The Commerce360 Blog

What should your conversion rate measure? Is the typical idea of how many of your visitors buy too broad? Aren’t some of those folks hopelessly improbable prospects? Aren’t there other successful outcomes besides a purchase?

These are the question being discussed in a recent blog post from our friend Avinash and many other respected analysts who have left thoughts in his comments.

Avinash suggests that visitors who bounce (stay less than 10 seconds or 1 page), bots (visitors that aren’t people), and those who display some intent other than purchase, be removed from the equation – with the result of a higher conversion rate to more accurately reflect how many ‘viable’ visitors were converted.

Like others who have commented, I think that the gross calculation (outcomes/uniques) has merit both as a baseline view of the site’s performance and as a reference for comparisons between sites within or across industries. As has been suggested (including by Avinash himself) just about all of our analytical results are directional or relative and not perfectly accurate anyway – so getting hung up on the relatively low absolute value of the numbers is in many ways not worth much effort – but it does give us a reference for change and comparison.

If everyone starts subtracting swaths of visitors before they calculate the conversion rates they report it will be impossible to understand or learn anything comparatively. Numbers like those we get from shop.org, disappointing as they may be, are useful so long as we can assume we’re all measuring the same thing.

On the other hand, internally and incrementally, there is an absolute need to dig deeper and measure conversion both more narrowly and against a broader range of targets.

I also agree with John Quarto-vonTivadar that tossing bounces wholesale isn't the right solution - those occur in quantities way beyond indiscriminant clicks or an inability of the page to impact the visitor. Before you discarded them you’d have to consider the sources of these ‘bouncers’ – those coming from paid search, email replies, banner ads, and even organic search clearly had some expectation and goal when they clicked. Could a few percentages of them suffer from muscular tics and poor hand-eye coordination? Sure. But more than that and I’d really want to think about why the site/page was so non-compelling as to result in their near instant departure.

Given that bounce rates are commonly quite high (Avinash suggests 30% as typical and we’ve certainly seen them much higher) and it would make sense to segregate that problem – may I suggest a ‘Lingering Conversion Rate’ which applies only to visitors who visit at least two pages or stay at least 20 seconds?

I really like the ‘conversion pie’ Avinash created, as a way of visually expressing the fact that many visitors come with intents (or potentials) which exclude purchase. It would be great to take it further and have the ‘non convertible’ segment replaced with success and failure slices for the other intents which users manifest via their behavior. Determining and categorizing these intents and their achievement is of course quite difficult today – but I think this is an area where we need to do some work and we need our vendors to help.

For example, suppose we know that a good number of our visitors come to do pre-purchase research. Wouldn’t it be nice if we could flag users based on terms they use in their searches (‘compare’ ‘reviews’ etc.) or some combination of links clicked or pages viewed and then get associate success events with each of these segments and get both detailed and comparative reporting? I’ve mocked up the kind of report I’d love to see. (click to enlarge)

Thanks and congratulations to Avinash (again) for the original insights.

Posted by Craig Danuloff at 1:06 PM | Permalink | Comments (2)

Opinions and Reviews are in many ways the basis of Marketing2.0. Consumers now have a voice, Brands are not defined by the company, Etc. Etc. Etc. The reviews on Amazon.com and Epinions were the first wave, followed by affinity forums where the merits of just about anything is endlessly debated (my favorite example continues to be The Remote Control Forums) and of course the opinions-are-oxygen world of weblogs. New technology and services to foster reviews are emerging with BazaarVoice and others.

Fake reviews, or those posted by shills, have long been a problem in the world of user-generated content. From authors' friends writing Amazon reviews to completely fake blogs discussing the merits of products and companies, the utopian dream of helpful and unbiased information had long required a healthy dose of cynicism and fact-checking.

So how do reviews posted on blogs for which the writer is paid, but discloses this fact, change the equation? ReviewMe.com has opened a marketplace for reviews. Anyone with a product or service that wishes to 'buy' reviews from bloggers, simply says how much they're willing to pay. Bloggers who want to receive these offers register their website and are then ranked and rated to have their 'value' determined. ReviewMe then plays matchmaker and the opinions and the cash flow.

The reality, of course, is that most 'review buyers' will be more interested in the links that come from the posts than the reviews themselves. Buying links in blogs for one-time fees isn't a bad deal if the ReviewMe pricing system essentially values PageRank and Authority, which it does. The impact on word-of-mouth and actual click-through traffic will be considered a bonus and almost certainly be positive on balance, assuming that both most reviewers aren't too harsh (a truism) and particularly lousy products and services probably won't participate.

Over time of course, the Google-Juice from these posts could diminish. Given the disclosure requirement it won't be hard to for the engines to algorithmically discard these posts from having PageRank influence.

One of the paid reviews of ReviewMe itself calls it a "marketplace for buying influence." I think that's a pretty good phrase and concept. Why shouldn't someone who ads value to the sales or marketing of a product be compensated? (Isn't that the premise of affiliate marketing?) Adding reviews of the reviewers (Was this review helpful?) could further benefit both the readers and the review-buyers in sorting out the legitimate and valuable from those just BS'ing to make a buck.

