And here, as promised, is a blog post about how to calculate the sales price - probably also explaining why we can't give any fixed price yet Be warned - it is quite a lenghty post! I hope you enjoy reading it though, it's full of information you probably didn't think about Basically, the price of a device being sold is being calculated using many different factors: Development Costs (one-time costs) Production Costs (costs per unit) After-Sales Costs (Support, Repairs, etc.) (monthly varying costs) Monthly company costs (fixed costs per month) Marketing Costs (monthly varying costs) Some money put aside for bad things that can happen (i.e. the touchscreen issue we had with the Pandora) Profit and Distribution (varying costs) Let's take a look at all of these, shall we? 1. Development Costs The development costs are one-time costs. It means: Once development is over, they won't change anymore, so they will be a fixed value in the final unit price. Development costs include everything you need to do to create the unit up until the part the mass production is going to start: Wages for the designers Costs for prototype parts and production Costs for samples, evaluation parts, etc. Costs for doing different tests (CE certification, reliability testing of various parts) Costs for parts needed for mass production (case molds, stencils for PCBs, machine programs, etc.) Wages for the designers should be clear - someone needs to create the device and he / she needs something to live from as well Of course, using a company for the design (i.e. for the case design) will increase the costs, as the company needs to cover their costs and wants to make profit as well, but in exchange you have additional safety. In our case: The case design company is also the company producing the actual molds. The designer knows their machine quite well, and if something doesn't work, it's THEIR job to fix it. This was different with the Pandora - the case had issues (as you know), and the company blamed the designer for that whereas the designer blamed the company. I am in no position to tell who was right (I'm no expert here), but this is something that needs to be prevented - as in the end, it cost more time and money. Next are the costs for prototype parts and production. You don't just create and design a product and then mass produce it. There are many many tiny little steps until the final product is ready. With such complex technology, it's nigh impossible to have a design fully working on bug free on first try. The Pandora had 5 revisions before it went into mass production. We already had 3 revisions for our dummy PCBs as well. It is possible to save costs here, as we did it with the Pyra: 2-layer PCBs only partly populated are less expensive than creating a full working 8-layer PCB with the SoC, etc. each time. Then we have the costs for samples, evaluation parts, etc. As you probably know from all the discussions on these boards and due to my previous blog costs, there are multiple solutions for everything. Be it SoC, Storage, etc. To find out the best possible solution, you can't just read the datasheets (though it's a good start). It's better to make your own testings. If we want to find out which SoC is best, we'd need to buy devboards for all of those. These cost money. Doing all the tests also needs time - and cost wages for the designer as well. This was just an example for the SoC. There's a lot more - touchscreen, LCD Cable, etc. These all need to be produced and extensively tested - and if they don't work as planned, you might need to find another company or solution and start from zero again. These costs all add up - and need to be covered with the sales later. For the Pyra, I expect development costs of about 50,000 EUR - but it's impossible to tell until everything is finished and working. It might be less, it might be more. But I think we won't hit more than 50,000 EUR, I'm always pretty careful with that. Next are the costs for doing different tests (CE certification, reliability testing of various parts) To be able to even sell the units to the public, we need to be sure they follow all the given guidelines. Which means: The prototype will need to pass tests like CE, FCC, etc. These tests cost money as well - and if they fail, you need to change the prototype until it passes. These are one-time-costs like the development costs. Additionally, we need to expect to sacrifice things as well. The OMAP5 devboard has a heatsink, and we also need to create one for the Pyra as well. However, what happens if the OMAP5 doesn't have a heatsink? Does the SoC die when it gets to hot, or does it simply get very very hot and crash? The only way to be sure is to try that - so it might be we have to sacrifice a devboard for that. I expect about 10,000 EUR for the Pyra here - so certainly less than the development costs. There is something else before the Pyra will finally be ready for mass production: Costs for parts needed for mass production (case molds, stencils for PCBs, machine programs, etc.) Once everything has been confirmed to be working, mass production needs to be prepared. Molds will be created (or finalized), the mass production machine needs to be programmed, stencil layers need to be produced, etc. My expected costs for the Pyra here would be 100,000 EUR. These are all the fixed one-time costs you have before you can start mass production. Now, the expected sales are VERY important for that. Whether we sell 10, 1000, 10,000 or 100,000 units - these costs won't change. If we sell 10, it would break down to 15,000 EUR per unit. With 1000, it would be 150 EUR per unit. 10,000 would go down to 15 EUR per unit, 100,000 would be 1.50 EUR per unit. And so on. This is one of the reasons why low-quantity devices are always more expensive than devices that sell in millions. It's also the reason why we can't try as many different approaches and solutions as a huge smartphone company. With low production quantities like about 10,000, it makes a huge difference whether the production costs are 150,000 EUR or 500,000 EUR (from 15 EUR to 50 EUR per unit) With quantities like you have for smartphones, it doesn't really change the price at all (from 0.15 EUR to 0.50 EUR per unit). 