How did 3Com lose the Fast Ethernet battle to Cisco?
Andy Gottlieb a former Apple engineer turned 3Com Product Manager, helped explain the impact of the decisions the company had made on the product line-up in the 1995 - 1997 timeframe:
We did do a good job with layer 2 Ethernet switching in desktop Ethernet, Fast Ethernet, Gigabit Ethernet, and in a stackable form factor. But the biggest error we made was betting almost equally across FDDI, ATM, Token Ring, and Fast Ethernet switching in the backbone or chassis form factor, rather than putting most of our resources into Fast Ethernet.
Focusing on Fast Ethernet switching for the backbone is what I was repeatedly advocating to everyone in my role as the one-person systems marketing department – but I didn’t get up on the table and threaten to quit, or actually quit (and go join a company like Foundry or Extreme Networks) when they didn’t listen to me on this fundamental point. We needed this in early 1995 when Cisco was coming out with their 5000 [via Crescendo] switch.
For Gigabit Ethernet (GE) switching, at one point we actually had three different divisions within NSOps building stackable GE switches before we had produced a chassis GE switch (the CoreBuilder 9000). The "each division as P&L" model that Bob Finocchio promoted, coupled with the reality of 3Com’s distribution/ channel strength, meant that the Premises Distribution Division [PDD, via the BICC acquisition], the Switching Division [SD, via Synernetics] and the NiceCom Division [NCD, via NiceCom] all had stackable GE devices, before there was a chassis version. That Bob and Eric both allowed simultaneous building of three stackable GE switches was both understandable and inexcusable.
The Cisco Catalyst 5000 chassis (which over time morphed into the 6000 and 6500) had this dense Fast Ethernet switching long before we did. For about a year they had a far superior switching product versus everyone in the industry, not just us. We were far behind in having dense Fast Ethernet switching in a chassis form factor, and behind having Fast Ethernet switching in any form factor by about a year.
Some executives noted that when the divisions had met to review the needs for a big chassis backbone product, the consensus was that resources were not sufficient to fund it at the time. This was due in part to their diverse investing in supporting Token Ring, existing FDDI customers and technology, and ATM technology.
Eric was somewhat philosophical about this whole product development mishap. He said, “It no longer mattered what products and technology solutions we might have had in the late ’90s. What really mattered was market coverage, the number of enterprise sales people actively engaged in large accounts, feet on the street basically. This is what made the real difference and we were outnumbered and outgunned, by at least three to one.”
Dave Tolwinski, the Synernetics Division GM, offered his perspective on this pivotal moment to not quickly move towards Fast Ethernet switching. Clayton Christensen’s theories about the “innovator’s dilemma” at work seem to apply. Dave explained that much of his architecture was based on FDDI, and that it didn’t lend itself to easy porting to Fast Ethernet switching. With their great growth and success, they were slow to make the switch, and they had a larger installed base with FDDI. He agreed they should have built something to compete against Crescendo and Cisco sooner. Dave noted, “One of the challenges about an acquisition strategy is that the acquirers have to make certain that the architecture, the technical architecture of the product that they’re buying, is malleable to their strategic need. Because developers can be trapped by the architecture. So, we were trapped by the FDDI embedded in the core. NiceCom was somewhat trapped as they had ATM in their core. PDD was trapped because they didn’t build chassis products – their whole vision of the world was small stackables.”
Tolwinksi also raised an important point, noted by some others as well - divisional independence may have affected outcomes. While no doubt Cisco and 3Com both gave their divisions some independence, 3Com suffered from a federalist versus centralized state of affairs. Bug tracking systems were in multiple divisions, making a customer’s solution to a problem more complex. Products, whether real or perceived, did not appear as heterogenous to the customer. Not recognizing when the end of a product cycle was over – these all were issues to confront. Dave noted, “After a couple of years, you finish the product, and then there’s got to be an absorption mentality more like Cisco’s, in which you wind up cutting some guys loose.” In Cisco’s case, it “cut people loose” very successfully – via spin-ins that went on to invent great things for Cisco again.
A spin-in was also a clever way to save on R&D expenses. Funding a startup with technical talent in a separate company would appear as an investment, preserving pressure on profits and stock price. They would later fold the company back into Cisco, at great profit to the employees in that venture. Cisco did a fantastic job of this. Technology executives Mario Mazzola, Prem Jain, Luca Cafiero, Soni Jiandani were most successful at this, with CEO John Chambers funneling them over $2 billion over the years for these spin-ins. 3Com attempted the spin-in model with Intransa, an IP SAN startup experiment that failed, and Atrica Networks, which had huge VC investments that did not ultimately pay off as well.