For any business, two pieces of information are most important to its survival. One, in order to make decisions, a business needs to know the ground reality of where it stands in the market now. Second, in order to plan forward and determine progress, it needs to know where it stood in the market in the past. These two sources of information individually don’t convey much information. But combined together, they provide actionable insights. Where am I now and where was I is what you need to know if you need to plan for where I want to go.

Making parallels to web analytics, current set of tools (unfortunately) only provide information on what is happening now. Your favorite tool will churn out data on number of visitors, page views, countries, referrers and what not. This exactly tells you how your website is doing today. However, this completely misses out macro trends. Sure, you can see a historical graph of number of visitors and all sorts of other metrics but that is only the first step towards knowing what has changed.

Ideally a web analytics tool should go deep on the segment level and mine signals in the (historical) data and correlate different metrics automatically for you. Here are some of the examples that I expect an analytics tool to mine automatically for me:

• Accounting for other variables, correlation between Twitter activity and number of organic searches. (Which isn’t obvious but could indicate that I’m doing a great job on Twitter building my site’s brand)
• Show meta-trends in the drift of the type of traffic I’m getting, countries visitors are coming from or the time spent on page. Am I slowly getting more traffic from niche blogs as compared to Google? That tells me what I am doing is working with niche blogs and should probably do more of that stuff.
• Automatically deduce what content or website sections see atypical visitor behavior so that I can act fix it (if it leaks) or use it for further gains (if it works).
• Tell me top 5 common paths my visitors take on the website and if there has been a significant shift over time.
• Source data from all different inputs: social monitoring tools, newsletter tool, etc and automatically correlate my activities outside the website with what is happening on the website.

Most of what I have written above isn’t super hard. Some of it can be done by having simple heuristics built into the tool. Moreover, data mining and machine learning has progressed a lot and I am surprised web analytics industry has been so slow at adopting the methodologies. Though Google is taking the right steps with their intelligence feature but it still it leaves a lot to be desired: where are correlations, recommendations, trend mining and other interesting stuff? Nuconomy was doing the right stuff but they took far too long, didn’t innovate a lot and end up getting bought by a company for in house analytics.

Web analytics shouldn’t be simply a data collection and reporting tool. It should actually be collection, reporting and mining tool. My tool gives me 100s of metrics to look at which I can’t keep looking at day after day (unless it is my full time job). Instead it should mine all 100s of reports for me, and show me interesting nuggets on what has changed (and possibly what could change). So I ask: where is the innovation in web analytics? All I see around is dumb reports ready to get mind by a human.

What is your perspective on this? Do you think web analytics is ripe for a major change?

Posted in Conversion Optimization, Split and Multivariate Testing on April 15th, 2010 | 4 Comments »

Scientific research papers on how to increase sales or conversions are rare.  Most of the articles you read on the Internet (admittedly, including a lot of posts on this blog as well) are based on what the writer thinks and what makes logical sense. But scientific research works in a different way. Authors of research papers must produce accurate, reproducible results.  And their articles are reviewed by peers before getting published. Hence in most cases you can always trust results of a scientific paper.

So I fished out research papers on the internet which tell you how to increase your revenues online. Here is the list for your reading pleasure:

Usability

• Web site usability, design, and performance metrics [PDF, good paper with lots of information]
• Influencing the online consumer’s behavior: the Web experience [Good, detailed paper]
• The impact of initial consumer trust on intentions to transact with a website: a trust building model [PDF]
• The role played by perceived usability, satisfaction and consumer trust on website loyalty [needs subscription for fulltext]
• An empirical study of Web browsing behaviour: Towards an effective Website design [PDF]

Miscellaneous

• Practical Guide to Controlled Experiments on the Web: Listen to Your Customers not to the HiPPO by Microsoft Research [PDF, Good paper for A/B split testing internals]
• Predicting bounce rates in sponsored search advertisements by Google [PDF]
• How well does result relevance predict session satisfaction? by Yahoo Research [PDF]
• Generating Web Site Traffic:  An Empirical Analysis of Web Site Visitation Behavior by UC Berkley [PDF]
• Predicting online-purchasing behaviour [needs subscription for fulltext]

After compiling this list and reading most of the articles, I realized there is a large disconnect between what happens in research and what is actually applied in the market. Do you agree? Did you even know such kind of research happens at all?

