Coffee, Sun & Technology

3rd post on the session length report (see part 1 & part 2 ) - there will be a 4th and last one.

Last time, we were analyzing the influence of the landing page for a campaign on the session length report. Here is the distribution we considered. It’s pretty easy to see that when the landing page is a custom page, the initial exit rate is extremely high and the people are churning out fast, after 1 or 2 clicks, maybe 3. Yet, in this example the custom landing page drove higher conversions than its home page rival. Why? Here is an hypothesis: highly targeted landing pages tend to polarize the audience into people very interested and people absolutely not interested. The ones not interested leave instantly. The ones interest have a very high likelihood of ending up buying the product, which explains the higher conversion rate.

Bringing conversion rates into the picture provides further insight. Take a look at this dual distribution: left axis and blue line represent the session length distribution, right axis and green bars represent the conversion rate for a given session length - both charts are for users entering the site via the home page. As expected, since on this site you cannot buy a product in less than 6 clicks, the conversion rate is 0% of any session length <6 page views. Then the conversion rate slowly increases slowly as the session length goes up.

This leads to a key hypothesis: all other things being equal, if you can somehow reduce the number of pages it takes for visitors to get to the product they’re interested in, then you will dramatically increase your conversion rates.

Let’s run some numbers in this example - assuming that we managed to cut the session length by one page (with better site navigation, guided navigation, internal search, etc.) - without impacting the overall user experience on the site. As far as the resulting session length and conversion rate distribution are concerned, you would effectively be shifting the conversion rate distribution by one notch to the left - see here. Using Excel, you can easily compute the incremental orders/revenue which in this case gravitate around 22%. Not too shabby!

This 22%, which again is the incremental orders you would theoretically get by cutting your session length by one page view, can be called the Session Length-Conversion Multiplier. What’s yours?

In case you haven’t heard yet, Google announced today its acquisition of Urchin Software (see story). Eric Peterson from Jupiter wrote a great but scary post which in essence questions the survival of smaller players in the Web analytics industry, now that Google is in. Google is so rich, so talented, so bold that is it almost impossible to predict with any degree of certainty what they will do with Urchin. Remember, these are the folks who introduced the 1GB free email account last year, when everyone else was offering 25MB.

But here again, I don’t think Google is in a mission to save the Web analytics space (nor Urchin). Their mission is to make money: drive more clicks on their banners and drive higher CPCs. By providing adverstisers free tools to run more effective campaigns, Google will naturally (a) further distance Overture and MSN in terms of performance, (b) expect higher CPCs since the site is converting better and customers are therefore willing to spend more.

I might be wrong, but it is my belief that Google has no interest in becoming a fully featured Web analytics vendor. Sure, they will maintain the core Urchin feature set - but will likely repurpose the tool to help their customers make more money with AdWords and… spend more on AdWords.

Another “To be continued“.

Francisco partners, a private equity fund, announced today they acquired WebTrends (see story) for $94M in cash. WebTrends is back to being a privately help company, after going IPO in the late 90’s and selling to NetIQ in 2001 for $1B. Let’s not kid ourselves. This move is not meant to “rebuild WebTrends”, nor to create a “Web analytics leader”. It’s primarly about money and return on investment. Private Equity Funds aren’t philanthropists - they probably already have their little exit strategy in a nice deck of PPT slides.

So what are they thinking? With revenues in the $50M/yr ballpark, and a public competitor trading at 8x TTM revenues, a $300-400M valuation seems very doable. You pay $94M, you get $300M. Not bad. Question is how. A couple of strategies come to mind:

1 - Acquisition strategy: Reaching the $80-100M revenue range by bulking up the offering (email, SEM, SEO, etc). Then go public as something “bigger” than a Web analytics pure play.

2 - Focused strategy: the opposite of (1) - streamline the business, focus on profitability. Don’t allow any more distractions. Maybe merge with another Web analytics vendor and become the 800-pound gorilla in the space. Then sell the business at the right time for the right money, or go public.

