Four Years Later…

Today is my four-year anniversary at Plexxi. I was in New York the week before Christmas to attend an investor conference focused on security and networking. It was a two-day trip that I expected to go by quickly as it was full of meetings and dinners. A colleague and I met with a number of crossover investors, analysts as well colleagues in compatriot companies. In our very first meeting an investor asked “four years in, how has it turned out compared to how you thought it would go when you started?”

I rambled through a semi-coherent answer wondering if the person listening to me had any idea what I was talking about. As I was talking, a flood of memories about the past four years came back to me. Some events seem so long ago, it is almost as if they occurred at a different company in a different era. By the end of the day, I had answered the same question three times. At dinner that night, I saw an old work colleague that I have known for twenty-two years, but had not seen in seven years. That entire day in New York has been gnawing at me for the past two weeks it motivated me to write a blog post.

1. We will never sell to the government and education markets: Not on day one, but on day two I said that we will never waste time selling to governments or universities. Few market segments have a slower technology adoption process than those two market verticals. I was certain that our target market would be financials and cloud providers. Four years later the simple truth is I was wrong. The government and education market verticals are among our largest and most active.

2. Large enterprises (e.g. financials) are full of technology experts: In the summer of 2012 a colleague and I took the first Plexxi switches (post Alpha designs with Trident+ silicon because Trident II was not available yet) to New York City. We reserved a conference room in the city for a few days and booked two-hour private showcases of our technology (we did this multiple times that summer). Judging from the interest, I was certain we were going to find some early customers who were willing to try out some new switching technology. I was wrong. We met many technology scouts, architects, planners, etc., but this never turned into business. The one decision we made as a company was to walk away from this market after twelve months of trying to sell to it. Maybe twelve months is too long or not long enough, but in hindsight these end-users were a Bridge Too Far for a company our size. Navigating the political process of a large financial institution in the networking space was too difficult and dilutive to the company.

3. SDN is an overused and confusing term: All the SDN washing of 2013-2015 destroyed any hope of having a coherent conversation around SDN. I do not even try anymore. I just tell potential customers that we build a variety of switches, some of these switches we make and some are from ODMs, but they are all designed to operate as a system through our controller and because they can operate as a system, a Plexxi network has a different set of behaviors and values compared to a network using traditional protocols and loop avoidance techniques, which are a collection of independent systems. If the customer expresses no interest in such a system, I suggest we get an early start on lunch or dinner.

4. Sell to the storage and application teams: We tried selling to the storage and application teams in 2013. That failed. Never once did a storage or application team have the political power to tell the network team what technology to deploy.

5. Networking subject matter experts (SMEs) are difficult to sell to: Incumbent vendors, especially Cisco, benefit from an extremely loyal customer and channel partner base. I spent a lot of time over the past three years selling to the CCIE crowd and I pretty much determined that it is a waste of time. If there is a strong anti-Cisco force in an account, my bet is they are going to buy Arista. Just showing up and trying to win the order for the next ten ethernet switches when the customer has only bought Cisco for the past ten years is waste of time.

6. Most networking experts are not networking experts: When I am told a person is a networking expert, I have learned to carefully ask questions to determine the level of expertise. I think I know a lot about networking, but I am not a networking expert. Most people I encounter who consider themselves a networking expert are trapped in a time span between 1998-2007. It is very challenging to sell ethernet switches to someone who wants to talk about OSPF and Spanning Tree and does not understand that using a Controller of some sort as a steroid driven NMS platform and switches that talk to each other using traditional networking protocols will not going to change how the network operates. The network is defined by protocols. I often suggest to people to watch this video from ONS 2012 and not to get trapped that is about using OpenFlow, it is really about constructing a network that operates differently. It is about (i) managing network bandwidth as a pool, (ii) managing the needs of applications, (iii) not managing each network box, (iv) not being at the mercy of how traditional protocols interact and re-compute topologies which sometimes find the optimal state, and sometimes not, and sometimes they are fast to find optimal state, sometimes they are not fast and (v) in general how distributed state systems are really hard to tweak to create a desired and deterministic outcome. It is about designing a network that operates differently with the infrastructure and the applications.

7. Cloud Scale. Web Scale, Open Networking and White Box terms have become so conflated they have lost their individuality: 99.9% people who use these terms have no idea what they are talking about and when they do use them, 99.9% of audience have no idea of the context. Most of the people who opine on these subjects should try selling ethernet switches for a year or two. PowerPoint presentations and shallow white paper thought pieces do not make anyone an expert.

8. Social networking/marketing is not a replacement for in person sales: The people who actually buy my products are not on Twitter and they are not catching up on Facebook during the workday. It seems the only people on Twitter are other colleagues in competitive companies having some esoteric debate about things that people who actually buy stuff do not care about. We have gotten leads through social networking, but the truth is nothing replaces meeting with a customer in person whether that relationship is direct or through a partner, which is how Cisco maintains dominant market share.

9. Old school marketing takes time to germinate leads: We tried live morning seminars, webinars and email campaigns and pretty much determined that these methods of marketing work, but not as well as in the past. Most customers want to buy ethernet switches in the same way they make a large purchase. People do not want to speak to anyone in the beginning of the process. They want to do their research on-line, look at others who have bought your product and if it all looks good, they will then contact the company. A few months ago the CEO of a potential customer contacted me on email and copied his staff. He and the team had spent the weekend going through our web page and wanted to talk to someone at Plexxi. He emailed me over a weekend to setup a call. A few days later, he told me he had gotten my email address from the cover page of a presentation I had given in New York three months earlier that someone had sent him. It took 90 days for this lead to germinate as the content had to shift through a chain of people, but when we spoke on the phone they were well prepared with questions and within 30 days had met with our CEO and brought a team to visit Plexxi at our location.

10. What customers did become our market?
Earlier this year we did some win/loss analysis as a team. One of the conclusions that we reached was we do not do well when we just sell to the network team, or the storage team or the application team. The last thirty years of enterprise technology consumption have been about best of breed and subject matter experts (SMEs). Enterprises made vendor choices and consumed technology in silos. The storage team really had no say in the networking decision and the application team had no say in the storage decision and vice versa. These groups might report up into a common leader, but the decisions were very much independent of each other. Smaller enterprises that consume through the channel, relied on their partner to make these decisions, but the structure of the consumption model was the same.

About mid-2014 I saw the first changes to this consumption model. I had the standard conference call to introduce Plexxi to a potential customer. The call lasted 45 mins with a set of slides. It led to an on-site meeting. When we arrived at the meeting, we were introduced to a team that consisted of the storage SME, networking SME, VMware SME, compute SME, security SME and a leader who described the team as a cloud networking team. The team was not making siloed decisions, but was making vendor selections as a team to deliver a new infrastructure build they describe as a cloud.

I have now seen this pattern of technology decision-making and vendor selection begin to repeat itself in other customers. Customers are self-identifying as cloud builders, which is a term we are using at Plexxi for our target customer. There are some details on how we use the term here.

I think there are many reasons for the Dell/EMC deal, but one of the reasons is the early breakdown of the legacy consumption model structure. EMC sales teams know the storage team, but do not know the compute, app and networking teams. The Dell sales teams definitely know the compute teams and maybe the network and storage teams, but not the app teams. Cisco has the same challenge in that their relationships are with the network teams, but not the storage and app teams.

Referring back to point #7, the large web scale players built their infrastructure as a system, meaning they automate their entire infrastructure, not just siloed technology portions independent of each other. This is the change I am seeing in the market.

/wrk