly count every single sheet metal bend and add up the costs of doing all those bend operations. Whew.
But in the past, that last way is how sheet metal quotes were actually made by hand, but now there are much easier processes. In early stages, you should use SOLIDWORKS Costing tool, which is built right into the CAD and does all the painstaking manual steps listed above, but just automatically, since the CAD system always knows the part perimeter and the number of bends in the geometry. Doing so gives us the following report.
Which estimates about $14/part if I build 100 parts. To get to that number, the software flattened the sheet metal part, estimated the cost for each cut, using a standard rate for laser cutting machines.
And did the same thing for bends.
And added all of that up.
Does that mean I can go out and book 100 of these parts at a sheet metal house right now for $14.28 a part guaranteed? No, each house will give me a different quote, based on their rates for each operation. But for 10 seconds of work and a few button clicks, it’s a darn sight better than the guesses we were making before. ($5/part?! Were we crazy?!?)
The SOLIDWORKS Costing add-in is good enough to get some first numbers to start estimating what your sheet metal or CNC machined parts will cost. (The big benefit is that the cost updates instantly with every design change you make, letting you quickly compare alternatives.)
Example of quoting process, not calculating above brace part model.
But as you’re getting closer to production, you’ll want to engage a real shop and get quotes based on their numbers. Rapid Sheet Metal has an online quoting tool, but also a free SOLIDWORKS add-in called SolidQuote that gives you a real quote based on the model you’ve got open.
And another company that has an on-line quoting tool is Protolabs, who does many types of small, flexible production runs as well. Protolabs has a range of manufacturing options, such as 3D printing in plastics and metals (the latter is pretty rare), CNC, and injection molding of plastic, metal and rubbers (the last two molding types also being rare).
Protolabs is the type of place a lot of start-ups use as a bridge manufacturer. The first 1-5 prototypes entrepreneurs usually build themselves. An order for 10,000 units usually leads to the set-up of a fully running assembly line with all its associated sunk costs. But what do you do if you need 500 of something? You want a flexible but repeatable solution to bridge the gap, without all the same sunk costs, and a company like Protolabs might help you out.
This same sort of calculation happens with manufacturing plastic parts. Injection molding can lead to costs per part in the pennies, but the set-up costs can be $100,000 for making your own steel mold. Since Aluminum is easier to machine than steel, Aluminum molds cost 15-30% less to first make than steel molds but can wear out 50% faster, perhaps as quickly as 20,000 shots.
For this reason Aluminum molds are sometimes used as bridge molds for companies needing to make a short run of injection-molded parts while saving up enough capital to invest in a steel mold. And even smaller scale then that, some of our customers are using the larger scale 3D printing machines, such as the Fortus line, as bridges to their aluminum molds, printing parts at a fixed rate of $4/cubic inch instead of plunking down $20,000 at once for an aluminum mold that may get drastically redesigned in a month.
CAPINC offers $5,000 injection molding simulation software, so that you can virtually test to see if your $20,000- $100,000 injection mold will even fill right the first time.
Gray area is a thin part that injection molding can’t fully fill. Found out early in the design stage thanks to Simulation.
And after looking at all those manufacturing costs, you’re ready to move on to the final hidden cost to consider.
4. People Costs
For a design engineer in the New England area, if you want round numbers you can estimate paying around $100,000/year, including all overhead costs. Which is why it’s amazing when companies decide not to spend an extra $500 to get an engineer a faster computer or training to make their core tasks faster.
Sure, there’s always that runway counting down, even after your company has “made it”, but if a small extra cost could get 2x more out of that $100,000/year investment in each engineer, why not take it? We’ve seen that replacing an old computer to one capable of running your biggest, baddest CAD assemblies easily exceeds this 2x factor.
This is why the book “Rework” by Fried and Hansson, another excellent resource for start-ups looking to create the right product and culture for their company, suggests “test driving” employees before you take a gamble on them, because “Resumes are Ridiculous”.
Fried and Hansson also suggest sending everyone home by 5 o’clock, no matter how many overdue tasks are waiting on that employee’s desk. That’s because having overworked, stressed-out, and rushed people pulling late nights just leads to more errors in your design, which are even more costly to remove later. This final reason is why CAPINC offers engineering consulting to all sizes of companies, so you can add our engineers during surge time to conduct FEA or CFD studies, design iterations, or train your new hires how to use CAD, so that your more experienced hires can keep their work schedules sane.
We’re committed to making New England companies more productive, which is why we’ve collected all the tips, costs and tradeoffs mentioned in this article plus more, and collected them into one, unified infographic below to get start-ups considering all the details and hidden costs from day one. To get an even deeper-dive into predicting and planning for start-up expenses, watch the on-demand webinar now on calculating costs.