Or, 3 - Something in between: streamline now, grow, use the profits to make a few key acquisitions, then go public or sell. This will take a longer time and I doubt their investors will go for that.

Regardless, let’s temper any enthousiasm - today Acxiom bought Digital Impact for $140M, at a nice 50% premium over the stock price. This in contrast of the Webtrends acquisition, which many consider to be at a disapointing valuation. Everyone in the Web analytics space would have preferred a higher valuation.

Clearly, To be continued.

See the Motley Fool’s quick write-up. Interesting comment in the brief:

“But trouble begets opportunity, and two companies now appear to be trying to take advantage of the climate of fear prevailing among investors in the data industry.”

“ChoicePoint…” - “Spyware…” - “Cookies…”. Climate of fear - or climate of ignorance?

Let’s take this report to the next level and slice it a bit. (BTW, Session length part 1 can be found here).

1 - First time vs Returning users
Take a look at this example. Interesting but you can’t be surprised. The 1-page exit rate plummets from 35% to 15% for returning users. Also note that the 3-page “bump” is more pronounced for returning users - as if they came just to check if their favorite site had the product they’re looking for - and then left.

2 - Slicing by visit number
We’re now looking at the influence of a user’s visit number over session length. Take a look at this distribution . This time, the chart is not a distribution but a cumulative chart indicating how many users leave after X page views, as a percentage. And we’re doing this for users on their 2nd visit to the site, 3rd visit, 4th and 5th. Again, interesting to see that the more visits to the site, the longer the session. People seem to have more patience for a site they know.

3 - Slicing by Geography
Let’s try to see if geography has any influence. See here the distribution for San Francisco vs New York. Again, interesting and not terribly surprising to see that browsing patterns aren’t that different based on where you live (*in general*, there are expections).

4 - Slicing by Entry page
In this final (and most interesting example!), we’re looking at the influence of the landing page on session length. In blue, campaigns with the home page as the landing page, in red, campaigns with a custom page for the landing page. Take a look and email me your thoughts (or drop a comment)…

Until next time - for part 3.

Session length = number of pages users viewed during their session on the site. With a little training, dissecting this simple report can provide invaluable insight to assess what a site’s strengths and weaknesses are.

Take a look at this distribution from a leading retail site (sanitized and published with permission). It is for all the users of the site. This is how to read the chart: it gives you the percentage of users who read X pages in their session. In this example, 25% of all users only viewed one page in their session.

Couple of comments about this:

1 - 25% of users see one page, then go. This is a fairly sticky site, 25% isn’t bad compared to others.

2 - The longer the session length, the fewer the people, EXCEPT a “bump” at 3 page views.
a) This is why (in general) it is important to help visitors find the right product as soon as possible, before they churn away
b) The bump a 3 page views is typical for sites with a very large selection of products, where Internal Search is extensively used. People check to see if the site has their product, they try the first search result, and a large fraction of them just leave the site at that point.
c) For sites with a small number of products and no Internal Search typically you don’t see any of these bumps. Here’s an example.

All right - what does this all mean?

1 - If your 1-page sessions are too high then you need to fix your top entry pages. Track your exit rate for your top 10 entry pages on a daily basis.

2 - For many sites I have worked with, a simple rule of thumb applies: you approximately lose 1/2 of your visitors between click 1 and 2, then 1/3 between click 2 and 3, then 1/4 between click 3 and 4, etc. Therefore, it should be your goal to help your visitors get to their product as soon as possible: work on navigation, Internal Search, have better landing pages which are relevant to the user, etc.

These are well known site design principles but they are way too often forgotten or simply ignored. More on this tomorrow or day after…

Let me start by saying that I am a true advocate of A|B testing. eCommerce sites should A|B test *all the time*, in fact it makes sense for larger online businesses to have “A|B testing Managers” who would plan, execute and analyze A|B experiments on an ongoing basis.

