Reverse Cost Accounting

First published on the DigiFabster blog

A few weeks ago a friend asked for help on a project he was developing. He needed pricing for 3D printing a prototype. The model he wanted to produce was heavy (roughly 100kg of PLA) and big, with a bounding box of almost one cubic meter. I warned him it would be expensive if the model wasn’t optimized…but I didn’t realize how expensive.

Well, the first offer in was for $16,000.

Amazing. I asked the service provider to explain a little.

He did:

The quote included $2,000 for material, which equated to roughly $20 per kilo, which is more or less the market price.

He also wanted $14,000 for the actual printing. His hourly rate was $23, which is about average. But due to the size of the model, even with the machine running 24 hours per day, it would take 25 days to complete.

The 3D printer in question costs $30,000. In theory they would have their money back in two or three jobs like these.

But seriously, let’s have a look at what this printer costs them per month, which is more or less the period they would be using it for my order.

1. Finance. Assume they took money from the bank and plan to amortize it over 5 years. They need to pay off the principal and interest in 5 years, so the bank will lend them the same sum when they need it in the future. At an assumed interest rate of 5%, their cost here is $560 per month. This job takes roughly a month, so from the $14,000 that leaves $13,500$ for other costs. (Tip: Use excel ‘s “=PMT“ function for calculations like these.)
2. Electricity: The printer takes 220V, 16A (110V, 32A in US terms), that’s 3.5 kWatt installed power. Assume half of that when running, so 1.75 kWatt for 25*24 hours, 1000 kWatthours @ $0,25 per kWatthour=$250for this job ($300 per month).
3. Spare parts and service. Assume those two together cost 20% of the price of the printer itself over the 5 year period, so $6,000 total, divided by 60 months = $100/month. Seems about right.
4. Space. The printer takes up space in a room in an office for which rent is paid. If the price per square meter per year is $120 ($11/ft2), that’s $10 per month/square meter. With space for servicing, loading and unloading, the printer takes up 5 square meters, so that’s $50 per month.
5. Now really scraping the bottom of the barrel: Climate control in the environment around the printer. Let’s say it costs 10% of what powering the printer itself costs, so $25/per month.

What do we get: 560+250+100+50+25=roughly $1,000/month.

The company is charging $14,000 for 25 days of printing (or valueing their service at approximately $16,800 per month). Quite a gap between revenue and cost.

Why? I asked that question directly, and from the answer I got, I decided that my service provider wasn’t really sure either. He told me they were still discussing some technical issues with the OEM, plus they weren’t sure about the machine load, in other words, how many orders they would get for this printer per month.

The last argument made me wonder: How many orders can you expect when your competitor can easily undercut you by 90%?. Isn’t this like cutting the branch on which you sit?

Obviously I couldn’t tell this service provider he’s got it all wrong, he should give me a discount of $13,000. That’s not how it works. But maybe, if we figure out where the “acceptable” rate of $23/hour originates, we can make it “unacceptable”, as not reflecting the state of the art, and thus get more attractive prices all around.

Let’s get our tools out

Traditionally, with an analysis like this, you start by dividing costs into direct and indirect, variable and fixed costs. In this case we’re lucky, because a major variable cost, the material consumption, has already been taken out of the equation, thanks to the openness of the service provider ($20/kg)

We can, for now, focus on the fixed and variable costs of the machine and see what costs are always there, whether the machine works or not, and which costs are variable, happening only when the machine is working. With that information we can work out a price per hour based on an assumed run-time.

A month with the machine working the whole month will cost:

Now a month with the machine not working at all will still cost:

Based on a scenario where the printer works for one month and stands idle the next, it has to earn $1,035+$610=$1,645 in the month it’s working, or $1,645/24/30 per hour, is $2.28 per hour.

Make that $2.50 per hour, and the 100 kg, 25 day job described above would costs $1,875, instead of the $14,000 in machine time they are asking for. The $23/hour is only explained in this way if we assume they planned the printer to work two weeks per year. That is absurd.

So I’m still faced with a gap of more than $12,000, and stuck.

The same issue, inside out

Our earlier approach came up with just $2.50/hour, so we are still an order of magnitude short. Maybe, by looking at the whole operation instead of only at only one printer and its attached costs, the riddle of “why $23 per hour?” can be solved.

I made a guesstimate of what this company pays each month, based on what I know about it from a visit, and started analyzing it as if I had their monthly bills in front of me. Looks like this:

Office: Roughly 40 square meters of space where the printers are installed, with a meeting room to one side and an office to the other, plus a shipment preparation room, which takes up an additional 20m2. So a total of 60 m2, at $120/m2 per year, makes $600 per month.

People in the office: 1 manager (the owner),1 sales engineer, so two full time equivalents (one FTE is 22*8=176 hours). That’s roughly $300,000 per year (taxes included, $75/hour), so $25,000 per month.

People in the workshop: 1 operator, $50,000/year, $4,200 per month.

Also in the workshop, 4 printers:

• One big-size FDM printer, see above, $30,000.
• One FDM printer for high temperature materials, $30,000.
• One professional FDM desktop machine, $6,000.
• One desktop SLA printer, $5,000.

Total value $71,000, financing cost per month (assuming the same conditions as above) $1,300. I propose to lump in all other equipment, for a cost of $200 per month, to make it $1,500.

Utilities: Given an electricity cost for one printer per month of $250/month, I think $750 for 4 printers will cover those, plus electricity in the office, plus other utilities, plus ICT, make it $1,000 per month.

