Xenakis on Technology

Productivity Improvement

Written by John J. Xenakis for CFO.com, Dec 20, 2000.

A software company has come up with a system that it says will make factories more efficient, provided you can afford the million dollar price tag.

They used to say, "Build a better mousetrap, and the world will beat a path to your door." But before that happens, you'll have to convince the world that your mousetrap is not just better, but so much better that they'll be willing to pay a premium for it. That's the problem facing Maxager Technology (http://www.maxager.com). Its decision support software has substantially improved factory output for a few customers, including Motorola Computer Group, Seiko Epson, and Wheeling- Pittsburgh Steel, but it has not yet obtained wide recognition and acceptance. In all, Maxager currently has 14 customers in various stages from trial program to full production, about half in North American and half in Japan.

One site in full production is Motorola's huge factory in Tempe, Ariz., which manufactures computer boards. The factory's seven production lines run around the clock.

"We produce 100,000 boards per month," says Dan Lombardi, director of operations at Motorola, "with 1,500 different part numbers per month, with all different customers and selling prices." All in all, several hundred million dollars of products roll of the Tempe assembly line every year.

The 100,000 units represent a 30% increase over last year, and Lombardi attributes that increase to implementation of Maxager's application software. A 30% increase on a base of several hundred million dollars is a very great deal of additional revenue.

The methodology is complex, but works as follows: In any factory, each manufacturing assembly line consists of a series of steps  preparing raw materials, assembling, heating, attaching, cleaning, trimming, painting, and so forth. Crucial to the methodology is the identification of a specific step, a "strategic control point (SCP)," which constrains the entire assembly line. Identifying the SCP in each assembly line is the basic starting point of the methodology.

Any step in production can be a bottleneck, or the step that slows the entire line down. If you're short of raw materials, it could be the first step; if an employee is out sick, it could be a manual assembly step down the line.

But the SCP is a strategic bottleneck. It's one that doesn't come and go, but is in many ways the heart of the entire production line. In most cases, it's a very expensive machine or asset.

The thing that differentiates the SCP from an ordinary bottleneck is that it's a step in the production line that can't be easily expanded. If the problem causing today's bottleneck is a shortage of raw materials or an employee illness, managers can add materials or another person.

But if your bottleneck is a machine that costs $2 million, then you can't expand production at that point without purchasing another $2 million machine, and that normally isn't practical.

Once the SCP is chosen, that's where the controversy starts, because Maxager's software computes manufacturing costs and profits in ways that differ from conventional financial systems. What's more, these new numbers lead to different decisions about such things as factory loading and product mix, which can result in some political friction.

In conventional financial systems, manufacturing costs are computed by adding together labor and other variable costs at each step, and adding on fixed costs via a formula incorporating the time spent in production, factory space used and other measurements.

"The ways that costs are determined by finance are steeped in tradition, and they don't always give you the level of detail you need," says Jack Maynard, analyst with the Boston based Aberdeen Group. "The finance department or the cost accountants do a lot of averaging, coming up with average costs of goods sold, for example."

That's good for a general picture of the company, but it doesn't always help make the best decisions. "For example, some studies show that some high cost products that are complex to produce make less money than other low costs products that are easier to produce," says Maynard. "The Maxager system provides the granularity of detail to determine true costs and profits."

There's another big change that Maxager implements: Maxager treats labor as a fixed overhead cost, rather than a variable cost.

"Labor is not a truly variable cost," says David Shucavage, director of consulting methodology for Maxager "It's fixed, since you can't send someone home an hour early. With union contracts, your goal is to make as much money each day with the labor you have."

Maxager then allocates overhead, including labor, in a specific way: according to how much time product manufacturing spends in just one single step: the SCP. Other steps and other factors do not take part in the overhead allocation formula, as they would in standard financial computations.

For example, if a product spends five hours in the SCP, and the SCP is running continuously, then that product is allocated 5/168 = 2.98% of the production line's overhead, including labor.

This is quite different from the way things are usually done, and a lot more precise, according to Shucavage, who used to be a production manager before coming to Maxager.