3 tips however,

1 - Take your time. Don’t rush to fast conclusions, allow at least 7 days - even if your traffic is very high. Traffic patterns go through weekly (even monthly) cycles and the types of visitors you get on a Sunday are different from the ones you get on a Tuesday.

2 - Analyze results across your key segments. Examples of such segments are: Returning buyers, returning users, first time users, etc. This will help you better understand who your customers are and what they are looking for. Often, you may alter your test and converge to a better final result.

3 - Focus on high traffic pages. Home page, product templates, general landing pages for your marketing campaigns. These tests take time to execute, so it’s important to work on what can make a difference.

Greg Edwards from Eyetools Research recently published 2 fascinating stories on the Washington Post and the Google search results page. Another study from the same firm focuses on eTrade and demonstrates that many areas of their homepage aren’t getting read by visitors.

I think these studies are insightful and worth the money (pricing starts at $1,500). Nay-sayers will argue that views don’t mean clicks. But users will never click on links they can’t see. And that is exactly the point: what do users see?

Not so long ago, marketers were buying 200 or 300 keywords on Google. Today, 2000 or 3000 is the norm - and it’s not uncommon to see sites bidding on 20,000 - even 50,000 keywords per engine. Marketers are getting their own “scalability challenge” aren’t they? “How many campaigns can I run concurrently”, “How profitable can I make them”.

Luckily, some great tools exist on the market today to manage keywords on an individual basis, and in near real-time. AtlasOnePoint, formerly GoToast, is one example - dozens of other solutions are available at a moderate cost in what can almost be considered a commodity market. Marketers are in charge of the creative and the landing page - bid automation tools manage the spending in each of the search engines.

Given the current market conditions, a few questions come to mind. First, the rules for doing business online are changing - still - as this process (theoretically) makes business profitability more important than ever. If all bidders compete with smart tools which optimize their spending, then virtually everyone gets his fair share of the pie, no one pays more that he should at a given point in time for a visitor to the site. There is no “cheating”, no possible “great deal”, all businesses are equal under these rules. On the flip side however, business profitability becomes incredibly important. If Business A makes average profit of $2.00 per visitor, and Business B only makes $1.00, then A can bid more and get dramatically more traffic than B.

Second, marketers have to change their modus-operandi. This model pressures them to take a short term view when assessing the quality of their search programs. With so many variables changing every day, so many programs, it is impossible to compute the long term returns of each individual campaigns and use this information to intelligently adjust the spending. Marketers are forced to consider immediate returns, and completely ignore the longer term (6 month returns is considered long term nowadays).

Third - the complexity of doing online marketing is dangerously increasing. MSN is launching adCenter (read this excellent article from Kevin Lee), Yahoo and Google are promising new enhancements, the smaller fish will shortly follow, etc… All vendors are trying to increase inventories as much as possible while prices keep climbing.

If prices keep going up, if the rules keep changing, if the inventory keeps increasing… isn’t this starting to look like a… bubble? Frankly who can say - but it’s a very real possibility.

SCIENCE FICTION - Palo Alto - December 12th 2019. Motorola announced today the first Track-Me-Better(R) enabled cell phone. This new phone allows users to be tracked in all brick and mortar stores equiped with Track-Me-Better technology. The benefits of this technology, which was introduced long ago in the form of a micro-card, are well know to the general public - it allows better targeting of promotional offers, more personalized customer service in a store and automatically accruing free miles at their favorite airline when customers enter a participating store. With this move Motorola continues to build its swiss army knife phone platform - and strengthens its position in the marketing services space.

Fireclick’s CEO Etienne Casanova said: “Track-Me-Better(R) coupled with powerful analytics and actionnable marketing tools such as Visual email, 3D-interconnect, and GoogleUltra provides an unprecedented opportunity for marketers eager to connect with their offline visitors.”

In a separate release the Marketing Research firm Fremont Partners estimated to total number to Track-Me-Better users to reach 83 million by the end of 2020 in the United States.