Good Behaviour

This is an article I wrote for issue 176 of .net magazine in the UK.

I used to be Head of Online Planning and Buying at a London based media buying agency. I was there for 3 years between 1999 and 2002. In my first year our nascent online media planning and buying department experienced a 1000% growth in billings and some growing pains. Of course overall spend was much lower then than it is now as online media was also in its infancy relatively speaking.

Then in 2001 things slowed dramatically. At the time, growth in online media had been fed by new internet start ups with lots of VC capital looking to advertise to help grow their businesses and drive inexorably towards IPO! Additionally it was driven by a growth in interest from mainstream advertisers looking to dip a toe in and check the temperature.

Advertising is often considered a bellwether of economic decline as it’s one of the first things to be cut from budgets as belts tighten and when the slow down came in 2001 billings pretty much flat lined in our corner of the online media world, but other channels fared worse.

Part of the reason why online advertising may have fared better is due to much greater levels of accountability compared with other forms of advertising. Now consider the level of accountability we have with web analytics.

Back then in the early “naughties” web analytics was barely a twinkle in a webmaster’s eye, now it is proudly sitting at the boardroom table.

Not only can web analytics bring even greater accountability to on and offline advertising (if set up correctly) but it also completely opens up the level of business accountability for the website itself. It can be used to drive growth and cut costs through improved efficiencies across the whole spectrum of online communication.

If we are truly staring down the barrel of worsening economic conditions, especially looking forward into 2009 then arguably the worst thing any organisation could do would be to cut its web analytics budget.

Back in January I was working with a client that operates in an industry that is itself suffering but the saving grace for this particular client was their new website which had proved a great success in the face of a generally poorer trading climate.

If economic conditions deteriorate web analytics and the insight that it provides should be safeguarded and pored over with even greater intensity in the same way that normal business reporting and results are.

Analytics seems almost joined at the hip with conversion but improving profits doesn’t just have to be about selling more.

As times (look set to) get tougher improving operating efficiencies can be as important to the bottom line as improving on-site conversion rates. Spending money on valuable resource when it’s being under utilised is effectively wasted capital.

Thinking outside the web server

Improving operating efficiencies of a website is what web analytics was made for but it can also provide valuable information that can guide off-line operating strategies, an example of this would be in businesses running call centers.

Some businesses that have a presence both on and offline, and I refer not just to the “clicks and mortar” (hackneyed expression!) operations but also to others such as catalog retailers, may have found that their offline channel customers behave differently in varying degrees to their online punters. There are also those customers that come through both channels.

For potential customers driven by the web it’s not unusual to see day part website visitor patterns showing two spikes, one at lunchtime and one later in the evening after people have gone home and had supper.

For business that drive response to their call centers the online phone number will most likely be prominently displayed on the site.  A call me back form may also be provided so customers don’t have to waste their money waiting on hold & listening to mind numbing elevator muzac as they move up the queue.

Whether it’s via call me back forms or via the web only contact number, potential customers will pick up the phone at a time that will suit them but staffing call centers is not as easy.

Within in normal working hours businesses can run fully staffed call centers at standard rates but after hours rates for call center staff typically increases by 50% and between 200 & 300% on other “special” days such as Christmas and Easter.

If there is a lull in call center traffic at any point within normal working hours then staff will be under utilised and money potentially wasted. This becomes more acute if call center traffic peaks after hours when either there isn’t the staff to deal with it resulting in lost revenue or there is the staff to deal with it but the associated cost is higher.

To mitigate increased costs through up-weighted call centre staffing at peak times after hours it makes sense to re-coup as much as possible at other slack(er) periods.

Using web analytics to shed light on when web traffic is likely to put pressure on a call center can help improve efficiencies and bring down running costs while keeping customers happy and revenue flowing in.

Where the web is driving a significant contribution to revenue it can pay to keep the web analyst up to speed with issues that don’t necessarily relate directly to the website.

If you use Google Analytics and you’re hot on your SEO and your site has an internal search functionality then you may be faced with a dilemma.

This post is about the apparantly confilcting relationship between URL rewriting and tracking internal search using Google Site Search and why it’s an issue worth addressing.

SEO & Dirty URLs
Everybody wants to optimise their site so it appears high in Google’s SERPs. “Dirty URLs” produced by dynamic web pages are considered harder for search engine bots to read, additionally they are harder for visitors to understand so there is also a usability argument for cleaning them up.

URL rewriting and associated best practice rules effctively turns a dynamic URL with all its query strings into a nice clean easy to understand URL. For a (very basic) example: www.site.co.uk/?page=SearchResult&SearchInput=widget&search=yes into www.site.co.uk/searchresult/widget . As a result the rewrite URL will obviously mask the original URL. The practice of using URL rewrites has been gaining some momentum over the past few months so keeping this in mind now consider internal search on your site.