Beyond the moral issues and logistics, there is the very real problem that most people probably aren't qualified to blog-review very many things, and won't be paid that much for each post - so will probably stop participating after the novelty wears off. Alternatively, if you blog-reviewed too heavily the opinions or quality would likely be suspect. I'm sure a few particularly opinionated writers will craft new paid-review blogs on which they dish out interesting and informative opinions and make a nice side-income, but they'll be the rarity. In the end I don't think the ReviewMe idea scales very far, but is another piece of the evolving landscape.

NOTE: This was NOT a paid review. I do however own the domain name influencepeddler.com (and have for years) so maybe I've gained some value there. Here are links to other interesting comments on ReviewMe, some of which were bought.

Posted by Craig Danuloff at 12:27 AM | Permalink

The back cover of the new Dell Business Catalog has one of those amazingly cheap desktop systems - $399 complete with 19" LCD monitor. I need a very basic PC to run the Omniture Desktop Player 24-hours a day - this could be the answer.

Should be easy to buy, right?

1. Go to Dell.com
2. Type the E-Value Code listed in the ad into the search box. Not Found.

3. Maybe I shouldn't actually type 'E-value'. Doesn't Help.

4. Enter to custom URL listed on the same page: Doesn't Exist.
http://www.dell.com/shopdellbiz/November

5. Choose Desktops / Small Business / Dimension / 5150 / Customize (5 clicks)
6. Find no sign of bundle from the ad.
7. Notice 'Purchase Help' tab on top of page (wasn't on home page)
8. Buried in left nav bar, click 'Shop Promotions'
9. Finally, EValues code icon. Can't enter it yet. So I 'click here'.

10. Enter the code and find the system as advertised.

At this point I was able to politely refuse 475 offers to modify the system and put it in my cart at the advertised price.

Think the telephone number on the back of the catalog was any better? I tried it. 5 levels into the phone system I was told that I reached them after hours (10pm Friday night).

Posted by Craig Danuloff at 10:00 PM | Permalink

Word of Mouth Marketing by Andy Sernovitz is a surprisingly good albeit imperfect book. I give it very high marks for clearly defining the often hard-to-pin-down concept of ‘word of mouth’ and for elaborating on the definition with very clear rules for it, reasons to do it, components of it, and a built-in implementation workbook that can help many people solidify their understanding and move into application.

My favorite passage is the definition itself:

Word of mouth is about genuine conversation. Word of mouth marketing is joining that conversation, participating in it – but never manipulating it (sic) in any way.

I also like the practicality of the ‘Five Ts’ which show how to make the concepts actionable:

• Talkers: Who will tell their friends about you?
• Topics: What will they talk about?
• Tools: How can you help the message travel?
• Taking Part: When should you join the conversation?
• Tracking: What are people saying about you?

There are however two areas where I wish the book were stronger:

First, it could really benefit from more extensive and interesting examples and case studies. Pat little stories about how popular Krispy Kreme was or how much Andy likes the place where he gets his hair cut undermine the majority of the book which compellingly makes the case that WOM is a lot more than just the buzzword-of-the-week. Five or six deep case studies with compelling applications that demonstrated real-world process and results of the principals the book promotes would have really helped both educate and persuade.

Second, the book does occasionally fall into the trap I feared would have consumed it – with shallow descriptions and arguments - essentially lazy passages where hype overcomes depth. One glaring example is the use of TIVO as a WOM case study. Sure Tivo users are passionate – but the clear business truth is that despite this passion the company has failed throughout its life to transform that passion into financial success. Mentioning one fact without the other is disingenuous or sloppy. Or worse it gives the impression that WOM is the goal rather than a means to a business objective.

On balance however, I strongly recommend this book for anyone who is struggling to get a handle on word of mouth and wants to decide if it’s real and how it relates to you. It’s a quick read and worth $17 to get a clear handle on what is and will be an important marketing tool.

PS: If you want to really learn more, consider the WOM Summit coming up in DC in a few weeks. I'm planning on attending.

Some other reviews: Here, and Here, and here.

Posted by Craig Danuloff at 11:48 PM | Permalink

Amazon continues to be one of the few online retailers that doesn't think like a retailer. Yesterday they announced ClickRiver, a proprietary pay-per-click network through which they will auction text ad space on Amazon.com product pages. On one hand this isn't new because they've run various forms of Google ads on Amazon pages for some time now, and the store long ago became a multi-retailer location with their Amazon Marketplace and other partnership initatives.

But ignoring the debate about whether they should be the click broker themselves or tie into another network, this move shows that Amazon understands a lot more than most about the way people use the web and the role their pages play in that process. In essence:

• People are interested in a huge array of products and services, many of which even Amazon doesn't carry. The ClickRiver ads are limited to products/services complimentary to any given product and non-competitive to Amazon.

