|
If
you want to know when new interviews go online,
subscribe to the WebWord.com
Usability Newsletter!
Web
Site Optimization
An Interview
with Andy King (Author of Speed
Up Your Site: Web Site Optimization)
Source:
5-February-2003
Lyle the Usability Guru asks: If you only had an hour (in one
shot) to spend on WSO [web site optimization] for a given site each year,
what would you do for that hour, and why?
I'd pick the
low-hanging fruit. Eliminate excess (graphics, multimedia), cut your prose
in half, and optimize the rest. (you could install mod_gzip etc. in less
than an hour also). The main thing is to make sure that your home page
loads quickly.
Lyle the Usability Guru asks: Have new web technologies like XHTML,
CSS, Flash, and XML changed the WSO game or the ways search engines work?
Yes, CSS has made
it possible to transform table-based layouts into CSS-based layouts.
Typically this reduces page size by 25 to 50%. The ratio of content to
markup improves dramatically.
Lyle the Usability Guru asks: What's the ROI [return on investment]
for WSO activities? Can you give any examples? Another related question
might be "How do you build a business case for doing WSO?"
This is akin to
asking what is the ROI for usability. Speed is a key component of
usability. Small improvements in speed can take critical pages below
typical attention thresholds, and dramatically lower bail-out rates and
abandoned shopping carts. I talk about this in the book, but compression
alone can save 30-50% in size and bandwidth costs. Webmasters who have
employed compression and optimization typically save 30 to 50% off their
bandwidth costs, and retain more customers, and have improved conversion
rates.
Lyle the Usability Guru asks: Have you seen UIE's research
on users' perceptions of download speed? Doesn't that really debunk the
notion that code/graphic optimization improves usability? Isn't it really
all about 'scent of information' and users "feeling" like they
are consistently making progress?
Yes, I've read
that. That is poorly supported elsewhere, I spend the first chapter
showing why response times are important, summarizing key research into
HCI and response times. However, there are factors that can affect how we
perceive delays, like feedback and task complexity. Attunability is
another interesting area of research, with users adjusting
"subjective time bases" based on the pace of particular systems.
If Domino's usually delivers in under 30 minutes, and then one day took an
hour, you'd certainly notice it.
Lyle the Usability Guru asks: 528 pages!!?? How much
'optimization' did you do on the manuscript of this book? :-)
Yeah, my editor
kidded me over that. We did trim and condense a fair amount, but there is
a lot to cover, and I wanted a font size that everyone could easily read.
There is also a sample chapter from another book in the end. Next big
thing? I'd say delta compression and wireless optimization.
Mac - Dont Attack Iraq asks: Can a page load too
quickly?
As far as I'm
concerned, no. But according to the response time research that I read,
and mention briefly in the book, you can have response times that are too
fast, and this increases errors. But on the web, that is unlikely to
happen anytime soon.
ry rivard asks: How can the weblog be optimized to load
better but also show first-timers that work is being done?
Many of the
weblogs I view source have lots of embedded formatting, like font tags and
complex CSS classes. Many weblogs are also by their very nature verbose :)
I'd advocate using higher-level type selectors in CSS, contextual
selectors, and to be brief in decks and point to longer stories for those
who want to read more. Writing succinct headlines is also important.
In general, cut
your prose as much as possible, especially on high-traffic pages. Users
don't read as fast on the screen. On the web, users are information
foraging, trying to maximize the value of their time. They flit about like
hummingbirds, looking for nuggets that interest them. One study I read
showed that on average, users spend about 1 second per page, and rarely
stay more than 10 seconds. Once they find an article they want, they'll
stay longer.
Mike C asks: Does precompilation of loop limits in Web
programming languages like ColdFusion, Java, and JavaScript help?
Yes, this is also
called coding motion out of loops, and is one of Bentley's 27 rules for
code tuning. Also, using local variables is much faster. Many of these
refactorings are covered in Chapter 10, "Optimizing JavaScript for
Execution Speed."
Mike C asks: What are the downsides of stripping out every
non-printable space, tab, and line break from an HTML document, so the
entire code essentially resides on a single line? I thought Netscape 4 had
trouble with very long HTML lines.