Posted in Conversion Optimization on January 29th, 2010 | 6 Comments »

All right, let’s admit it: increasing conversion rate on a website is still a voodoo science for many. With new technologies and terminology being thrown around (on Twitter, blogs, etc.) every other day, it doesn’t get any easier for people just starting to understand conversion optimization. In this post, I will try to briefly talk about all technologies and methodologies being used today for extracting more juice out of existing traffic:

• Web Analytics – the first step is to actually measure conversion rate of your website. Tools like Google Analytics,  Clicky and Woopra make is quite simple to define your website goals and then measure the goal completion rate. Different tools may have different terminology but essentially every web analytics tool worth its salt with have ability to track conversions, knowledge of which is an absolute must. Think of it in this way: how can you optimize a metric if you don’t know what the metric’s value is. So if you are not measuring your conversion rate, you should forget about improving it.
• A/B Split Testing – these tools allow you to make different versions of your website in order to test which of it leads to maximum downloads, sales, signups or conversions. Main idea is to keep testing your ideas of what works and what doesn’t, instead of relying on intuition.  With introduction of easy to use tools such as (our very own) Visual Website Optimizer, split testing has become a very simple, yet extremely effective methodology. Nothing beats increasing conversions than testing your ideas against hard, cold facts of your visitor clicks. Google Website Optimizer is another tool for split testing which is available free of cost, though it does require technical understanding of HTML, page tagging, etc.
• Behavioral Targeting – this methodology requires extensive planning and segmenting of visitors into different groups to show customized offers/pricing/promotions. Ultimate idea of BT is that you  tailor  your website according to behavior of the visitor (e.g. number of pageviews, entry page, etc.). BTBuckets and our labs tools are the ones which let you try and use behavioral targeting for conversion optimization.
• Clickmaps and heatmaps – these tools let you visualize which parts of your website get most clicks and which parts don’t get even a single click. Such tools come handy when you have doubt that perhaps your call-to-action is not very visible on the page. What these tools will tell you that while your button isn’t getting any click, people are clicking onto your bullet points (which are not hyperlinked) in search of more information. So perhaps it is a good idea to hyperlink those bullet points. Crazyegg and Clickdensity are some of the most affordable tools  that let you generate a clickmap of your website.
• Visitor movies – tools such as Clicktale and Userfly give you a magical power of peering over your visitor’s shoulder to see what exactly they are doing on your site. What these tools enable you to do is to see each and every action by your website visitors. The tools will literally replay all mouse moves, clicks, keypresses that a visitor does.  Though it ultimately becomes cumbersome and tiring to watch 100s of visitor movies, this method is still a fantastic way to gain qualitative insights.
• Voice of customer – when every other method fails, you should resort to surveying your visitors:  why exactly they didn’t make the purchase, was your pitch clear, did they understand what you are offering. All other tools tell you what is happening on your website, customer feedback tools such as Kampyle or GetSatisfation tell you why it is happening.
• Live Chat – taking surveying users idea to next level, live chat tools (Olark and  numerous other tools) actually let you chat with your live website visitors. They can ask you questions or you can initiate the conversation. Having a live chat feature on your landing page and sales page can help boost conversions because visitors may have N number of doubts which your page may not be clearing; chatting with you can answer most of their concerns.

Apart from these categories, numerous other tools in search analytics, PPC analytics, affiliate management, etc. are available but the  above ones are the most useful ones. Even amongst the above, I will argue web analytics and split testing tools should be an absolute must for anybody serious about improving his/her website conversions.

Leave a comment here if you think you have additional toolssuggestions for conversion rate optimization which I missed here.

Posted in Conversion Optimization on December 18th, 2009 | No Comments »

There are number of metrics that startups and websites obsess on. Some of the most overanalyzed yet non-useful metrics are number of visitors or pageviews on a website. The reason startups get obsessive about them is that these metrics are easy to use and are no brainer. Just slap some code on the website and you are ready to get insights on your startup’s progress, supposedly.

In fact, there are three metrics that a startup (or for that matter, any website or online business) should single mindedly obsess on accurately measuring and hence optimizing for:

