(13min 56) Web analytics packages can track everything your site users do. So how do you make sense of it all? Is it really a case of information overload? And why pay for software when there are free tools like Google Analytics?
In today’s BTalk Australia Phil Dobbie talks the practicalities of web analytics with Conrad Bennett, Technical Services Director at WebTrends.
Add your thoughts in the Talkback section at the end of this post.
Coming up: 5th-9th of January 2009 is Web Week on BTalk Australia. Hear a range of experts talk about web design, measurement and marketing.
View all BTalk Australia podcasts here
Subscribe to BTalk Australia on iTunes.
- Today’s Transcript
Phil Dobbie: Hello, I’m Phil Dobbie and welcome to BTalk Australia. Today, making sense out of web analytics. Sometimes I wonder whether the web provides us with too much information. We can find out who’s been to our website, how long they’ve stayed, what they’ve clicked on, which is all great stuff but how do you make sense of it? Well, Conrad Bennett is the technical services director at web analytics company WebTrends. So Conrad, you know, it’s a simple question. How really do we make sense of it? Most people get burdened down with too much information don’t they?
Conrad Bennett: Absolutely, and one of the challenges is we primarily deal with marketing departments and they’re not used to having huge amounts of data available. So for them, it can be a real eye opener to have that much. Really the focus that we try to aim for is the objectives of the business and the site, so that you’re working backwards from your requirements rather than looking at a set of reports and deciding what you can do with them.
Dobbie: So out of those requirements are there a couple of common top-line statistics that you should look at just to give you an indication of how your site’s performing?
Bennett: There are — usually standard metrics for different site types that will work as a good barometer. So that might be length of visit, certainly if you’re a content site, you’d be very interested in the number of pages that somebody views. With other types of sites it’s might be more interesting to look at specifics like whether they look at product detail pages if you’re a commerce site, for example. So there’re some simple metrics that work for different site types and certainly they’re a general barometer of the success of the site — they’re a good idea. And then there more advanced metric depending on the business that you’re in and what the objectives of the sites are. So you start to look at conversions, things like “basket analysis”, booking processes, etc.
Dobbie: A basket analysis?
Bennett: Yes, if I’m an eCommerce site or I have eCommerce as an element of what I do, I want to know what products people view. What are they putting into the shopping cart? Which of those do they go on to purchase? And then as you get more advanced, you can start to look at affinity analysis — which products people buy together.
Dobbie: So that’s like the “people who bought this, also liked this”, sort of approach.
Bennett: Absolutely. So that’s the basis of you know, site-side products for recommendation. And that can be interesting because often, products will fit together that you don’t necessarily expect. I mean obviously, now certain products will but, you can get a lot of insight from that kind of analysis.
Dobbie: These are things that any store owner would like to look at as well, I guess, but the web allows you to get so much more access to information — information that you’d love to get in your bricks and mortar store.
Bennett: Absolutely. And one of the interesting things for me is that quite often we get the discussion about privacy of data on the web and actually when people walk into a store and look at products, they are on closed-circuit TV the whole time. It’s just that nobody generally sits and looks at that data. Stores will do analysis on, on how people move through the bricks and mortar store, but they don’t really have the luxury of having that kind of data. So they can’t see how many times have you visited over the past two months. You know, how often you looked at a particular product. And the web gives you that ability to identify and research behavior. So when people return to the same product time after time, because if it’s not something small like a CD or a book, they’re probably not going to buy in the first visit. So they may visit several times before they build themselves up to having made the decision and go ahead and purchase or book.
Dobbie: There are a couple of indicators that things aren’t going well. You know, the warning bells that you really need to change your site. Are there a couple of statistics that are really shining through as bad statistics?
