July 2006


31 Jul 2006 12:07 am

perspective thumbIt is only fair to follow up a post titled "Stop obsessing about conversion rate"  with this post.  (Just in case you have not read the Stop Obsessing post that please read that first for more context).

Conversion rate is a very important metric, used properly. Here is my point of view on the basics and best practices for measuring conversion rate.

Definition first: Conversion rate, in percentage, equals Outcomes divided by Unique Visitors during a particular time period.

What are Outcomes: From a macro perspective any reason for which your website exists. Most frequently this is the total number of orders submitted or total number of leads collected or total number of newsletter/email sign-ups. For non-ecommerce websites it can be number of people who completed a task, so for a support site the number of people who got to a faq or a answer or a knowledge base article (this is a really crude measure of success for a support site but it is something).

Why use Unique Visitors: There is a lot of heat around this topic. Some people are in the Total Visitors camp and others are in the Unique Visitors camp. Remember trends are important so if you stick to whatever you prefer just be consistent over time and you’ll be fine (don’t feel the need to get “un-brainwashed” from whatever camp you are currently in : )).

My personal view on why we should use Unique is that every session is not an opportunity to get a customer to hit submit order (I do realize this sounds scandalous, more so because like a majority of you I live in the world of ecommerce : ).

Shopping / web surfing is a delicate dance of come to see who we (the company) are, go to another site read reviews, then come to see our benefits, then to ask you wife it is ok for you to buy and then come purchase. I am probably missing a few other steps in there.

Using Unique Visitors is a better read of what is really happening on your website because it accommodates this dance and gives you "credit" for those prior sessions when the dance was on.  More importantly being a practitioner I feel metric definitions should incorporate on the ground reality and using Unique Visitors accommodates that reality. (Matt Belkin, you have read his interview on this blog, has a great alternative point of view on this, click here for that.)

Uniqueness is currently measured by setting a persistent cookie (call it shopper_id) most of the time and they are a bit unreliable (I have to stress that certainly not as much as the hoopla that surrounds them) and hence this is not optimal. But still the best we got. If you disagree please recall the statement above about consistency and trends.

Time period: If you are measuring Weekly Conversion Rate it is the sum of orders during that time and sum of unique shopper_id's during the same weekly time period. It is not the sum of unique visitors on each day and then a sum of that daily unique visitor number for a week. For monthly sum of orders during that month and  sum of unique shopper_id's during that time. It is not recommended to sum daily unique visitors to get a total for month or week.

All of the above might seem to be too much detail just to get started on the metric but all this detail reflects the importance placed on this metric and how most of the time we don’t even agree on what the definition is and how it should be measured. So hopefully it was helpful to you.

Ok so here seven simple best practices related to conversion rate:

# 7 : Start with overall site conversion rate. (Then promptly forget about it.) There is no other metric that will tell you less about your website than overall conversion rate (assuming you have more than $ 25,000 in sales on your website) but overall conversion rate is easy to measure so you should do it, I recommend using the definition above, and get it out of your system.

# 6 : Trend over time and don’t forget seasonality. If you have read this blog for any amount of time you know I love two things: trends and segmentation (more on this later). Most definitely trend conversion rate numbers but what is unique about this metric is that more than others it is really impacted by seasonality and so do things like 13 month trends or look at 5 quarters or 8 days.

In each case you will have same period from history to compare with. That will give you a lot more context and avoid sub optimal reactions / actions by comparing time periods that have nothing do to with each other, say last month vs this (unless your “season” is that, which it could be).

# 5 : Understand exactly what the acquisition strategy of your website / company is. This is not a report, it is a conversation / investigation with your business partners and it is an extremely important step that any analyst needs complete. Figure out what is your core acquisition strategy and then measure conversion rate for those elements. Is your company heavily into Direct Marketing (email, snail mail etc)? Are you spending excessively on PPC (Pay Per Click)? Or maybe you are about to plunk down a million dollars on a new affiliate marketing strategy or maybe on SEO.

Do this before you get too deep into conversion rate measurement because it will without a doubt lead to more meaningful analysis from you and, hyper importantly, you will measure what is important to your company rather than what your friendly neighborhood web analytics tool is throwing up at you.

# 4 : Conversion rate by top five referring url's. This sounds really simple and silly but there is usually a disconnect between what the company strategy is and where the traffic really comes from.

So for example if you are a complex business with many websites you might not necessarily be measuring conversion rate for your corporate site which might be sending you tons of traffic. Or from some blog that started to praise you a lot. You get the idea. I always find “hidden gems” in the referring urls and measuring conversion for your top five referring url's is a great insights finding insurance policy.

If you find that this report shows you the same stuff as some of the above or below you can stop it. But if you are not doing this I bet you are missing something delightful.