1. Monthly growth in number of paid users – what is the number of paid users you have this month v/s you had last month? Is the difference positive month after month? Focusing on growing number of paid users month after month should be top priority because it is what ultimately brings more revenue and gives you confidence to achieve even more success. Any growth less than 50% should be a cause of worry for startups because during a startup’s early days you are looking for hyper-growth. Eventually the number would stabilize as the business stabilizes but during the first year or so target for more than 50% growth for sure. And if by chance, growth starts going towards the negative territory, you should get alarmed and act accordingly.
2. Retention rate of existing paid users – how many of your existing paid users canceled your product/service this month? Is that figure decreasing month after month? This metric is complementary to the metric above (growth in paid users) because while the above metric helps you focus on acquiring new customers, this one helps you focus on retaining existing customers. If customer growth rate is high but retention rate is low, then no matter how many new customers you get this month, next month you are back to square one. If, on the other hand, you have good retention rate, you will build great momentum month after month. You should definitely go out of your way to retain existing customers. If they leave, do (polite) exit interviews. Ask them why they leave and fix it in your offering. A good retention rate that you should be targeting at would be 75-100%. This means no more than one fourth of your existing customers should cancel your product or service in a month.
3. Monthly Revenue – the amount of revenue you are bringing in month after month. This is a n obviously important metric but a lot of startups just don’t focus on it as much as they would like to. Instead they start focusing on number of visitors or other such proxies. Monthly revenue ties together the above two metrics plus one more important ingredient: cross-selling or up-selling. If you are acquiring new customers and your existing customers are getting retained, are they purchasing more stuff from you than they currently do?  As your startup evolves, you and your customers should move up the value chain and you should continually find new avenues to provide value to your existing retained customers.  And in the process, make more revenue per customer.

The three strategies of: a) getting new customers, b) retaining existing ones and c) up-selling and cross selling new offerings are not new. Management gurus have been discussing them for ages. Even then startups and websites get drowned in a flood of metrics and forget that they are there to make money.  They should better be optimizing how to make more money. And only way to optimize that is to focus on the right metrics. Do you agree?

Posted in Conversion Optimization on December 2nd, 2009 | 3 Comments »

Let’s face the truth; Alexa is not the best source of traffic data out there on the Internet. Plus, it does not have statistics on conversion rates. But, hey, Alexa is free and we are going to use it to benchmark (approximately) the conversion rates for your competition. Here is how to do it two simple steps:

Step 1. Establish industry norms using your actual conversion rate data

Suppose you are SEOMoz (I am using this website as an example and have no real data for them) and you sell paid tools for SEO. Let’s suppose your current conversion rate is a conservative 4% (again, hypothetical data) and you want to estimate how your competitor SEO Book is doing.

You and your competitors (since it is the same industry after all) follow a similar trend when it comes to relationship between conversion rates and other site metrics such as bounce rate, time on site, and page views per user. In this step, we try to calculate values for parameters which relate conversion rate to all these metrics. Using your actual conversion rate data and the stats that Alexa shows about your website, calculate X, Y and Z as the following:

• Your Conversion Rate = X * (1 / Bounce Rate as shown by Alexa )
• Your Conversion Rate = Y * Time Spent on Site as shown by Alexa (in seconds)
• Your Conversion Rate = Z * Page views / user as shown by Alexa

The reason we don’t use your actual bounce rate, time spent and page views data is because you don’t have that data for your competitors. You only know what Alexa says about your website and what Alexa says about your competitors. So it is better to work on the Alexa data that is freely available and uses the same methodology all across.

As an example of SEOMoz, Alexa tells the bounce rate, time spent on site and page views / user is 50.7%, 219.7 seconds and 3.2 respectively. Using this data, we get the values of X, Y and Z as:

• X = 4% / (1 / 50.7 %) = 201.6
• Y = 4% / 219.7 = 0.018
• Z = 4% / 3.2 = 1.25

Step 2.  Use the parameters to estimate competitor’s conversion rate

Now we have obtained the parameters which relate your actual conversion rate to the data that Alexa shows about your website. Next step is to simply use those parameters on your competitor’s data (as shown by Alexa) to get estimates of their conversion rate.

Competitor Conversion Rate:

• Estimate 1 = X * (1 / Competitor Bounce Rate as shown by Alexa)
• Estimate 2 = Y * Competitor Time spent on Site as shown by Alexa
• Estimate 3 = Z * Competitor Page views / user as shown by Alexa

Finally, to get an idea of what their real conversion rate, we simply average the estimates.

Competitor Conversion Rate = (Estimate 1 + Estimate 2 + Estimate 3) / 3

Continuing the SEOMoz example, if we were to estimate the conversion rates for SEOBook, we calculate 3 estimates of conversion rate (based on the data shown by Alexa for SEOBook):

• Estimate 1 = 201.6 * (1/69.4) = 2.90%
• Estimate 2 = 0.018 * 135 = 2.43%
• Estimate 3 = 1.25 * 2.43 = 3.0%

As you can observe, the estimates are quite close. Hence, we can be pretty confident that the actual conversion rate is close to the average of these three estimates, which is:

Estimated conversion rate = 2.77%

Of course, the above estimated value is only valid if the real conversion rate for SEOMoz that I assumed (4%) is true, which may or may not be the case as I don’t have access to their real web analytic data.