Bennett: It’s a difficult one to say, because every site is slightly different. So there will certainly be areas of the site where the metrics will appear to be backwards. For example, if you’re in the customer support section of an online banking site, you really want to get people in and out as quick as possible. So there you’re looking for short visits where people get the answer that they want and then leave. Whereas obviously within a magazine site, you’re interested in people staying as long as possible and consuming lots of content. So again, the metrics will, will differ site by site. Typically though they’ve got a fairly standard theme which is length of visit, number of pages consumed — those kinds of things. And when those numbers are changing over time and not in the right direction, that’s really what you should be looking for.
Dobbie: Yeah, so you’re looking at trends.
Bennett: Typically trends are where it’s at. It’s very difficult to look at a site in a snapshot way and say, this is successful or not. I get very nervous when people talk about having absolute numbers for metrics. I’d rather see somebody telling me that they’ve got a three percent improvement since last month.
Dobbie: People do love to do that snapshot though, don’t they. I guess one of the dangers is that you really don’t know what’s a good statistic or bad statistic. Are there ways that you can look at benchmarks for example, for your industry and say, well hey, look, we’ve got a shorter-than-average visit or people are looking at fewer pages. Is that a dangerous step to take.
Bennett: It is a little bit. It’s a very common requirement. People very often ask for benchmarking information because they tend to obsess about what competitors and the rest of their industry sector are doing. But typically what we see is that looking at your own data over time is much more valuable because ultimately you’re making your own changes to the site. I’ve had a number of situations in which a customer has basically argued with me about the metrics we’ve provided because they’re not what they consider to be industry standard. And usually that’s because they’re more successful than industry standard. So people are expecting a five percent conversion rate for example and you tell them their conversion rate is eight percent, often they won’t believe you because the industry standard is four to five percent.
Dobbie: Right.
Bennett: So you can get very tied up on that. I know every business is different, so your conversion rate will not be the same as your competitor’s because your site isn’t the same and your products won’t be exactly the same.
Dobbie: So what do you say to those people? Hey look, let’s make your site a little less effective so that you fit in with the benchmark?
Bennett: Yeah, absolutely. Yeah.
Dobbie: Making a change to the site for improvements is the key be all and end all, isn’t it. That’s why you’re studying these analytics. So what sort of changes are you likely to make?
Bennett: It’s quite varied actually — typically what we’re seeing is that a lot of the analysis is around the kinds of campaigns that are bringing people to the sites. So where we’re actually making the changes is not on the site itself, it’s in the acquisitions — the messages that they’re delivering. Partly because they’re easy to change, because they’re fairly fast moving. If you think about things like email campaigns, a retailer will send out several emails a week. So it’s relatively simple for them to make changes there, rather than changing their shopping basket process, for example. But at the same time, there will be times when you want to make changes on the site. Nowadays, it tends to be done in a much more organised way. There tends to be more testing involved. I know five, eight years ago, it would be something that was relatively quick and easy to do — nowadays it tends to take longer. Because if you’ve got a booking process, for example, on a travel site — you don’t want to play with it too much. You know, you want to make changes and, and you know, measure whether they’ve been effective or not, but you don’t want to make fundamental changes not knowing what the impact’s going to be. So we tend to do things gradually. It tends to be step by step. We tend to test variance of changes, knowing that we can change back quickly if it isn’t successful.
Dobbie: Now there’s a lot of science to this isn’t there? I mean, it is exactly the same as the science of how products are laid out in supermarket and it can be time consuming. As you say, you’ve got to test it at every stage. You really need to base it on analysis. So is this something people can do in-house or are they better off getting audits from experts, studying that trend and analysis and getting them to tell them the changes they need to make, rather than trying to tackle it in house?
Bennett: Typically it’s best if it’s a combination of the two. But the real value comes in-house — those are the people that will get, over time, the experience of the metrics and the site and the business, and they’ve got a much better feel for where things should be going what the objectives of the business are. Anybody that comes in from outside is going to give you a good flavour for the types of metrics you should use to measure, but ultimately they can’t really tell you in that snapshot, whether your conversion rate is good or bad, whether your metrics are heading in the right direction. Because they can’t really see that. So a combination of the two is usually best and, and typically within WebTrends, we try to recommend our customers that, rather than having our consultants do all the work, we train them do the majority and they use the WebTrends experts as experts. So when they have new things that they want to try or metrics they’ve never used before, we’ll get involved with them to help, but ultimately you know, the site’s owner knows the site and knows the business much better than we ever will.