The importance of internal site search
Internal search is one of the primary ways a visitor will find something on a site. Collecting insight from internal search can help inform and drive action; it’s as if the visitor is marching up to the e-store keeper and asking where to find to ketchup - or whatever. Internal search gives valuable insight on what visitors are actually looking for, in their own words. As a result it can drive tactical and strategic decisions ranging from words and phrases used in the pay per click aquisition strategy to product placement and stock control.

So, both URL rewrites are popular as part of the grander SEO strategy and internal search analysis is important for gaining deep insight into a visitor’s wants and needs as they browse a site.

Google Analytics has recognised the importance of internal search by introducing an internal search tracking function to its product, it’s called Site Search. But like the funnels in GA it needs to be set up using elements from the page URL. In this case it is the search query parameter that GA needs and therein lies the rub. If dynamic URLs are hidden using URL rewrites in order to optimise for SEO, then it is not going to be possible to see the search query parameter in Google Analytics - even though the rest of the data will be easier to understand. GA support confirm the proplem. In their own words:

“To set up Site Search, you’ll need to enter the query parameter. You may not be able to set up Site Search if the query parameter is masked using a URL rewrite.”

The tyranny of the “or”…
It seems like a choice needs to be made, but naturally the best of both worlds is most desired. The answer should in fact be relatively simple. Avoide using URL rewrites for internal search results. By doing this the search query parameter will remain exposed and therefore should be visible in Google Analytics. As a result it can then be used in setting up Site Search tracking.

Getting this sorted out will depend on how the rewrites are set up and the content management system in place but in most cases it should be possible to achieve.

The object of this post has simply been to draw attention to an issue that some may already be facing and which may become more prevelant given the ubiquity of Google Analytics and the increasing usage of URL rewrites on sites with dynamic pages.

The solution should be fairly simple and although it means that uniformity across all URLs will have to be forgone, the ends will most definately justify the means. If there are any other alternatives to solving this please feel free to comment.

I may get shot (down in flames) for writing this post.

Web analytics is not just about data, this is well documented and blogged by far greater minds than mine - so I won’t get shot for that I hope! Web analytics is simply the engine behind driving better performance online. Better performance online for most organisations that actually engage in web analytics is usually about driving more revenue and improving cost efficiencies- and of course improving conversion.

99% of companies in the UK are SMBs and I think this is the great challenge for the web analytics industry. Many SMBs have websites and many of those websites perform a function, but the hard reality is that amazingly they don’t have the same size budgets as the average blue chip fortune 1000. They still need to invest to improve performance so they must approach their performance optimisation from a different perspective.

Usability is arguably a part of web analytics (2.0 as it has been labelled). There are many great usability experts out there and several different ways of approaching usability; these range from individual lab based usability tests, remote sample based usability tests using services such as Ethnio to journey replay solutions like Tealeaf.

To be clear, after looking at click-stream data and having identified where a problem might lie, if usability is what’s needed to unearth the truth then the methods just mentioned should be the preferred route; but they aren’t cheap.

A more cost efficient option would be to use a click based heat mapping product such as ClickDensity or CrazyEgg. These are not new products, they’ve been around for a while and they’re like click maps on steroids. They record clicks regardless of the presence of a link or not. They show the results either as actual clicks on the part of the page where the click was made or aggregated as a heat map. The advantage here is that where a standard link overlay will only record a click if it occurs on a link (assuming the tag is set properly) these tools will record click activity regardless. In other words, if a visitor reaches a page and attempts to click on something that looks like a link but isn’t, it will be recorded and show up.

So how do you get the most out of these tools in 5 ½ steps?

1. Assume a customer journey based on a task – making a purchase or signing up to an email
2. Replicate the customer journey as best as possible using a funnel or scenario in your analytics tool
   • Start with the most popular entry page
3. Allow enough data to collect
4. Identify main points of attrition (try and think why this might be happening i,e, form a hypothesis for each page where there is considerable drop off)
5. Look at the offending page using the heat mapping tool. The heat map will of course only show where your users have clicked but because it records every click there may be some surprises regarding where activity has and has not occurred and this could lead to action resulting in improved performance. For example there may be a high volume of clicks on a piece of text which has been mistaken for a link, this is potentially lost traffic and could go some way to explaining the drop off.

Any tweaks to the page that are implemented can subsequently be A/B tested to verify performance.

Cons

• Again, I should state, this is not the real deal in usability circles (don’t shoot!)
• You can’t talk to the people viewing the page and you can’t hear their thoughts as they navigate the page
• You can’t see cursor movements
• You can’t run the test with users instructed to carry out specific tasks

Pros

• For the budget conscious business it is much cheaper and more cost efficient. Even for large organisations it is a good practice
• The sample size includes everybody that interacts with a given page
• You can run A/B tests using these tools and compare your results instantly and run the best performing page.

This final point is perhaps the crux of it. The objective here is to amend the page design so that it makes life easier for the visitor and thereby unblocks the path to customer satisfaction.

This post has been written in the hope that it will prompt the more budget conscious business to think about how they can approach usability from a standing start. It’s not an attempt to provide a definition.

Please feel free to comment with your own thoughts and experiences.