Summary

Simply plug in your conversion rates in the above methodology and you should have pretty good estimate on how you are doing as compared to your competitors. You can also try to triangulate your estimates by using other data sources (apart from Alexa) such as Compete.com or Quantcast.

Do let me know if you find this approach helpful.  As always, feedback and comments appreciated. Want a tool to automate all this analysis for you?

PS: The way I define Bounce Rate and Conversion Rate, they are not related.  But the way Alexa defines, the two metrics are definitely related.

Posted in Conversion Optimization on November 24th, 2009 | No Comments »

Lately, I have got quite a few requests for how effective call to action buttons are created in Photoshop. Though having persuasive text as call to action is important, button shape, size, color and style can also make a tremendous difference in conversion rates. So, here goes the list of free resources on creating buttons that convert and examples to get you started:

Photoshop Resources

• Make A Big, Beautiful Call-To-Action Button In Photoshop [Sitepoint.com]
• Create Clean and Fresh Call-to-Action Button + Embossing Text Effect in Photoshop [PSDVault.com]
• How to Design Call-to-Action Buttons with Photoshop [Woorkup.com]
• How to Create a Slick and Clean Button in Photoshop [SixRevisions.com]
• 30 Useful Photoshop Buttons and Badges Tutorials [SpeckyBoy.com]

Ideas for Buttons

• Call to Action Buttons: Examples and Best Practices [SmashingMagazine.com]
• Good Call to Action Buttons [UXBooth.com]
• 10 Techniques For An Effective ‘Call To Action’ [BoagWorld.com]
• 5 Tips For Creating An Effective Call To Action Button [SitePoint.com]
• 57 Strong Call to Action Buttons [xdxy.com]

This list is an ever expanding list, so feel free to suggest more (free) resources for call to action buttons. Leave a comment and I will add it in the list.

Posted in Conversion Optimization on November 18th, 2009 | No Comments »

Lot of people inquire about what an ideal bounce rate or conversion rate is and if their website metrics are in the right range. One size doesn’t fit all. In this post, I fish out industry standard conversion rates and bounce rates. Though your only competition should be you, having an idea of  industry metrics might help some.

Sources for these figures:

• Fireclick Index
• Anil Batra’s survey of Typical Bounce Rates [PDF]
• Coremetrics Monthly Industry Reports [PDF]

What is your bounce rate or conversion rate? Does it match with your industry average?

Posted in Conversion Optimization on November 17th, 2009 | No Comments »

To set the definitions right, it is generally agreed that bounce rate is the percentage of visitors who exit the website immediately after arrival. Conversion rate is the percentage of visitors who complete website goal, which may be a signup, subscription, purchase, download, etc.

Most people believe that bounce rate and conversion rate is inversely proportional. That is, if bounce rate goes up, conversion rate would go down and if bounce rate goes down, your conversion rate will go up (because apparently you will have more interested visitors). On the face of it, this seems to be true and hence the proposition that fixing the bounce rate OR the conversion rate alone will achieve business goals seems to be true.

Sadly, this relationship between bounce rate and conversion rate is an illusion. To understand that there is NO relationship between these two metrics, you need to know what bounce rate really is. Does the bounce rate talk about visitors who viewed just one page on your website? Or should it capture more nuanced idea of visitors who stumbled across your website by chance? Most web analytics tools define bounce rate as the former: that is, a single visit is considered a bounce. Bounce rate, defined in such a manner, conveys completely wrong information.

Increasingly, visitors are becoming goal oriented.  For example, if they need to see your shipping policy, they will Google it, read about it and leave your website.  That visit is not a bounce: visitor got what he was looking for. Similarly, most of you will exit after reading this post for say 3-5 minutes. Do I consider you a bounced visitor? No, not at all.  You were engaged for a long time, how could you be classified as a bounced visitor. However, the web analytic tool I use will classify you as a bounce because you just read one page on the website. Realize that bounce rate which you are reading out from your tool is not what it says. Scrutinize definitions and understand what the metric is saying to put it in the right context.

So, what is the best way to represent bounce rate? I think bounce rate is best captured by measuring what percentage of visitors spent less than 30 seconds on your website. Any time  spent which is less will signal that visitors arrived on your website by chance and is NOT at all interested in what you are offering, hence quickly went back to what he was doing. All other visitors spending >30 seconds, even if they just see one page, should be classified as non-bounced visitors. To summarize:

Bounce rate = Number of visitors who spent < 30 seconds on the website /
Total number of visitors

Unfortunately, measuring exact time spent by a visitor by web analytics tools is difficult and most of them will approximate this number. That said, I think bounce rate should be defined by time spent on website and not by pageviews.

