Seeking Alpha

In a previous SA article I wrote about Xyratex (XRTX), claiming that Xyratex's valuation is distorted and that the company has significant upside. I made this prediction based on the assumption that expenses for R&D will be substantially reduced going forward. This is based upon:

1. The intense objection of Baker Street Capital to development of new products before the company's profitability profile is restored.
1. Over the past three years Xyratex invested a fortune in ClusterStor. However, once this product is launched, the company will be able to support ClusterStor's R&D needs (new models/updates) with a lot less expense.

In drawing the profile of Xyratex's R&D division and examining its efficiency, it becomes clear why expenses for R&D will be reduced going forward.

Measuring R&D Efficiency

Measuring efficiency of R&D is not a simple process. The ultimate output of R&D is new products/new models and the revenue they generate. If we could determine the total expenses involved in each R&D project and then compare that figure to the sales that product generated, we could come up with a simple number for Net Present Value (NPV). If we had this figure for each R&D project we would have a complete picture of the effectiveness of R&D. However, without access to that data (primarily due to the need for secrecy before launching a new product), what alternative do we have? First of all, let's try to get a clear picture of the R&D division at Xyratex.

R&D Division

Xyratex's annual report usually contains data about how many employees the division contains, how many already launched products are being upgraded, and how many new products are in development. Summing up the figures from the last 7 years in the table below we see:

 Year 2005 2006 2007 2008 2009 2010 2011 2012 R&D expense \$54.3M \$71.3M \$77.5M \$85.8M \$71.0M \$92.7M \$115.5M \$103.6M R&D Employees 386 401 421 508 435 556 524 495 Core Product Projects 27 30 38 49 33 61 52 53 New Product Projects 10 16 7 11 12 13 15 18 Average Project Expense \$1.47M \$1.55M \$1.72M \$1.71M \$1.58M \$1.25M \$1.72M \$1.46M Average Expense Per Employee \$140K \$178K \$184K \$169K \$163K \$166K \$220K \$209K Average Employees Per Project 10.4 8.7 9.3 10.1 9.6 7.5 7.8 7

Development of products currently in production, considered "core products" are projects that support the current product line; this means development of updates or new models of current projects. Development projects for new products entail the development of new technologies or new products. From the above table, we can assume that approximately 35 projects per year are initiated to keep the company's products competitive. How much money was invested in those new products projects?

 Year 2005 2006 2007 2008 2009 2010 2011 2012 New Products Projects Cost \$14.6M \$21.8M \$12.0M \$18.9M \$18.9M \$16.2M \$25.8M \$26.2M

Average cost per project X new product projects

In 2011 and 2012, a sizeable amount of the R&D budget went towards ClustersStor. It seems that near the end of 2010 Xyratex recruited 121 new R&D employees. This reflects the fact that in 2010 there were 29 more projects than in 2009. In 2011 there were 22 more projects than in 2009, and in 2012 there were 26 more projects than in 2009. As soon as the first ClusterStor product was shipped, it is reasonable to assume it required R&D projects to keep it up to date. We can take the 2012 figure of 53 core product projects as an indicator of how many projects are required to maintain ClusterStor and Xyratex's core business. The cost of those 53 projects seems to be about \$77.3M per year.

R&D Efficiency

It would be wrong to say that an R&D division is as effective as the number of patents it generates, since some patents can support a product that generates hundreds of millions of dollars in revenues, while another patent can stay "in the drawer", forever unused.

Therefore, it seems more accurate to measure efficiency by how much money is invested in R&D (input) and compare that figure to the total revenue a product generates (output). A great article from the Harvard Business Review by Professor Anne Marrie Knott claims that the old-fashioned production function:

should be replaced by:

where R stands for R&D. Gamma stands for the change in output (revenue) that a 1% increase in R generates. I prefer to look at R as an innovation metric instead of just the numeric value of R&D dollars. A lousy engineer earning \$100K a year doesn't produce the same output as an engineer working for Apple (AAPL) who happened to earn the same amount, yet his ingenuity helped to develop the iPhone.

That said, how can we know if Xyratex R&D expenses are effective? Let's look at the ratio of R&D in a certain year to revenue generated 2 years later.

 Year 2005 2006 2007 2008 2009 2010 R&D/Revenue generated 2 Years Later 17.1 14.7 11.1 18.6 20.3 12.5

The decline in 2006 and 2007 can be explained by the recession of 2008. We can see that in 2008 and 2009 that efficiency metric jumped significantly into the 18-20 range. The year 2010 shows a marked decline in efficiency to 12.5, which decreases to 7.2 when running the same calculation on the second and third quarters of 2011.

Those numbers clearly reveal that Xyratex's expense for R&D have generated much less revenue during the last 2 years. The reasons for this could be:

1. The ClusterStor development project had a longer time horizon. Revenue from ClusterStor is just now starting to flow.
1. Xyratex expended high-quality resources trying to save the NetApp (NTAP) business through innovation. Needless to say, they did not succeed.
1. Xyratex started a lot of speculative R&D projects in the attempt to compensate for the losses entailed in the NetApp fiasco.

What's Next?

The best thing for Xyratex to do now is to "kill" all new product development projects except those aimed at new SMR and HAMR technologies (HDD technologies that will enable higher disk aerial density. Those projects are crucial for securing future HDD capital equipment revenues). If we assume the entire increase in R&D expenditures from 2010 to 2012 came from ClusterStor, then Xyratex invested over \$100M in developing the ClusterStor business line (excluding acquisition costs). Considering this as a sunken cost, future investment in R&D for ClusterStor is necessary to allow Xyratex to experience growth in its market share of the HPC field by ensuring that this product stays competitive.

If Xyratex "kills" 90% of its new product projects (about 16 of the 18 new product projects), it can save \$20M a year and still maintain a big enough R&D budget to support all three of its business lines: ClusterStore, storage systems, and HDD capital equipment. This will allow Xyratex to grow and return higher profit margins in the coming years.

Saving \$20M a year in R&D equates to firing tens of employees, an expensive and painful process, so scaling back on the R&D division may take a few quarters.

Conclusion

I highly doubt it that Baker Street Capital will allow the company to invest in new products until the company shows higher profits. At mid-term, Xyratex products can support enough growth to offset the loss of the NetApp business. After NetApp will no longer be a customer of the company, Xyratex could start to realize top-line growth. In the longer term, Xyratex will have to invest in new products and technologies, but this "experience" with Baker Street Capital will teach Xyratex an important lesson in R&D efficiency. The next time that Xyratex thinks about where to throw its R&D dollars, it will think twice when choosing its target.