Masterclass: The Revolution Masterclass on interpreting web data

Web analytics is becoming increasingly important to brands as they boost

their investment in the internet as a channel for communicating with and
selling to consumers.

Jim Sterne, president of the Web Analytics Association (WAA) and founder
of Target Marketing, defines it as the study of online experience in
order to improve it.

But, as web businesses have advanced, so the data they extract from
their web sites has become more sophisticated. Web analytics has sought
to help businesses track user activity and extract information they can
really use.

Sterne reckons the detail that can be collected online is unsurpassed:
"It's as if direct marketers can technically measure how many people
pick up their envelope, turn it over, look at the first and second page,
and send off the coupon. If you have that granularity, you have
knowledge."

However, using all this knowledge effectively is as important as
collating it. Marketers must know how to interpret web data properly if
they are to improve their efforts.

What's important?

For an analysis project to work, the starting point is to define what
you want to achieve, says Sterne. "I help companies make their sites
better, but they have to answer the question 'better at what?'. The
biggest challenge with web analytics is that it forces you to decide
what's important and that is the toughest thing in business."

Alain Portmann, global media director of interactive agency Webliquid,
says a lot of clients get frustrated because they don't know which
channels can maximise ROI. "It's a simple statement to make, but it's
difficult to identify the best-performing channels and partners," he
points out.

"What attracted me to the internet was the ability to be
accountable.

A lot of clients are doing good work at the initial level of raising
awareness, but they're not aware of what's happening at the point of
purchase."

Portmann stresses the importance of marketers defining their business
priorities, and says the type of product or service will determine their
approach. For example, an e-commerce site will have different success
metrics to a publishing or brand web site. The choice of metrics is
therefore vital in trying to understand what customers are doing and
this depends on your target actions, adds Portmann: "If your key
objective is to measure sales, you need to study the conversion rate
from points A to B."

Asking the right questions is essential, believes Tim Boughton, chief
technology officer at Holiday-Rentals.com. "You can be overwhelmed by
what you could ask and the process invites questions. We are a small
company, so we don't have a huge amount of time to answer questions that
may not lead anywhere." He reckons a good discipline is to only ask
questions about those areas you know you can change.

Stay ahead

Holiday-rentals.com is part of the vacation rental company. It has grown
steadily since it was established in 1996, but the past five years have
seen considerable growth, with revenue increasing by 60 per cent year on
year. However, as the sector has become more and more competitive, the
company has invested significantly to compete, especially in search
marketing.

"What's crucial for us, and one of the main reasons why we need
tracking, is to understand and improve the conversion of people looking
for holiday homes to rent to people making booking enquiries," says
Boughton.

Holiday-rentals.com has two distinct groups of users: advertisers, who
pay it to host details of their properties, and renters who find
properties on the site and book direct with these advertisers.

"We run a very large, paid-search campaign, and we have to judge the
conversion rates and bid prices carefully to make it economical," he
continues.

"Unlike most retailers, who can take direct revenue from the people they
attract from paid search, we rely on the knowledge that the renters we
attract make enquiries, which generates revenue for our advertisers. If
we do that enough, those advertisers will renew their advertising year
after year."

In its early days, the site had a basic web analytics package that
examined broad traffic numbers, but, as competition increased, the
company implemented a more heavyweight package, called VBIS, from Site
Intelligence. "It helps us to understand how people move through the web
site and gives a visual demonstration of the site," says Boughton.

However, faced with this power, it is easy to drown in data as there is
a tendency to collect without purpose, warns Portmann. He recommends
that once marketers have defined what they want to look at, they should
clarify the specific metrics relative to their commercial and site
objectives.

They should make sure their terms are clear as it is easy for clients to
accept web agency jargon without really understanding what they're
talking about.

"There are not many standards in this area," says Portmann. "Your web
agency might talk about the 'interaction rate' or 'committed visitors',
but are you sure what they mean? It could be someone who has looked at a
page or has a product in the cart."

Portmann believes web analysis should stay true to the principles of
traditional marketing and map the customer process through awareness,
recognition of need, consideration, research and purchase, and
post-purchase behaviour. It is possible to define these five stages and
the abandonment rate at each.

An important issue is to decide where your data will come from, says
Boughton. It can be collected from server logs or via client-side
tracking.

The latter can be more time-consuming and possibly more costly as it
involves tagging all your pages.

Holiday-rentals.com works off templates, so it only had to tag
those.

It uses client-side tracking because it provides a truer picture of who
is accessing the site by filtering out non-human traffic, such as search
engine spiders. It can also provide additional information such as
screen resolution and operating systems, which is useful. A separate
function looks at where spiders go on the web site to help improve
optimisation.