2. Production Costs (costs per unit) Next up are the production costs. Unlike the development costs, this is no fixed value, of course. Producing 100,000 units costs more than 10,000 units The production costs itself consists of different factors. Parts Production time Assembly, testing and packaging Failure Rate Parts costs probably are pretty clear: All the parts that are needed to produce a unit are in there. That's the case, the PCB, the SoC, the keymat, the LCD Cables, the Battery, the packaging, all those tiny little components that are on the PCB, etc. For the Pandora, these costs are about 200 EUR each. I expect slightly higher costs, as the SoC, memory and LCD will cost more. However, nubs are a lot cheaper (since we won't need two ATMELs for the firmware) and we also try to create a 2-layer base PCB and a small 8-layer CPU module PCB (which should also lower the costs). So we should expect something like 250 EUR for the Pyra. Production Time Costs is the time the machine (and the people doing the manual soldering for some comonents) needs to populate each PCB. The machine at Global Components can populate 400 parts per minute - so the fewer parts we've got, the cheaper it will get. With the Pyra, by making use of two PCBs (module and simple main PCB), where we place the complex parts (BGAs) on the way smaller CPU module, the overall production time costs should be a bit cheaper than for the Pandora. The Pandora was 40 EUR per unit, so I expect the Pyra to be around 25 - 30 EUR. Assembly, testing and packaging Each unit needs to be fully assembled, tested and packaged up, so it's ready to be shipped to the customer. With the Pandora, this is pretty complex. We're trying to improve that for the Pyra as well (mainly case design changes). Also, as the Pyra has a faster SoC, the self-tests should run faster. The Pandora cost for this was 10 EUR per unit, I expect 7 - 9 EUR for the Pyra. Failure Rate If you calculated everything up to now, you should have about 300 EUR production costs per unit. So 1000 Pyras would cost 300,000 EUR. However, there's always a failure rate. Global Components guarantees a failure rate of less than 5%. Anything higher is covered by them. The real failure rate usually is lower though (with the Pandora, about 2 - 3%). These are PCBs that could not be made to work, so they can be thrown away. So 50 out of 1000 Pyra PCBs could be broken from the very beginning. The parts can't be reused (well, except for LCD, Case, etc. of course), and they also cost production time. That's would be about 180 EUR per unit. With 50 units, it would add up to 9000 EUR. So in reality, when calculating costs to produce 1000 Pyras, you need to calculate with 309,000 EUR (or 309 EUR per unit). 3. After-Sales Costs (Support, Repairs, etc.) (monthly varying costs) So, unit shipped, customer happy, no more costs? That would be great - but unfortunately, things can always break. Warranty repairs cost time and money (wages, shipping and part costs). Thanks to our experience with the Pandora (and to the fact that my shop is making additional money with other things), these won't be a huge issue for us. But you need to expect them - so I'll calculate with 4 - 5 EUR per unit. Of course - later in the production, the after-sales cost will go down, as the production process will constantly be improved. 4. Monthly company costs (fixed costs per month) A company has monthly costs as well. Wages, room rent, tax counsellor, etc. For my shop, this is about 4000 EUR per month. Luckily though, these costs are mostly covered by all the sales in my shop already, so it's not a huge concern for us. But what if sales go down and I mostly only sell the Pyra anymore? Now, as these are costs are monthly costs, they heavily depend on sales. If I sell 100 units per month, I would need to add 40 EUR per unit just to keep the shop going. If I sell 1000 units per month, it would only be 4 EUR. The reality is probably somewhere inbetween. There's another concern though: The more units I sell, the more people and space I need to hire to be able to ship them and have enough space for the stock. That's why I want to grow slowly - what would happen if I have 10,000 sales per month, and after 6 months it suddenly goes down a lot...? I still had to pay the rent and wages - and then would probably run out of money at some time. Right now, I'm calculating with about 10 EUR per unit. 5. Marketing Costs (monthly varying costs) If sales go down, more marketing is needed. Thankfully, this is not really an issue for us - most of the marketing happens due to retro-gaming magazines, reviews and the community itself. So it's only about 300 EUR per month at a maximum - something we can easily cover 6. Some money put aside for bad things that can happen (i.e. the touchscreen issue we had with the Pandora) Even though the production is running, something bad can always happen. Like the recent issue with the touchscreen in a full LCD batch, which did cost us thousands of EUR just for shipping. That's why you always need to put a bit aside - to be able to cover that. For the Pandora, that was 10 EUR. I don't expect that much for the Pyra though. I've taken care that all our companies are sitting in Europe, mostly in Germany. So if the touchscreen issue would happen, I'd simply return it to our German distributor and let him deal with the rest. So I'm planning with 4 EUR per Pyra. 7. Profit and distribution (varying costs) Last is profit and distribution. If you calculated everything together, we should have a rough price of 350 EUR per unit (expecting a total sale of 10,000 units) Usually, to be able to work on future projects (successors, improvements, docking station, etc.) you need to make some profit as well. Often this is up to 50% for some huge smartphone companies, but I'm calculating with about 15%. Or, if it's a company that also sells games and apps, they often don't make profit on the hardware, as they plan to make profit with the games. This is not the case for the Pyra though. So that would increase the price to 402,50 EUR Then there are also distributors (like ThinkGeek, Ithic.com, GetDigital.de, etc.) They also want at least 15 - 20% profit (they've got huge shipping costs from Germany to US as well) - which leads to a price of about 500 EUR. Of course, they won't be happy if I sell it for a lower price as they do - so I would need to calculate with that price as well. In our scenario, we'd have a final price of about 499 EUR. However, as we don't know the final development costs, production costs, etc. right now, this is just an estimation. But thanks to my experience with the Pandora, we should roughly get that. It shouldn't be a lot higher though. Wow... this was a long post. I hope you enjoyed reading it and learned something new Calculating costs certainly is no easy task. There are a lot of different ways optimizing the production itself, to save costs. You just need to find the best way (which sounds simpler than it is). Also, estimating sales is not easy. Having more sales will lower the production costs a bit, but if your price is low and the sales aren't too good, you make a loss as well. But having done that for the Pandora (and my shop in general as well) surely will help us a lot.