"People often lie and cheat to promote the products they want to see made," he says. "The numbers that represent the standard costs are wrong because they're old and because there's game playing and politics. If you don't have the right data, then you don't know what's really going on."

For that reason, many companies installed data collection systems at their manufacturing workstations, in order to implement real time activities based costing. For companies that haven't already done that, Maxager installs real time data collection.

Using all this data, Maxager's software then makes a number of computations -- and Maxager has obtained some software patents on these computations -- to implement the methodology's cost and profitability formulas.

The software computes a particular metric for each product, called "profitability per minute," based on the profitability of the production line divided by the amount of time that product spends in the SCP.

According to the methodology, you change your product mix to favor products with the highest profitability per minute, according to Michael Rothschild, President and CEO of Maxager. "When you look at the world from a time based measure of profitability, as opposed to the stagnant snapshots of profit per unit as traditional systems do, you find that there are many products that have the same unit costs, but some generate very high levels of profit per minute, and some are very low."

"You have to understand what you're using Maxager for," explains Rick Prohammer, a partner with Arthur Andersen LLP who is responsible for Andersen's alliance with Maxager, and who has been involved in several Maxager implementations.

"It's the ultimate decision support system to help you figure out how to load your plant. It implies cooperation of the sales people, to sell the right things, so that they allow the fixed asset investment to sell as much product as possible. You have to avoid the situation where the sales people sell products, and then you have to make the stuff to fill the orders as well as you can. Maxager is tool to get them focused."

Motorola's Lombardi puts it differently: "The way we look at it, Maxager is the tool," he says. "Maxager likes to say that they give you $2 million or $10 million or whatever in savings, but the reality is that the tool only gives you information. You have to make the necessary changes."

Why does this methodology even make sense? Should a company really make decisions based on "magic" numbers computed by Maxager's software in a way that few people understand?

And granted that the SCP is an important step, maybe even the most important step in the production line, but does that really mean that every other step in the line should be so completely subordinated to that one? Isn't the reasoning a bit circular? Aren't we making the SCP into the most important step because we believe it should be the most important step?

Setco Corp., an Anaheim, Calif., manufacturer of molded plastics, and a division of McCormick & Co., ran the Maxager software in a pilot study that lasted six months, and ended in a decision not to continue. "We found that our manufacturing processes were pretty clean, and the bottom line was that there wasn't a lot of additional money to be made," says controller Hal Hendrix.

The Setco factory production lines had characteristics that did not match the ideal situation for implementation of Maxager software: the factory did not have clearly defined strategic bottlenecks -- the SCPs -- and the factory is not running 168 hours per week.

According to Prohammer, there are two major criteria which decide whether the Maxager technology will be successful in particular plant. "It works for asset intensive manufacturing operations, something like a steel mill or a paper plant," he says, "and though it's risky to make broad generalizations, this is as opposed to a chemical operation which is a fairly simple mixing together of materials. This won't be asset intensive, and to add capacity, it's fairly simple just to get some extra warehouse space."

Prohammer's second criterion has to do with the different types of products manufactured at the plant.

"It thrives on variability," says Prohammer. "It works if there's a wide variety of products that flow out of the manufacturing operating. For example, a company that makes only ten products -- even if it's asset intensive -- may be a tough fit [for Maxager] since there's not enough room to maneuver. Once you get to a hundred products or more, that's where you hit the sweet spot of Maxager."

Implementation

Putting all this together, there are several steps to adopting the Maxager methodology, and implementing the Maxager software, and all of them are difficult:

Analyze each of your production lines, and determine the Strategic Control Point for each of them. If the production line has several different bottlenecks, choosing the SCP may not always be easy, according to AMR's Prouty. "It takes a significant amount of analysis to pick the right one, and a huge risk if you pick the wrong one," he says. "Also, in some plants, it can appear to shift between two or three different places, depending on how the company's product mix shifts."

Install real-time data-collection equipment. This is an important step by itself  regardless of the methodology is used. Many factories do not capture data, and financial executives have no way to be certain that product cost estimates are accurate. Maxager's algorithms use all this data to make its computations.