# 3 : Don’t measure conversion rate by page or link. This is a request we all get and often execute on. It is rather sub optimal. In a complex multi page and multi link web environment how could you possibly measure “conversion rate by page”? Unless your web scenario is: Two people come to the site, one enters at the home page and the other at product overview page and you can get to checkout directly from both pages measuring conversion rate “of” each page is very misleading.

In the click density (site overlay) report some web analytics tools show conversion rate for each link on the page. The hypothesis is that x % of people who clicked on this link purchased. Unless all these links lead to the checkout this is a useless piece of information.

In a multi page complex path web experience simply the fact that someone saw a page or clicked on a link is not enough to attribute any credit that page / link in terms of conversion rate. (If you want to measure value of a page see the approach described in this post and look for the section where we talk about “page influence”.)

# 2 : Segment like crazy. Most websites convert in single digits (usually low single digits). With such a small ratio of people converting the “insights gold” is hidden deep in your site data and it is extremely critical that you segment “like crazy” in order to find your insights. This means showing top 5 (or x where you pick the x right for your company) segments of conversion rate, we have discussed two above already, referring urls and core acquisition strategies (DM, PPC, SEO, Affiliates etc).

That is just a start. You should have indented sub segments for each of the top five. Something like conversion rate of top five sub segments for each segment of conversion report so you can really show where the desired outcomes are coming from. So top five direct marketing campaigns, top five PPC key phrases driving conversion or top five affiliates etc. This is not too complex for any top website all this information will fit on one page (and with 10 size font : ).

Segmentation is not hard, for a simple view and some ideas see the post Segment Absolutely Everything.

# 1 : Always show revenue next to conversion rate. Another extremely simple trick to provide context. We usually create a report that will show various conversion rate buckets (like the idea in #6 above). But conversion rate just by itself can be misleading in terms of opportunity for any website. So my recommendation is to show the actual revenue number (or leads or newsletter sign ups or whatever is your conversion) next to the conversion rate %.

What you will find is that some of your highest conversion rates don’t bring in most of the revenue, it could the line in your report with the fifth highest conversion rate (or whatever).

Typically people see a high conversion rate and think “let’s do more of that”. But the highest conversion rate could come when you give deep discounts (which you can’t do all the time) or come from segments of customers you can’t find more of (say existing customers). Providing outcome numbers next to the % gives more meaningful data to your decision makers. Sounds simple but is very effective.

One last bonus recommendation for our dear blog readers……

# 0 : Never measure conversion rate without a goal. This is not always possible but a highly recommended best practice. Having a goal gives context to your actual number, asking your business decision makers forces them to think about where the revenue (or other outcomes) will come from causing them to really analyze their execution strategies and try to plan them ahead of time as much as possible.

Having a goal guides the conversation and analysis and the deep dives that will yield insights and asking for a goal makes you integrated with the business decision making process (because almost always the user of the web analytics tools is not the one that will set goals).

Like best practice #5 above this recommendation is not so much just to have on the report, which is important, but to insert a social / cultural change in your business. In Web Analytics goals are hard to find and yet there is enormous pressure from your business leads for insights and action recommendation. Ask for a goal, force them to think and create a environment where you, dear analysts, lay some responsibility for business planning and execution on their door. In the end you will look like a hero if you help push this change in the typical company culture.

I do realize the irony that all of the recommendations above might, just might, end up getting you to obsess about conversion rate more than you should. : ) Do please avoid that tendency and see the post on why you should not obsess about conversion rate. : )

Agree with the best practices recommended above? Are there tips and tricks that have worked for you better? Please share your thoughts, critique, feedback, random musings with us via comments. : )

[Like this post? For more posts like this please click here.]

27 Jul 2006 12:20 am

beeOn July 24th I had the privilege of giving a presentation at the Google Mountain View headquarters as a part of Google’s TechTalk series (thanks Brett). The topic was Customer Centric Web Decision Making.

The talk covered some of the topics that I have already blogged about here but there is a lot of new material I have not had a chance to blog about.

I really enjoyed the session and thought that I’ll share the video with you all. The video is 46 mins long and there are some great questions from the Google audience at the end.

I have to forewarn you that the quality of the video is good but not exceptional. Most of the time you’ll be able to make out the slides great but the presenter, yours truly, will hover around like a shadow (!). The audio is very good, maybe I should have converted this into a podcast. : ) Here is the talk…….


The video is located at google at: http://tinyurl.com/qhdk9

If for any reason you want to download the video (and you are a patient person with a broadband connection) you can use this link: http://tinyurl.com/n72wl

I have a 5g ipod so usually I download these videos into my ipod, for that you can use this link: http://tinyurl.com/rhbxu

If you watch the video I would love to get your feedback / critique. Anything is fair game: style, slides, content, message, presenter, ok even the lighting. : )

I hope you’ll find some value in the video, thanks in advance for your comments and feedback.

Next Page »