A wild guess at material consumption: 120 kg of thermoplastics to feed the FDM machines, at $30 per kg on average, is $3,600, plus 10 kg of resin @ $200 per kilo = $2,000. Total $5,600 materials. I add this for completeness’ sake: They seem to be selling their materials at cost price, so it’s not going to influence the overall outcome.

To summarize:

$37,900 per month, or $32,300 without the materials. This is the amount of money those four printers have to make to allow the company to meet its expenses. How can this be split among the printers? Maybe by the ratios of initial investment: 42% (big FDM), 42% (HT FDM), 9% (pro desktop FDM) and 7% (desktop SLA)

According to this logic, the big FDM printer would have to contribute $13,566 per month, but that would work out as $30 per hour @50% “occupancy rate”. Maybe they count with a 75% occupancy rate? That makes $25 per hour, close enough to the $23 we started with to assume we’re on the right track.

Reloading…

This is now our assumption:

They try to recover all their fixed costs and overhead by appointing rates per hour to the various printers according to their respective price/total investment ratio. At an occupancy rate of 75% they will stay afloat. The huge minus is that big jobs, like the one we started from, will never be awarded, they’re just too expensive. This is a great pity, since big jobs cost the same in overheads as small ones.

Let’s have a look at those overheads, and why they are there.

The task of the two people in the office is, basically, to sell print services for $37,900 per month in materials and print time. Using our own DigiFabster statistics, that’s 250 orders per month, $150 per order on average. That’s -again based on DigiFabster statistics which consistenly show a conversion rate of models quoted to orders received of 10:1- 2500 models to be evaluated.

Assuming 15 minutes per model spent on repair, calculation and further processing, that’s 625 hours of work per month right there. Given that one full time employee (FTE) works 176 orders per month and they only have two people (a total of 352 hours), it just can’t be done.

They will need a better-than-average average order size, or a better conversion rate, or both. The big printer, obviously by and in itself, helps towards bigger average checks. At an average check of $300, with the same conversion rate, they can do it. It means 315 hours per month of quoting, out of the 352 available, and they will still have to handle the 126 orders they generate in the remaining 37 hours per month, but given a little unpaid overwork now and then, this seems feasible.

If this is what they need, what can they do to achieve it? What can they do to get bigger average checks?

A good first step would be to assign office overhead to each order. Let’s say they introduced an “order overhead cost” of $200 per order. They could then drop their price per hour for the big FDM printer to $6 . “My” job would then cost: $200 order cost, $2,000 material cost, $3,600 for print time, total cost $5,800. Add a 30% profit margin for good measure, makes $7,500, which I would gladly pay.

Over time people with smaller orders would go somewhere else, people with big jobs would come to them, which would improve the conversion rate and average check . They would still have to work unpaid overtime to get the orders out.

Maybe automation?

An alternative scenario, easier to achieve, is automating the quotation processitself. As seen above, at average checks and average conversion rates, the sales office becomes the bottleneck: two people just can’t handle 2500 models per month manually. Adding people is no solution, the sales target, now $37,900, would go up too -in increments of $12,500 per FTE (fast calculation: $12,500/$150 is another 80 orders needed, thus 800 quotes, thus 200 hours of work to be done in 176). Automation, however, could greatly improve productivity.

Automation of 3DP sales process from the ground up costs, according to different sources and specs, anything from $50,000 to $100,000. The minimum set of features should include:

• A tool for model uploading;
• A tool for file repair;
• A tool for geometric measurements;
• A tool for slicing;
• A calculation engine with an interface to set up basic pricing rules, to replace the traditional spreadsheet;
• An input-and-feedback interface (front office), to give the end-user the possibility to tweak his order interactively (quantities, materials chosen, needed execution) and inform him of the cost for his selection.

That’s the bare minimum, but building it, even if it is done by a third party, will take focus away from the current, very intensive work of receiving emails, generating quotes, accepting orders, etc. Even if enough money were available, it will be a long drawn-out project. Two years to get it right is what we hear.

Let’s say it will cost $100,000 to build such a tool. Even if they got a loan from the bank (with similar terms as they get for capital investments) it would cost them an additional $1,880 per month for five years after implementation. Also, that number doesn’t factor ongoing maintenance or updates to the software, which could add as much as 20% to the base investment.

A better option, instead of in-house automation, is to subscribe to a software-as-a-service (SaaS) like Digifabster. Given a price per quote of 0.50$ and an average conversion rate of 10:1, the “order overhead cost“ would drop from 200$ to 5$, which would completely change the picture.

The time spent on quoting 2,500 models and producing and shipping 250 orders per month would drop to the time still needed for bookkeeping, debt collection, and shipping, but handling 250 orders per month is certainly possible for a two-man team, especially since DigiFabster takes care of payments, invoicing, and status updates for the end-user.

This would leave the office team with time on their hands to do more interesting stuff -like full cycle projects (idea on a napkin-brainstorm-design-prototype-small series) or model optimization- where their expertise can add real value and can be paid for accordingly.

From that moment on, the mission of the office team changes: It’s no longer selling print services and materials for $37,900 per month: they can delegate most of that work to the software and the printer operator (whom they should give a raise!), while they focus on key accounts which want premium service and can pay a premium price for it.

Plus: having gained a correct understanding of the real costs of 3D printing, they can now use pricing more strategically, get more orders, get more equipment, improve their software integration to reduce time spent on orders even more. There will be no stopping them.

For more details on DigiFabster, click here.