I'm an avid reader of Avinash Kaushik's blog, Occam's Razor, and I quietly anticipate each new post. In general, Avinash is incredibly insightful, thought-provoking, and just fun to read.
But as a fan and follower of his work, I was surprised by many of the statements in his recent post, "Measure the Real(tm) Conversion Rate & Opportunity Pie." It’s almost as if Occam cut himself shaving.
It’s possible that Avinash didn’t have time to examine every angle—especially since he was putting presentation slides together for London at the time, and we all know how that goes—so I’d like to address the more dangerous and misleading conclusions that can be drawn from so-called “Real™ Conversion Rate” thinking.
The article first caught my eye when it described, among other things, how to "…improve your Conversion Rate in 10 minutes by doing nothing more than applying simple math." Sounds fantastic! After all, here we are as an industry spending tens of millions of dollars measuring marketing efforts that cost tens of billions of dollars when, all the while, a solution needing less time than a cup of coffee sits under our collective noses. “This will be great,” I thought.
The problem with the argument presented in the article is that it strikes me as a marketer's version of the Anthropic Principle: the notion that the universe is the way it is because we're in it. Such scientists are that the forces of Nature are balanced just-so because every other variation would not allow Life (i.e., us) to exist to observe it.
Applied to marketing and analytics the message becomes: "We're professionals who toil to effectively persuade. All those industry-typical, outrageously-low conversion rates can't be our fault because we work too hard, so there must be something wrong with the math." In other words, let's fit the data to our expectations, rather than the other way around.
This fascinating approach to improving conversion consists not of converting more traffic, but filtering out what qualifies as “total traffic.” I like to call this our “Unreal™ Conversion Rate.” It's kind of like deciding the crime rate is too high, and therefore the solution is to make murder legal -- thus lowering the crime rate by simply not counting an entire class of crime. Of course, our government does this every day in under-reporting the budget deficit when it uses current Social Security receipts to offset current year expenses, despite the fact that those receipts represent future debt still to be repaid. Ah, but I digress.
The same logic was applied to women’s dress sizes in the U.S. when the Department of Commerce withdrew the standardization of woman’s apparel in January, 1983. It’s no secret that American women—and men, and children—have gotten heavier since then. So, in a blatant appeal to vanity, a “perfect Size 6” became a “Size 4” or lower! Today, there are even negative sizes for our svelte ladies. One might otherwise conclude that the thin have gotten thinner. Yet, sadly, thinking thin doesn’t quite do the trick.
The article then states something spot-on: "The fastest way for you to improve your conversion rate is to figure out what is the number of people who are in play for even remotely being converted.”
Yes, I totally agree. But then the piece goes on to describe three suggestions for achieving such improvement and, in so doing, demonstrates the H.L. Mencken witticism that "For every complex problem there is a solution that is simple, neat, and wrong."
#1 Bounce Rate
An interesting approach on bounce rate, defined in the article as "…the minimum time that someone has to commit to your website, just ten seconds, for you to even have a chance of convincing that visitor of anything." At first, this sounds reasonable, at least until you start thinking a bit deeper of some actual scenarios in which it would be in play. The first time someone comes to your site, it may absolutely be true that they do a 10-second-ish scan of your homepage. But what about regular, repeat customers, the backbone of any successful business?
When you go to Amazon.com to buy a book, do you stare at the site for 10 seconds, waiting to be convinced? Or, already familiar with Amazon's fairly standard navigation, do you instead quickly type in the book/DVD/gadget you were actually looking for? In fact, wouldn't a repeat customer spending more than 10 seconds at a familiar interface indicate something were wrong rather than right?
And wouldn't this suggest several other insights? For instance, there are certain situations where the bounce rate is an indication of something wrong in the selling process rather than customer disinterest. Further, that there are some situations where the mental state of the prospect indicates they know precisely what they want, rather than just idling around.
These two points combined suggest that there are any number of scenarios and states of mind of the personas visiting the site—all of which must be accounted for in fulfilling a full spectrum of customer needs and goals. It seems unreasonable that we can infer intent based on bounce rate. (Actually, I'd assert that we can't infer intent by any analytic-centric approach; we can only use analytics to measure the efficacy of modeled intentional paths. But that's a different discussion altogether).
So, this huge discount of10-20-30% of traffic caused by "bounce rate" is really just a fudge factor so we don't have to feel as bad for not really planning out complex sets of scenarios.
#2: “(If you use Web Logs) Filter out search bots, image requests, 404 errors, website monitoring software ‘visits’ etc.”
To be frank, this advice has been around 1999. The article claimed that a conversion "realization dawned" thereby shedding some insight. If you're still mining your own web logs and don't know to filter image requests, or still using software that doesn't filter it automatically, then you don't need a filter, you need a good hard kick in the ass. Maybe it's time to splurge on indoor plumbing, or a color TV.
Then again, if you're still doing all that AND having industry-standard conversion rates of about 2.5%, then what does that really say about the analytics industry in 2007?
#3 “Use Customer Intent”
starts out so well; so delightfully, painfully well: "One of the biggest
mistakes business[es] make is thinking that every visitor to the website is
fair game.” I just knew the next
statement would be something insightful, such as "not every visitor is
there for the same reason,” or "visitors might be motivated to achieve
different things on your site" or something along those lines.
Instead, we are treated to a stereotyped car dealership analogy that belies a fundamental misunderstanding of the sales process. What's next, ‘All Democrats are Commie Pinkos?’ or, ‘The darker the berry, the sweeter the juice?” You can read the example yourself in the article, but the gist is that the big evil car salesman is going to do every unethical thing in his stereotyped book to trick you into buying a car. And you, the noble visitor, have loftier goals that none dare call conversion.
So close to making some headway, the article continues: "not every visitor ... is there to buy" (true! true!), followed by the disappointing "not every visit ... is an opportunity to convert" (false! false!). In fact, your visitors are there precisely to convert as long as you realize that in order for your business to achieve its goals, the customer must be allowed to achieve her goals first.
I think that the fundamental misunderstanding of the article is best summed up in the following paragraph where it states "Using Market Research or Website Surveys or other methods, attempt to compute why Visitors come to your website.”
Wait, you're gonna COMPUTE why Visitors come to your website? If you don't already know why people come to your website, what will "computing" it do for you? I can compute the path of the earth around the sun, but does that tell me why gravity works the way it does? I think not.
And what self-respecting business owner needs Market Research or Surveys to determine why people come to the site? To be sure those things can give insight into why people don't come to the site, or why they come to the site and then leave too quickly (after all, that's the proposed benefit to using Bounce Ratio). But if you're in business and you don't know why people come to your site, then you're in a heap of trouble, and a survey ain’t gonna save you.
There's no amount of analytics crunching that is going to tell you why people come to your site. Rather, one must do the hard work of putting oneself in the visitor’s place and empathically plan the persuasive scenarios in which they come to the site. Then, aggressively use analytics to measure how well you've implemented those scenarios; THAT's the secret to significantly higher conversion. Otherwise, we’re merely forcing the numbers to rationalize our own poor planning.
A “Real™ Conversion Rate” might help your company’s self-image, but it certain won’t improve your bottom line.