The "Pepsi Challenge" and "New TV" analogies: Think about these when evaluating VDI solutions

When I talk to people about how they evaluate and ultimately choose a desktop virtualization solution, I often use analogies. In today's article I'll share the two most common analogies that I use and explain how they related to desktop virtualization.

When I talk to people about how they evaluate and ultimately choose a desktop virtualization solution, I often use analogies. In today's article I'll share the two most common analogies that I use and explain how they related to desktop virtualization.

The Pepsi Challenge

Coca-Cola has traditionally been more popular than Pepsi in the US. In the 1970s and 1980s, Pepsi famously ran an ad campaign where they conducted "Pepsi challenges." They found random people on the street and asked them to taste Cola "X" and Cola "Y" and to pick which one they liked better. It turns out that more people chose Pepsi rather than Coke.

Does this mean that more people like Pepsi than Coke? Not exactly. What is means is that more people like a single sip of Pepsi rather than a single sip of Coke. In his book Blink (which I highly recommend), Malcolm Gladwell argued that because Pepsi is sweeter it will almost always win a single sip test. But when it comes drinking an entire can of a soda, most people are sick of the sweetness of Pepsi before they finish the can, while drinking an entire can of Coke is no problem.

Surely Pepsi knew this which is why they came up with the concept of the Pepsi challenge in the first place. They designed a situation where the consumer could "pick" a winner but where they knew they would most often win. Did they lie? No. They just changed the conditions of the test.

The same thing happens in the desktop virtualization world. It's often the case that the product that does best in your initial test isn't the best for you long term. Take VMware View for example. It is simple to setup and tests very, very well. (Look how much we loved it during the Geek Week: VDI challenge. That was based on the experience of setting up a five user environment in a single day.) So VMware can talk about how awesome and simple their setup is, but in your environment you only set up a product a single time, so do you really care how easy or complex the setup is? (I'm not trying to get into a View versus XenDesktop thing here. Yes I know VMware claims that simple setup means their product is simpler overall. Yes I know Citrix claims that View has scalability problems. I'm just pointing out that ease of setup or how good something works in a pilot doesn't always related to how good something works in production.)

So when you evaluate products in your lab or pilot environment, it's important to test, ask, or figure out how everything will work. How many connection servers will be required for your final environment? If you have more than one, how easy is it to figure out which server a user session is on when they call for help? Does your vendor offer a capability to clone a drive? Maybe that worked great in the lab, but in production how many of these can you do at the same time? Can you do it during the day?

Going to the store to buy a new TV

Have you gone to the store to buy a new TV recently? It's amazing how many options are out there. Even if you decide on the type, size, and resolution you want, there are still probably 20+ models to choose from. So when you're in the store trying to narrow it down, what do you do? You look at color. You squint your eyes can get up close. You watch the same thing on two TVs side-by-side and ask your friends which image they like better. Then you buy the TV with the "best" image and take it home and hang it on your wall.

And you know what? Once you get that TV home it doesn't make a damn bit of difference which one you choose. At that point you're not comparing the "Model X" TV you bought to the "Model Y" TV you didn't—you're comparing the new "Model X" TV to your old blurry glass TV From 1992. Model X, Model Y, or any of the other eighteen options from the store would have all looked equally as awesome to you!

When you were in the store trying to pick a TV, you were comparing the images directly against each other. You might have been able to say that one was better than the other because they were right there side-by-side. And you know, maybe one was better than the other, but that doesn't matter. You'll never notice at home.

In desktop virtualization, the vendors often talk about whose remote display protocol is the "best." Citrix loves to cite Video Clarity test results that quantitatively "prove" the video over HDX is better quality than video over PC-over-IP. (Well, the specific video they tested on their specific hardware with their specific network configuration was better.) But does this matter? Are they asking the right question? Do users really care that some test equipment told them one vendors product has better video than another's? The Video Clarity test is like the little stickers that vendors put on their own TVs to brag about the features. While the stickers are technically accurate, Citrix is selling something that users don't care about. (To be clear, sure, users want good video. But we don't need a color spectrum analysis instrument to tell us which video is better. As long as video plays smoothly at the resolution people want, they'll be happy regardless of what the data shows.)

At the end of the day, neither of these analogies means that you should evaluate the products you want to use before you implement them or that you shouldn't make decisions based on features. But take these lessons to heart. Make sure that you're asking the right questions and testing and picking things that matter to you and not making decisions based on marketing materials that vendors have constructed to show why they win.

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

Thanks for the topic Brian.

I'm born 1971, so with you being a bit younger I still believe that we might be sharing the same generation.

While I am Aspergers  it was in quite anybody’s guess what drove my peers – be that of the same or indifferent – I surely don’t know. From all of my work-life and things IT, I have gathered that there is a certain stigma, perhaps a way of being, a way to think that is quite commonplace amongst us.

In all this, I beg for neither agreement nor disagreement. I ask: what is your thought?


You mean like a stigma for or against a certain brand that would lead people to want to buy or to avoid that brand, regardless of actual testing or results?


Brian, please understand I was merely attaching to the topic in a way to have some further understanding among us peers in IT.

To be clear, I have personally no understanding. As of the stigma, the way of thinking that gathers us all, I just wonder and query of the thoughts among us peers.

Please people, contribute.


HA!  I commonly use a similar TV analogy from when I was upgrading from a Tube to either LCD or Plasma.  The end result, it did not matter... it was still a major upgrade and vast improvement either way.

When I have customers asking me View vs. XenDestop, I come back to the TVs.  Glad to see its not just me!


You make a great point, Brian!

The current protocol debate is a wonderful diversion from many of the deeper questions around choosing a platform or platforms to host computing resources for your users.  It's a lot easier to choose the viewing protocol that looks nice than it is to struggle with an assessment of how a platform will scale/integrate/deploy.  You can hardly blame a poor administrator (much...) for leaping at the simpler decision.

While it may be heresy, I contend that there is no single viewing environment that's right for all user communities.  There is no single platform for computing that's perfect for all users.  As with other IT decisions, we can do better when we can optimize a solution for specific requirements.

There may be economies that force us to choose one path over another.  This is just optimizing for cost.

You make a valuable and often overlooked point!


The dumb continue to lead the blind (general comment not directed at the commentators or this blog) and still think this is a protocol comparison. Forget BS analogies that don't apply because that implies your are comparing similar things.

Instead move the conversation to common sense. Once you use some of that and focus on real world implementation the answers are obvious. How many View scaled deployments are there, across networks, WAN accelerators, secure remote access? Why? It's obvious....