I don't advocate
making your entire HTML page into one single line. Some editors can choke
on long lines, older versions of the Oracle info server can choke on long
lines, and if you email your pages (as we do at webref), some email
programs can flag a virus in longer lines. So I advocate a max of 255
character lines to be safe, or a max of 2000 character lines to avoid
problems with Oracle. Also, removing whitespace can break some JavaScript
code, and make your code hard to read. You can avoid these problems by
keeping unoptimized versions for any edits, and punctuating your
JavaScript statements with semicolons.
Joshua Kaufman asks: What is the most common problem to be
solved when optimizing web sites?
Too many HTTP
requests. This is due to the overuse of images and external files. We're
also seeing a trend of too many external JavaScript and CSS files in the
HEAD. This delays the display of your content as they must load first.
Joshua Kaufman asks: What major web sites do the best job of
optimization?
Yahoo.com has the
most highly optimized home page I've seen. They use URL abbreviation to
save nearly 30% off their home page HTML. View source to see what I mean.
But even Yahoo has bloated up, they have nearly 300 links on their front
page. WebReference.com of course :) I like most anything from Zeldman and
company, very clean and CSS-based. Though there's always room for
improvement.
Joshua Kaufman asks: There is software that assists and, in
some cases, automates the accessibility process. Is there any software
that does the same for web site optimization?
Yes, there are a
number of products, mainly for optimizing HTML and JavaScript. SpaceAgent
from Insider Software, VSE Web Site Turbo from VSE Online, and of course
automated graphical tools. I test and demonstrate many of these products
in the book. To convert to CSS-based layouts, and to do it right, you've
got to do it manually.
Rotwang asks: I could create a stunningly beautiful entirely
graphical page, or a simple page with no graphics, or something in
between. How do I determine the safe point; the point where a page is
acceptably attractive, authoritative and/or creates the right impression,
and yet loads fast enough to serve my customers?
It depends on the
type of site. For informational sites like WebWord.com or news-related
sites, the graphics should be kept to a minimum. With the advent of
widespread support of CSS, you can now create many pleasing effects
without graphics. I cite a study in the book on this ratio. For shorter
delays users prefer documents that include graphics, for longer delays
users prefer text-only documents.
Laurel asks: Do you feel that information architecture (in
this case I mean the categorisation of web pages for findability) can have
an effect on site optimisation? I suppose I'm asking if things like
intuitive URLs and labels can reduce the need for extra context on a page.
How would you separate site optimisation and usability/IA?
Great question.
Yes, there is a tradeoff for some techniques between IA and WSO, and with
Search Engine Optimization (SEO). Good IA has a logical hierarchy and
clear, unambiguous labels. Some WSO techniques uses short abbreviated
names and URLs to achieve savings, which can preclude descriptive terms.
SEO also can conflict with IA and WSO, with some sites sacrificing logical
hierarchy to create keyword-filled directory and file names. Balancing
these three disciplines is an art in itself. For high traffic pages like
home pages, I favor speed over IA and SEO. In the book, I discuss mapping
techniques that you can use to have the best of both worlds. The book
gives you the tools you can use to optimize your content. How far you go
is up to you.
Joe Clark asks: I have some long prose pages, and I wonder
at what point should they be broken into separate pages. I hate the
scroll....but I hate the little chunks per page. These will not be
shortened; they are the length they are. So, what is the optimal number of
words per separate html document?
Great question. I
haven't seen any studies specific to this, although I have some
observations. I assume you mean for readability and usability, and not for
SEO. For a multipage article, we found at WebReference.com that page views
dropped off dramatically after 4 or 5 pages. When we made an article
longer, fewer people read pages 6 or higher. Also, page length is related
to page size, and without feedback you need to make sure your pages load
in at most 8 to 10 seconds. That is about 30 to 34K total. With a 10K
banner and a logo say, that is a maximum of 20K.
Optimal line
length is another story. There are a number of guidelines on this, dating
back to the 1800s. Max optimal line length for print is 1.5 x the length
of the lower case alphabet. Jakob would trim this by 25% for onscreen
reading. This article from Human Factors International on optimal
line length found that people prefer moderate line lengths, but read
faster at longer lengths. Of course, when you make your line lengths
shorter, your page length increases, and people will have to scroll more.
THANKS ANDY!!
Important Note: All
content originally posted on WebWord by respective owners.
Dont miss Andy's
special web site: WebSiteOptimization.com
What Next?
|