"This allows us to see whether spiders are reaching all the points we
want them to reach and if there are any dead ends," says Boughton. "We
have 50,000 pages and our content changes a lot. If it drops off the
cliff, we can act quickly."

Examine trends

Sterne warns that companies should not get too hung up on exact numbers
in web analytics as any number is an interpretation. Instead, the
important thing is to examine trends and see where they are heading. He
says: "Don't install analytics software and start cranking out reports
and expect wisdom - it's just data."

Portmann agrees. "Clients need to trend the data. You can track
seasonality and things like that, but you should forget about the depth
of information and think about relationships. There is an obsession with
standalone metrics, like abandonment rate, but it isn't meaningful until
you can relate it to something else, like the wider marketing
activity."

Boughton considers it impossible to look at web metrics in
isolation.

They should be considered as wider business intelligence, such as
revenue or number of orders, to put them in context.

"Information on how people use the site is worth nothing if you can't
relate it to invoices that were paid, ads that were renewed or HTML
emails that were sent," he says. "We are always looking to improve
conversions at each of these steps to increase the lifetime value of our
customers.

Only by joining together our paid-search, bulk emailing, ordering and
customer systems with the tracking database could we ever fully
understand our ROI."

When it comes to interpreting data, there is often a limit to what you
can understand through pure metrics. At that point, says Portmann,
marketers may want to undertake some qualitative research with customers
through an online questionnaire. "You can't really get to grips with
things like purchase intent or perception of brand through metrics. You
may see a trend like a 20 per cent drop-off, but you can only make it
actionable by finding the reason behind it. Web analytics arms you with
the information, but a decision needs qualitative data."

Portmann cites the example of Excel Airways, which launched a site in
2004. Traffic figures were strong, but sales didn't follow suit.
Webliquid therefore combined data from web site analytics, third-party
ad-serving and qualitative research to identify the problem.

"About 75 per cent of visitors to the home page were not going on to the
later stages of booking and one in two visitors were not completing a
booking," he says. "Yet, many of them were calling the 0800 number,
which pointed to an issue with the web site infrastructure."

By using an online questionnaire, the company discovered that about 30
per cent of visitors were purely researching prices. Others were
dropping off because of sale-related charges that were not clear earlier
in the booking process. These findings enabled the company to modify the
booking process and display special offers more prominently, which led
to a steady improvement in the conversion rate.

Boughton says companies should prepare to have their preconceptions
blown away and embrace any suggested changes. A key lesson for
Holiday-rentals.

com was that 74 per cent of visitors didn't see the home page. "That was
a shock," he says. "We assumed that everyone starts there, but the
majority found us through search. That changed how we prioritised the
work on the site."

Based on its previous assumptions, the company had been spending a lot
of time updating the home page, but now it had to ask what site visitors
saw first. "It was content pages, and that made us ask how many people
went away because they didn't know it was a larger site," says
Boughton.

As a result, the company has added a search navigation link, which only
appears if a visitor comes from an external search enquiry. This helps
to improve the abandonment rate.

However, it's a slow game and improvements can be very subtle, says
Boughton.

Holiday-rentals.com started by looking at its conversion rates. It split
the process into four stages, examined each page in turn, looked for the
biggest drop-off point, and then asked whether anything could be done to
counter it. It would then test the site for a couple of weeks to see if
results improved.

Massive increase

"The incredible thing is that there is so much of a win," says
Boughton.

"You only need to improve conversions by one or two per cent to produce
a massive increase in business. If you look at how much extra traffic
you'd have to buy to get that, it's huge."

Sterne says businesses must be realistic about what they can achieve in
a specific timescale. A big-bang approach seldom works. "With
incremental changes, you get a handle on it. The only time you'll want
to do a big-bang update is if there is a fundamental navigational
problem."

Part of the iterative process of using web analytics is learning from
small improvements and seeing which have the most impact. "Companies
need to pick the most important processes in their business and optimise
them one at a time," says Sterne.

With such an iterative approach, it is important to ensure there is a
wide appreciation of the relevance of your statistics. "You have to
invite more questions from everyone. It's quite easy to forget how
useful it is to the business," says Boughton. "We want to make the data
accessible to everyone, particularly those who don't quite understand
it. We are trying to whet their appetite. In implementing this project,
we've made absolutely sure that the information is not reserved for the
techies among us but is presented in daily, easy-to-swallow doses to the
decision makers within the business."