Measuring conversion on an ecommerce site pretty much always involves establishing an overall site conversion rate. For most ecommerce sites this figure seems to hover around the 2% mark. Overall conversion rate seems like a hallowed metric – the one true “anchor metric” by which overall performance is measured. As a result there seems to be a slavishness regarding the site conversion figure. I think this can be misleading.

Firstly, to be clear, increasing the conversion rate is an excellent goal to aim for, but if it drops off it is worth considering two other performance metrics:

1. Total volume of sales
2. Cost per sale

What are overall sales doing? Are they increasing or decreasing?

Has there been a marked increase in marketing spend or spend in any other areas of acquisition?

Volume Vs conversion
As a rule, the more paid for (solicited) traffic driven to a site the greater the pressure the site is under to perform.

Demand generated traffic, by which I mean the likes of online display, tends to deliver visitors with lower levels of interest and a lower pre-disposition to take a desired action. It’s not surprising that this often has a negative impact on overall conversion. However, driving traffic to a site is crucial to growth for obvious reasons.

Affiliate marketing is a good case in point. Say an agreement is made between a client and an affiliate network (such as TradeDoubler or Commission Junction) with remuneration being established on a cost per sale basis, effectively the client’s acquisition (CPA) cost is fixed at a certain rate. The result is that the cost of driving the mix of volume and quality traffic is passed on to the affiliate network. The affiliate network’s strategy (to begin with) may be to drive a large volume of extremely low cost traffic, with the result that conversion goes down while sales volume is met and CPA is met.

Because the affiliate network is paid on every sale that is attributed to them they are incentivised to provide as many sales as possible at the lowest possible cost to them. There could also be an additional incentive if they reach a certain target volume in a given time period. In other words these combined conditions can give rise to a situation in which volume of sales increases while conversion rate falls.

I have seen instances reflecting this where conversion has taken a sudden dip while volume of sales remained steady; on further investigation it was traced to an increase in traffic which in turn was traced to one specific site in another country – a direct result of rogue affiliate action.

Cost per sale Vs conversion
In the case of an e-commerce site where the desired outcome is generally pretty simple by definition – sell something – all efforts can be said to contribute towards that one goal. Therefore all expenses incurred can be said to be as a result of pursuing that goal. It would be reasonable then to establish a cost per acquisition figure which should also act as a key performance indicator.

The issue becomes how to measure cost per acquisition and what to include in the cost of sales: advertising, site updates, internal labour costs, 3^rd party costs (professional services, systems etc), administration and so on.

The more included the higher the CPA. But, fixing the equation will at least allow the CPA to be used as a benchmark performance metric over time.

It seems to me that if sales are increasing and overall cost per acquisition is falling then performance can still be viewed positively despite a conversion rate which may also be suffering as a result.

By contrast, if the conversion rate is going up but sales are falling and cost per acquisition is increasing in part due to an expensive but highly targeted acquisition strategy which has a low yield then the increasing conversion rate can be misleading.

Doubtless this is all stating the obvious but it is done so in the aim for a more balanced perspective.

I’ve worked with quite a few e-commerce sites over the past few years and in our dashboard of metrics visits reaching a product page has always been there. This is for two reasons:

1. In (almost) all cases its possible to add a product to the shopping cart page from the product page
2. Because the product page naturally contains the greatest volume of information relating to a product.

So it has always been considered good to get visitors to the product page because shop owners stand a better chance of selling from the product page.

On that basis marketers often have a fairly obvious eureka moment and decide what a good idea it would be to land traffic directly on their product pages. No!

Product pages are just that, they aren’t landing pages. There is a significant difference and it is hard to find a page that will do both.

Because of this strategy it is possible for the “Get to product page” conversion metric to vary from anything between 40-75%. I think there are two problems with this.

Firstly this standard way of viewing this conversion metric doesn’t take into account traffic that was just dumped there by an adword campaign or a banner of affiliate campaign.

Secondly, by including visits that enter on product pages it masks the ability of the site to actually drive traffic through to these pages.

Funnels
Ah, but what about using a funnel to accurately track conversion to product page?

It depends on which analytics tool you use but many now, including Google Analytics, quite correctly (in my view) do not base their funnels on rules that require a visitor to have passed through the previous step in order to be counted I the next step. That means all visits reaching each step are counted – including those that enter directly.

Don’t forget bounce rate

How could anybody forget bounce rate. In this instance bounce rate queers the pitch because not only do these visits enter on a product page but they also disappear again immediately without doing the one thing we want them to. In other words not only did the site play no part in driving these guys to a product page but also they were totally disinterested in what they saw. Definitely not to be included in a performance metric like “get to product page”.

In short then, first know what you want to measure.

1. Is it the site’s ability to convert visits to a product page or…
2. …just the total number of interested visits that arrive at a product page.
3. Or do you really want to know the total figure, warts and all.

If it’s 1 then look at total volume of visits reaching the product page less the total volume of visits that enter on a product page.

If its 2 then look at the total volume of visits that reach a product page less the total volume bounce visits on a product page.

If it’s 3 then send an email, I’d love to hear from you.