Dobbie: Now you mentioned earlier that marketers aren’t used to getting access to a lot of data, although, they fairly often are. It depends on the type of business, but subscription businesses have access to a lot of information. We also have access to demographics and the purchase behaviour of our customers. What we tend to do is cluster them together into segments so that we can manage them and treat each customer differently. We can cope with five or six different segments and understand how those people are going to behave. In the online world, we seem to talk about dealing with each person differently. I mean, should we be doing the same thing online and start to build segments based on online behaviour and customer purchase decisions?
Bennett: Yeah, absolutely. Typically that’s where we’re heading with the kinds of things we’re doing with clients. It’s very rare that an online business (and it’s exactly the same as an offline business) has the luxury of time and money to actually market to somebody one to one. So it’s very unlikely that you’ll be able to do that. That said, technology can help, but generally within analytics we’re looking at segmentations. We’re looking at grouping people together. Because your population will just be too big to do anything else.
Dobbie: Yeah. Now there’re a lot of web analytics programs around at the moment and some of them are free. Google’s got a pretty good slice of the market with Google Analytics, which is pretty comprehensive, so what can (here’s the product plug for you) what can WebTrends do that Google couldn’t offer, for example?
Bennett: Sure, the great thing for the web analytics industry actually that Google did what they did. I mean, they effectively took a product that was a few hundred dollars and made it free, but they’ve obviously developed significantly from that. What that’s done for the market in general is created a huge amount of interest. Because it’s amazing how many people will try something when it’s free, that wouldn’t previously spend a few hundred dollars to do the same thing.
Dobbie: Sure. And a great deal of understanding I guess as well.
Bennett: Yeah, and actually the big cost with any of these types of applications really is internal cost. Because that’s where you spend the time and the resources to use the product. So what you’re actually paying out for the product itself shouldn’t be the major part of the investment. So whether you choose Google or WebTrends or anything else, actually the internal expertise is really where you’re putting in the investment, if you like. Ultimately with any free product, there are going to be limitations in terms of things like scalability and in terms of things like SLA and service, also things like ownership of data. So those need to be taken into account. If you’re doing something relatively small scale, no offense to Google, but quick and dirty in terms of a campaign perhaps, then that might be absolutely appropriate. If it’s something that needs to be scalable, it needs 24×7 support, if you need to get access to the data, if you need openness, which is a big thing that WebTrends is now working on, and you might need integration with other applications. That’s something that is much more difficult with an application like Google because it’s really just not set up to allow that kind of data transfer. So that’s really where we find that we spend a majority of our time. If it was simple bake-off between simple metrics at a very low level, then I absolutely wouldn’t recommend that somebody went with WebTrends purely because there isn’t any requirement to do that. I mean Google can give you those metrics very easily. And while it’s never going to be free, because that’s just the cost of the application, obviously that makes it easier for people to try it. What we typically find nowadays is we very rarely talk to somebody who’s not doing any web analytics at all. Because at the very least they’ll be using Google. But it’s very good for us because it allows us an opportunity when they outgrow it, to move to a more enterprise level application, to what we’re seeing now is that it’s becoming much more mainstream. People want to integrate with their offline data and their offline information, and for sectors like finance and, and telcos, where they’ve got huge amounts of data already about their customers. It’s obviously very interesting for them to be able to combine those. But those are big projects and that’s not something you can do overnight, so that’s going to keep a lot of organisations busy for quite a while.
Dobbie: Takes a lot of work to make stuff very simple, doesn’t it really?
Bennett: Absolutely.
Dobbie: Thanks very much for your time today, Conrad.
Bennett: Thank you.
