Web performance for mortal companies

Web performance for mortal companies

In this interview with Mac Slocum at Velocity, Joshua Bixby talks about mobile web performance, the difficulties of data gathering, site speed and SEO, and easy ways that every "mortal company" (aka any company that isn't Google, Microsoft or Amazon) can optimize their site.


Video Transcript

Interviewer: So, the first question I have for you does, does mobile carry different user expectations particularly on the optimization and velocity front?

Joshua Bixby: I think the biggest challenge with mobile today is that there is very little public data to actually answer that scientifically, which is a huge problem for our industry. We, you know, if we look at where we’re at on the desktop/you know, / web the way we can see today. We – Steve and others have pioneered data, that data has gone mainstream, primarily because we have had tools to analyze it and now we’re making a difference and it’s becoming accepted. That same process has to happen on mobile and it’s got to happen a lot faster than it did in the desktop world.

Interviewer: Mm-hmm.

Joshua Bixby: So, the answer is we’ve done some preliminary studies, which suggest, yes there is a difference. I think along – between those study is actually what we need to see is that distance getting closure and we don’t have the data, you know, Apple has the data, other companies have the data and people aren’t sharing it. And it’s just so important for our industry to share, you know, so that we all were on the same playing field. So, I would say, I don’t know, I think yes, intuitively it makes sense that yes but I don’t have the scientific data to back it up, yeah.

Interviewer: Okay. So, this might be a previous question but are there mobile benchmarks that should be considered yet?

Joshua Bixby: I mean, I think we have to collect the data first before we…

Interviewer: Yeah.

Joshua Bixby: …start setting benchmarks. One of the things that we’ve seen in our industry is that these benchmarks are helpful but they also can hurt and they can hurt because if the benchmark is set too high than we are developing features or buying products that are too expensive because they don’t return the investment. So, we have to get the benchmarks around performance, we have to get the benchmarks around conversion whenever that means to you…

Interviewer: Mm-hmm.

Joshua Bixby: …and then we can decisions about where the benchmark can be. So, I would say too early but it’s never too early you start collecting the data. What is your curve, if you went from, you know, once, you know, five seconds to eight seconds how much more money do you make or…

Interviewer: Mm-hmm.

Joshua Bixby: …eight seconds to five seconds, how much more money do you make?

Interviewer: Sure.

Joshua Bixby: You need to be able to answer that question to make any rational decisions about where to invest in performance.

Interviewer: Now, you’ve been involved in some long-term research…

Joshua Bixby: We have.

Interviewer: …on acceleration and business benefits.

Joshua Bixby: Correct.

Interviewer: Can you share some of the results from what you’ve learnt so far?

Joshua Bixby: Yeah, absolutely. I mean, the first result is, it’s bloody hard to get customers to talk about the data.

Interviewer: Mm-hmm.

Joshua Bixby: That’s the first result, as I have to give bottles of wine away…

Interviewer: [Laughs].

Joshua Bixby: …I had to sit in meetings where I, you know, beg CEOs to say this is good for you, it won’t hurt you. So, number one that’s what I’ve learnt. The other thing that we’ve learnt and the most important thing is that for, you know, I like to call them mortal companies…

Interviewer: Mm-hmm.

Joshua Bixby: …not, not an organization where you can rouse an army of developers that will focus on things that might not have an ROI, you know, we’re – you just have the time. And that’s not to say Google, Yahoo, Amazon are wasting their time, it’s just to say that they can afford to invest in some of the future, where as some of the other companies can’t. What we find with mortal companies, is that there is a huge pay back for performance improvement.

Interviewer: Mm-hmm.

Joshua Bixby: We’re seeing, you know, one of our companies like AutoAnything, nine percent increase in conversion and then you have to times that by a six percent increase in cart size to equal with the revenue would be. If I can go to an executive and say I can increase your revenue, you know, your conversion by nine percent and your cart size by six percent by acceleration that’s compelling.

Interviewer: Mm-hmm.

Joshua Bixby: We see similar stats from JC Whitney, we just got a stat recently around search, we have a company Everyday Health that we work with, they have a very popular site called whattoexpect.com, where they found the number of Google searches – in another word the bot by acceleration, the bot was able to look at two times more pages in the same period of time. And they found, since they implemented acceleration but their search rankings are up 10 percent. So, 10 percent search rankings, you know, conversion improvement, cart size improvement, the chance that people are buying on first click all are directly correlated to performance, so really exciting stuff from my perspective.

Interviewer: Yeah, yeah, absolutely. You mentioned mortal companies, right?

Joshua Bixby: Yes.

Interviewer: Is there one initial thing that those mortal companies should be focusing on right now, is it a suite of things or is there just one thing that they could do, they could yield some of those benefits you’re talking about?

Joshua Bixby: Yeah. I mean, I think the first thing is to set some kind of benchmark, so to understand performance and when I say understand performance I mean really from the users prospective. There are so many tools out there and those tools are good, they tell you what it looks like from the elbow of the Internet over fiber from, you know, from some really high-powered ISP. What we really need is what – what we recommend to our customers is, you need to see what your user see.

Interviewer: Mm-hmm.

Joshua Bixby: And what your user see isn’t what your, you know, your database administrator looks at because they are probably in the same building as your server.

Interviewer: Mm-hmm.

Joshua Bixby: You need to get the perspective of your real user, so that will be the first thing. And then there is definitely some low hanging fruit, I mean everyone should have compression on, everyone should be, keep doing keep-alives. What we see with performance is that, there is some easy stuff, they get progressively harder and more expensive and progressively harder and more expensive. So, yes there is some stuff that absolutely ever site in a world should do, it would take you two clicks either in a configuration on a load bouncer or on your server and you’ll be faster.

Interviewer: Right.

Joshua Bixby: It’s the other stuff that is really hard.

Interviewer: Right. Now, last question I have for you. You ran a workshop, right and in a workshop you analyze an unsuspecting site.

Joshua Bixby: We did, yes.

Interviewer: Right. Can you show what the site was?

Joshua Bixby: Sure. We actually did the Velocity site…

Interviewer: Okay, great, that’s a good…

Joshua Bixby: …because we thought, we thought it would be a good – one O’Reilly is a customer…

Interviewer: Mm-hmm.

Joshua Bixby: …you know, they bought us. So, we had a relationship.

Interviewer: Sure.

Joshua Bixby: And two we actually – our intention wasn’t to embarrass the site. So, what we actually did is we took the O’Reilly site, made it worse than it is in the real world, brought it back to where it was and made it better, because the intention was to show that, you know, that was the process. So, yeah O’Reilly, the Velocity home page.

Interviewer: So, were there any big revelations from, from that process, do you feel like the people came away from that going okay, I think I understand this reference points now?

Joshua Bixby: I certainly hope so, I mean what we tried to do is provide enough data so that they could do that, you know, so that, you know, I think in 90 minutes when you’re presenting something that a normal organization would do in four months, although it doesn’t have to take that long. But if you were going to do it by hand this is probably a four-month, couple of person project, we compress it, so there is a lot of data. The feedback that we’ve got since then has been overwhelming the positive which is, this is helpful, you know, slide deck has been downloaded a few hundred times already and it’s couple, you know, day out. So, yeah, I think, I hope so, you never know when you are up there, right?

Interviewer: Right, right.

Joshua Bixby: You never know, but so far so good.

Interviewer: Excellent. Well, thank you very much for coming today.

Joshua Bixby: Thank you.

Interviewer: I really appreciate it.

Joshua Bixby: Take care.

Interviewer: Okay.