Holiday-rentals.com distributes a daily dose of data to its key
personnel, which includes things like daily visits, six-week trends and
the source of traffic. This information has proven to be a useful tool
for management as they can identify any discrepancies quickly, without
it being a drain on their time.

Boughton adds: "There was an incidence when our paid-search trend
dropped to zero in a day. We called Google and it turned out that it
needed faxed approval of budgets and its fax machine was out of paper.
If you only look once a week, you won't spot things like that."

Which just goes to show the importance of the human element. As Jim
Sterne says: "Web analytics is not really about the software; it's about
how smart the people who crawl through the data are."

MASTERCLASS PANEL

Jim Sterne is the founding president of the Web Analytics
Association.

His consultancy, Target Marketing, helps firms to measure the value of
their sites for improving customer relationships. He has also written
six books and produces the annual Emetrics Summit.

Alain Portmann is global media director at interactive marketing agency
Webliquid, which he co-founded after leaving itraffic/Agency.com, where
he managed media budgets for clients like BA. He has made diagnostics a
key part of Webliquid's proposition.

Tim Boughton was appointed chief technology officer at
Holiday-rentals.com in 2003, after a career consulting to blue-chip
firms on e-commerce and the web. Having led the firm into web analytics,
Boughton is now helping three US sites in the group to do the same.

VISITLONDON PROBES VISITOR BEHAVIOUR TO IMPROVE SITE

VisitLondon.com, the official site for the capital, has used web
analytics for several years. It decided to review its technology and
install a new system from WebTrends at the end of 2005 to better
understand customer behaviour and drill down to the individual user
level.

Rather than getting bogged down in what stats are available from which
tool, the firm based its decision on business needs.

"We looked at questions we regularly ask internally and came up with a
kind of wish-list of the things that would really help our
decision-making," says Jeremy Willmott, managing producer of
VisitLondon.com.

These questions included: 'what information is most important to
different user segments?'; 'how could the site be improved to help users
get to the information they require faster?', and 'what new content
areas should be added?'.

Willmott says it is vital to remember that the stats you get from web
analytics packages don't provide answers. "It's like you're getting a
thread to start pulling on. It's knowing the questions to ask up front,
so you can start to tease out data through reports. But, at the back of
that is the analysis of the data and what you get out of it.

"A lot of people take stats on face value, saying, for example, 'That
section is really popular, so let's move it up the site'," adds
Willmott.

He thinks there needs to be another level of analysis on top to
understand why these things happen.

VisitLondon runs regular reports using WebTrends to see what users are
searching for most; particularly if an item is searched for, but
produces no results.

"You can use those stats to feed into your editorial plan. For example,
ice skating was a very popular search over Christmas, so we put a big
ice-skating room and some content on to the home page."

The company is also using WebTrends to guide its usability testing, and
to improve its registration process and navigation paths, so as to guide
users around the site better.

TOP TIPS ON ANALYSING DATA

1. Make sure you have collected all the data you need in the first
place.

2. Make sure you know how each item of data has been derived or
captured, and you also understand any sources of error or assumptions
that are being made, such as visitor counts from web logs.

3. It is vital to cross-check your analysis. You can match web data to
back-office systems or take two approaches, but make it a rule to only
trust cross-checked results.

4. Don't try and guess what your visitors will do. Experience shows they
will often act in ways that you'd least expect.

5. Remember that users are real individuals and not 'unique visitor
counts'. You are analysing your business, customers and your audience,
not your web site.

6. Be focused on your aims. Decide what you're investigating and
investigate it. It's all too easy to waste time looking online for the
answers and learn nothing.

7. Don't ignore oddball results - it might hint at an important piece of
insight that could transform your business.

8. Don't embed business logic in the data-collection process. Just
record everything that happens, and then apply your business logic to
test your assumptions and theories on pure data.

9. Look beyond the site. Link online and offline data to provide more
insight and value.

10. Never underestimate what careful analysis can achieve.

Thanks to Malcolm Duckett, corporate VP of marketing & operations at
Speed-trap

CHECKLIST

Questions that should be considered when embarking on a web analytics
project

- Have you outlined your business objectives?

- How will an analysis of your user base help you to achieve these
objectives?

- Are you in a position to make any changes that are recommended by your
web analysis?

- Do you have support from all the departments that will need to be
involved in the project?

- Who will analyse your reports, come up with action points and drive
them through?

- Are you clear about the statistics you need to measure? It is easy to
get sidetracked and veer off into 'nice to know' areas.

- Are you realistic about what you can change in a given time period?
Remember, the business doesn't stop while you are undertaking web
analysis, so set a deadline for each project.

- Do you view web analytics as an ongoing project? The more you do it,
the better your web site will be.