Coming back to conversion rate, how is it related to bounce rate? As traditional thinking goes, the visitors who bounced bring the conversion rate down as they have no chance of completing the website goal. I fully agree that bounced visitors (by definition) have no chance of completing the conversion goal. Then, I ask, why to include bounced visitors in conversion calculations at all? To truly reflect the progress you have been making on your website, conversion rate calculations should NOT include bounced visitors. Bounced visitors never really cared about your website, non-bounced are the ones who engaged and spent time going through what you are offering. Conversion rate should capture how good a job your website is doing for getting those visitors (who care about your website) to complete the goals. Conversion rate, ideally, should be calculated as following:

Conversion rate = Number of non-bounced visitors who completed the goal  /
Total number of non-bounced visitors

So, now we have two metrics which are not at all related to each other: bounce rate and conversion rate. Both of these metrics convey different information regarding how you are performing. Hence, both of these metrics should be separately optimized. Optimizing bounce rate is for convincing more number of people to engage with your website. Optimizing conversion rate is for convincing the visitors who are already engaged to complete your website goals. Reducing bounce rate AND increasing conversion rate are two different activities.  Remember that.

What are your views on relation between conversion rate and bounce rate? How do you and your web analytics tool measures bounce rate?

Posted in Conversion Optimization on November 12th, 2009 | 6 Comments »

• F-Shaped Pattern For Reading Web Content – the most well known article that laid out the generally proved F pattern of attention on a webpage. The pattern suggests that  headline and starting paragraph are THE most important elements on your page.
• Eye gaze Patterns while Searching vs. Browsing a Web site – confirms the F style eye gazing pattern applied both to browsing and searching modes.
• An Eye Tracking Study of How Font Size, Font Type, and Pictures Influence Online Reading [PDF] – A research paper by IBM and Google. Font size and font-type do not really make much difference in browsing style.
• Effects of Hyperlink Density on News Web Page Reading: An Eyetracking Study – A research which concluded:
  • increased number of links on the page may lead to increase in number of pages viewed on the site but would not necessarily affect conversion rate.
  • If there are more links on the page, people may not see links below the fold at all.
• The Determinants of Web Page Viewing Behavior: An Eye-Tracking Study [PDF] –
  • More complex your page is, more the possibility that the visitor will take an unexpected eye gazing path on your page (which may reduce the chances of missing your call to action).
  • The first page the visitor lands on provides a lot of context to his total browsing experience on the site. What he may miss on the first page, may be not be noticeable during his total further interaction on the website.
• Could I have the Menu Please? An Eye Tracking Study of Design Conventions [PDF] –
  • This study indicates that the users rapidly adapt to unconventional user interface, however that “unconventional”-ness should be maintained across the website
  • Participants take 50% longer to complete the search task on the complex site as compared to a simple site

Posted in Conversion Optimization on November 9th, 2009 | No Comments »

There are NOT a lot of free resources available on the Internet for A/B Testing. This post tries to lists the best tools, guides and resources for A/B Testing. As it will be an ever growing list, feel free to make suggestions for additions into the list.

Tools

• Visual Website Optimizer – Our own hassle-free visual A/B, split and multivariate testing tool. Use invite code “wingify-blog” (without quotes) for a free account.
• Google Web Optimizer – From the land of Google, comes a simple but free to use A/B and Multivariate testing tool
• Five Second Test – Ultra simple A/B testing for images
• A/Bingo – A Ruby on Rails plugin for doing A/B tests
• Vanity – Experiment driven development framework for Rails
• Genetify – An open source JavaScript based framework for A/B testing
• DIY A/B Split Testing using PHP and MySQL – a very simple and short tutorial (with code)
• A/B Testing with CakePHP and JQuery – a plugin for CakePHP framework
• Django-lean – a split testing framework for Django

Guides

• Effective A/B Testing – a superbly made presentation cum tutorial on A/B Testing
• 101 Tips for Google Website Optimizer – a long list of tips, tricks and ideas on how to use GWO
• Microsoft’s Practical Guide to Controlled Experiments on the Web [PDF] – an excellent paper on how testing should be done
• Technical overview of How Google Website Optimizer Works
• From 20 Bit’s blog
  • An Introduction to A/B Testing
  • Implementing A/B Testing
  • Statistics for A/B Testing

Show case of existing A/B Tests

• Anne Holland’s Which Test Won
• ABTests.com

If you have any other suggestions for additions in the list, I will be happy to add them. Just leave a comment.

Posted in Conversion Optimization, Split and Multivariate Testing on November 5th, 2009 | 17 Comments »