Use the financial algorithms implemented in the Maxager software to determine manufacturing cost and "profit per minute" for each product manufactured. In some cases these figures will be dramatically different from the figures that are being computed by the company's existing financial systems. This will raise political issues that will have to be resolved.

Use the cost and profit calculations of the new methodology to drive decision making on how the factory should be loaded - - what product mix should be manufactured -- and therefore what products the sales people should sell. That's a tall drink of water for any company. No one's going to jump all of those obstacles without a lot of management work and support.

This situation is made even more difficult by the price of Maxager's software: $500,000 for the software license fee for each plant, with an implementation cost of $100,000 or more, depending on the changes on the production line the software says is needed. (Some earlier published articles on Maxager indicate that Maxager is charging 1% of the factory's revenue for its software. Maxager indicates that this idea was briefly considered at one time, but has been abandoned.)

This cost may be relatively insignificant for the Motorola plant, when compared to the 30% increase in revenue obtained for a plant which generates several hundred million dollars in revenue per year. But for smaller companies, that's a lot of money.

Recommendations

I spoke to numerous people for this column, always trying to get an answer to the following question: If this methodology is so great, how come more companies aren't using it? After several years in business, why does Maxager have only 14 customers?

The answers that came back were variations of the following: This is such a radically new way of measuring costs and profit, that it causes political problems between the CFO's office and the production managers.

In my numerous discussions with experts from the vendor and analyst community, and users, I reached my own professional conclusions about the technical validity of the methodology. I reached these conclusions based on my own background not only as a journalist but also as a mathematician (I studied graduate mathematics at MIT for several years) and as a systems analyst and programmer (which I've been for several decades). And I reached these conclusions about the overall methodology and approach, but without having studied the gory, nitty-gritty details about how Maxager's software performs each individual computation.

In my professional opinion, the Maxager algorithm should work in cases that fit the criteria we've described - where there's an expensive asset which serves as the SCP for the entire production line. Under these conditions, the Maxager methodology and algorithms should produce substantially better numbers and better decisions than traditional standard costing or activities based costing methods, and should produce the profit improvements that the vendors claim.

While I agree that there are political obstacles to installing this new methodology, my own belief is that the reason Maxager hasn't taken off more is because it's too expensive. The entry level licensing and implementation cost is almost a million dollars, which is a lot for a largely untested technology for even large companies, and is out of the question for smaller or mid-market companies which could never hope to improve profits more than a few hundred thousand dollars per year.

I'd like to see Maxager come out with smaller $50,000 versions of their applications, targeted to specific mid-market verticals. A model for such a system might be SAP's mid- market products. SAP's R/3 costs large companies several million dollars to implement, but SAP has products for $50,000 with built-in templates that target specific vertical mid-market companies. Maxager should look for a way to do the same sort of thing.

Companies wishing to evaluate this technology can start with a number of other resources.

The Maxager methodology is based on the Theory of Constraints developed by Avraham Y. Goldratt. The theory can be checked out at http://www.goldratt.com, the web site for the Goldratt Institute, and http://www.rogo.com, the web site on Theory of Constraints maintained by David Shucavage, the Maxager methodology expert quoted earlier in this article.

Some of the content on these sites tends to have an evangelistic flavor, but look particularly for the numerous Theory of Constraints user stories at http://www.goldratt.com/success.htm, including Ford, Lucent, and a number of other companies.

Can your company's programmers implement constraint-based algorithms in your own company's application software? Perhaps, but keep in mind that Maxager has spent several years developing and implementing its systems, and has refined the computations well beyond the ability of someone not full time in the field to duplicate easily. And, as we've noted, someone else might run into Maxager's software patents.

If you've done your due diligence, and you decide that your company might indeed see large productivity from Maxager's new mousetrap, then ask Maxager to set up a pilot or trial run. They've indicated a willingness to consider that option for promising prospective customers.

(This is a modified version of an article that originally appeared on Dec 20, 2000 on CFO.com at this location. )