At large companies, many CFOs are only peripherally involved in pricing decisions. It’s often different at smaller firms, where finance chiefs tend to wear more hats. But even there, it’s probably quite rare for pricing to be the CFO’s primary focus, as it is at Xometry.
In fact, the opportunity to instill more rational pricing into a market niche was a key formative element for the company.
A few years ago, Xometry’s two co-founders, chief executive Randy Altschuler and CFO Laurence Zuriff, were investigating business opportunities in the on-demand manufacturing arena. They were attracted by the high level of technological innovation in the space, particularly with respect to 3D printing (also called additive manufacturing).
The pair visited lots of machine shops and additive manufacturing service bureaus. “We discovered that they were largely mom-and-pop-type operations where a rollup of any kind would not make sense,” Zuriff tells CFO. “What we also discovered was that there was incredible price opacity in that marketplace.”
They observed extreme pricing differentials for very similar manufactured parts in different geographic locations, as well as variances from week to week or even day to day. That, of course, posed problems for the manufacturers’ customers.
Pricing “could change overnight, depending on shifts in supply and demand in very localized markets,” says Zuriff, a former managing director at investment manager Granite Capital International. “While that was probably pretty efficient on a local basis, it was inefficient on a national basis.”
The founders set about building a network of machine shops and 3D service bureaus. At present, they number 700-plus. The common denominator among them is their focus on handling low-volume orders — 10,000 units and under — for specialized parts.
Commercial operations began in February 2014. Engineers at customer organizations upload to Xometry’s website computer-aided design files detailing the specs for the parts they need. Here is where pricing comes into play.
Xometry has a team of data scientists that create and, using machine-learning technology, continually revise algorithms that govern price quotes. The algorithms take into account myriad factors, including materials, shapes, tolerance levels, and design features — holes, for instance.
“If you put a square hole in the middle of a piece of aluminum, the cost might be five to seven times what it would be if you drill a round hole,” Zuriff notes. “The distance of a hole from the edge of a piece of metal has a cost impact too.”
The scientists also perform computational geometry, breaking down the elements of parts’ physical structures, another important determinant of the pricing equations.
The Great Unknown
Based on those and other factors, Xometry will quote a price for the order — say, $100 per unit. Then it distributes the order to those network partners deemed to be a good fit for the job, listing a preferred price for the manufacturing, perhaps $80 per unit. The spread between the two prices represents Xometry’s profit.
The first partner to accept the job gets it — the speed with which a job accepted is crucial, as the customer, seeking to control its costs, is almost surely investigating other manufacturing options as well. However, that partner may quote a higher price. The $80 figure that Xometry established was merely the mean point in a probabilistic distribution of expected quotes from the partners, also determined by a machine learning-based algorithm.
Xometry thereby assumes pricing risk. It has already provided its price quote to the customer and has no control over what price the manufacturer will quote. If the latter price is above a certain per-unit threshold, the company will take a loss on the job.
“Imagine being a CFO when you don’t know what the cost of what you’re selling is going to be,” says Zuriff. “You probably have a general idea, but it’s not like you’re taking a SKU off the shelf that you bought for $1 and selling for $2.”
Xometry also may choose to keep a job in-house. Its internal manufacturing operation generates approximately 15% of the company’s revenue.
Rather than acting like a broker, taking fixed commissions on business it distributes to a roster of providers, the company acts more like an options trader, whose risk has a fixed upside and a theoretically unlimited downside.
In fact, Zuriff says, in its number-crunching Xometry uses math similar to that employed in the Black-Scholes model, which is used to determine the fair value of call and put options.
Customers that use Xometry’s services multiple times may gain trust that similar types of jobs will be priced similarly. And the pricing will, on average, be lower than what they’d typically get by dealing with manufacturers individually, according to Zuriff.
Small manufacturers often reject jobs they could do profitably out of fear that something will go wrong and they’ll end up with a loss, he says. “What we say to them is that yes, you’re going to make lower profit on some individual jobs — but you’re going to get so many more jobs that [your competitors] are unwilling to take because they don’t understand the risk.”
The manufacturers are vetted for quality in a couple of different ways. First, they take a test; Xometry sends them specs for a part that is fairly difficult to make. It pays them to do so, which Zuriff counts as a partner-acquisition cost. Second, a new vendor gets easier jobs for a while and must ship the parts to Xometry for quality screening. The quality scores are then factored into the pricing algorithms.
But lower-quality shops are also allowed into the network, because some orders don’t, for example, specify high tolerance levels.
Zuriff says he spends more of his time working with the data scientists than doing everything else. He provides them with feedback on actual production costs compared with the expectations embedded in the algorithms, allowing continued improvement of the algorithms. He has a strong background in probability and statistics, enabling productive interaction with the team.
They closely watch three variables that impact gross profit: “win rate,” the percentage of quotes to customers that result in orders; “take rate,” the margin achieved on jobs placed in the partner network; and “acceptance rate,” the percentage of jobs accepted by a partner within a preset amount of time after the job is distributed to the network.
Zuriff and the team conduct pricing experiments, adjusting prices to gauge both the elasticity of customer demand and the manufacturers’ sensitivity to prices for particular categories of parts. All of the resulting data is used to further refine the algorithms.
Another aspect of Zuriff’s job is managing relationships with banks to which it offloads receivables risk. The up-front cash Xometry gets in exchange for paying the banks a sliver of the receivables’ value is important to the manufacturing partners.
Most of the company’s customers are large organizations that can negotiate payment terms favorable to them. General Electric, for example, has 160-day terms. “The local machine shop guy can’t wait 160 days,” Zuriff notes. “He’s already bought the materials and made the parts, which probably took at least one payroll cycle. Another 160 days after he delivers the parts would be many more payroll cycles.”
The receivables-based credit lines from the banks are low cost because of the receivables’ high quality. In addition to GE, customers include BMW, the Department of Defense, and NASA. There are other customers, Zuriff says, but he declines to identify them.
Big Customers, Small Jobs
While its customers are large, they generate various small jobs that are appropriate for the manufacturers in Xometry’s network. For example, the company’s biggest vertical segment is aerospace. An airplane may have millions of parts, but an aircraft manufacturer might need only, say, 2,000 units of any individual part, depending on the number of planes it’s building in a particular line.
A mobile-device maker may make millions of device but need only a couple of thousand fixture units that hold them in place during the manufacturing process. Xometry also serves the medical devices field, where 3D printing allows customization of, for example, titanium inserts for knee replacements.
About 65% of the parts delivered to the customers are manufactured with 3D printing. However, because 3D manufacturing is much less expensive than traditional computer numerical control (CNC) machining, the latter accounts for 65% of Xometry’s revenue.
3D parts are also easier to price, and Zuriff expects that over time, with additional innovation, they will comprise more and more of the company’s business and allow it to scale. It’s been reported elsewhere that Xometry’s annual revenue is currently about $20 million.
The pricing complexity, he says, “is what makes the business interesting for a CFO. It’s not just plugging in numbers and getting an answer. It’s thinking about how all the various changes you can make in the pricing might affect your gross profit and margin.”