• People know how to click (and leave) a web page to go find what they need. Not putting these ads there will not prevent them from wandering off to find these accessories (or whatever) if that is what they want to do.

• Their SEO success brings a lot of traffic, which both creates revenue also drives bandwidth and hardware costs. A great many of these clicks/people probably are looking for something related to the items but not the items themselves.

• If they're going to leave, you may as well get paid. Even if Amazon has conversion rates 3-5x the industry average (and I don't think they do) there are tens of millions of visitors leaving every day without buying something. But Amazon knows something of their interest and intention based on their navigation - the odd of ClickRiver ads being relevant is very high.

I just did some back-of-the-analytics math on for another retailer, and found that at $0.50/click the value of 1/3 of non-purchasing uniques would currently be worth upwards of $2M annually. That's a double-digit percentage of total revenue in this case. Could they capture that high a percentage, or that CPC, and do so without cannibalizing their own sales?

Impossible to say. Worth figuring out.

A related thought: Google Adsense should enabled any retailer to do this by making it possible to filter the ads displayed on your page by item/brand/category in addition to just blacklisting certain competitive URLs. They're scanning the target page anyway to calculate their quality score they so should be able to do a pretty good job assuring that ads are in fact complimentary and non-competitive. It would be great to allow all retailers to apply Amazon's strategy.

Posted by Craig Danuloff at 6:20 AM | Permalink

Omniture today releases an update for SearchCenter, version 2.3 which is primarily aimed at improved compatibility with the various paid search engines. Here are the listed features:

• Additional Search Engine Support: Ask.com has been added to the high profile line up of search engines supported by SearchCenter. With these additions, SearchCenter now supports 8 primary search engines and 40+ total search Web sites throughout the content network.

• Multiple Currency Support: As companies employ multi-national search marketing campaigns, the ability to manage those campaigns from a single interface is becoming more important. SearchCenter 2.3 now supports the use of multiple currencies for reporting and management of search engine marketing activities.

• Support of Yahoo! (Panama) Release: SearchCenter has been updated to support the enhanced functionality of the new Panama release of the Yahoo! search engine that is scheduled for release in Q4 of 2006. Whenever you decide to migrate to Panama, SearchCenter will provide you with a powerful tool to manage your Yahoo! search marketing campaigns.

• Enhanced Exception Handling for Google: SearchCenter 2.3 has been enhanced to support the management of exceptions in regards to the Google search engine. This enhanced functionality streamlines the management of trademarked and other controlled keywords and text ad content.

• Support of Google Site Targeting: SearchCenter 2.3 now supports the use of the Google Site Targeting feature which allows the search marketer to choose which content network sites will display their ads.

We continue to add new clients to those we're managing with SearchCenter, and are glad to see timely support for Panama and other changes the engines are making. We've been testing 2.3 for a couple of weeks, and have just begun to import ASK campaigns and are getting our first MSN API-keys later this week. Watch this space for reviews on how SC performs with these newly supported engines.

Posted by Craig Danuloff at 9:07 PM | Permalink

Today Netcraft tells us there are 100 Million websites. CNN heralds it as a milestone. It strikes me more like the US hitting 300 million or the earth hitting 6 Billion. In other words, it isn't good news.

A little more color from the report:

The 100 million site milestone caps an extraordinary year in which the Internet has already added 27.4 million sites, easily topping the previous full-year growth record of 17 million from 2005. The Internet has doubled in size since May 2004, when the survey hit 50 million.

Think for a minute about 100 million websites, and the pace at which they're being created:

1. How many of these sites exist only to foster the pagerank of another site? For a long time (and still) many SEO's create 'fake' sites to increase the number of inbound links.
2. How many of these sites exist only to gain search traffic for Adwords revenue. The blog version is now called splogs, but these have existed in many forms for some time. Many are auto-generated and contribute nothing to the net, in fact they detract from it.

I'd not at all be surprised to find 25% or more of the world's sites fit into one of these two categories. Maybe a lot more. Of the rest, I'm sure far less than half are of any commercial nature, but this still leaves 30-40 million sites. Some questions about these:

• How many have no search engine optimization - the webmaster hasn't even done the basics to help the engines properly classify what they've got (let alone all those terms they would legitimately want traffic for but aren't getting).
• How many have horrible navigation and worse copy and make it really hard for the visitor to easily know what they're trying to say (or sell) and why?
• What percentage don't at least basic analytics code on them and/or have webmasters that do not look at that data AND do something because of it at least once a month?

I would easily expect the answer to each of these questions to be at least 75%. Does that sound harsh? Do you think there are 7-10 million websites which exist for a good reason, are reasonably well designed, optimized, and analyzed? Sounds optimistic to me.

On the other hand, my analysis would leave 25-35 million sites that need optimization, architectural and content improvements, and better measurement. That sounds like an opportunity.

Posted by Craig Danuloff at 2:46 PM | Permalink