The Emetrics Summit in Santa Barbara, CA, was held last week. Reviews and comments have been flooding my inbox. Although I couldn’t attend myself, John Quarto-vonTividar, co-inventor of persuasion architecture, presented on persuasion scenario planning and the call for an open standard for scenario definitions, including an XML standard.
I’ve also heard rave reports about the Web Analytics Association members meeting, which had over 150 folks in attendance. I’m really sorry I had to miss it.
In absorbing everyone’s comments over the last couple days, I’ve discovered there’s great progress on the analytics front. Enthusiasm for the space is growing exponentially. No doubt, any company worth its salt is finding itself more deeply invested in its Web analytics. My good friend Jim Novo, author of “Drilling Down,” shared some of the most encouraging news.
“Ford Motor Company has run a pilot where they are using the info from a ‘Build Your Car’ configurator on the Web to predict demand for certain cars and features,” said Novo. “When they matched the predicted data from the config to actual sales data, the fit was simply amazing. So amazing, you would immediately question if the data was tortured somehow to show this result.
“But that’s the nature of near frictionless environments like the Web,” he continued. “You tend to get behavioral data that’s simply more true then asking people their opinions, which is the more common way to get affirmation for auto design from the customer.
“What is probably more important from an analytical culture perspective,” he said, “is that this gigantic metal-bender car company with very long lead times is actually using this Web data to modify production plans because it has been such a reliable predictor of demand. If they can make this kind of thing happen at Ford, you can do it where you are. This is a monumental achievement.”
Quarto-vonTivadar, known around our office as “Q,” had his own insightful observations about the Ford Direct presentation. He also had a few thoughts about where people are at in relation to testing and analytics. The following points are an extrapolation of what he articulated at the water cooler. He’s literally our resident rocket scientist (having worked for NASA) and has been doing multivariate testing ever since.
Why Not Measure Micro Conversions?
The Ford Direct site accounts for 10 percent of total sales, a very respectable number. Obviously, Ford’s macro conversion can be measured as a sold automobile, so the site’s ability to generate a lead for a local dealer is the on-site macro conversion. But throughout Ford’s presentation, there seemed to be little notice of the several micro conversions that can be measured and optimized. Dealer searches, car configurators, and inventory searches within a fixed-mile radius are all micro conversions that can clue the marketer into where a visitor might be on her path to a macro conversion.
The lesson to be learned? Take inventory of all possible actions a visitor can take on your site, actions that may fall short of an actual macro conversion but that can indicate a visitor’s on a buying path. Then create persuasion scenarios that build persuasive momentum for those actions. Seeing their desired car in their desired color might instill resolve in one customer type, while another would become resolved from knowing his neighborhood dealer has a large variety of a particular model to choose from and he may be able to make a deal. Just because visitors aren’t leads today doesn’t mean you can’t optimize making them leads tomorrow.
Hot, Medium, and Incentives
Quarto-vonTivadar also found it interesting how Ford Direct scored leads generated on the site. Leads are put into one of three buckets: hot, medium, and cold. The hot leads convert into sales at a ratio of six to one over cold. But due to incentives offered to cold leads, their average sale price was significantly higher than that for hot leads. He wondered if there’s a way to make the hot leads equitable to the hot in terms of average sales price.
Process vs. Talent
As the last presenter, Quarto-vonTivadar got to hear a lot of folks tout their conversion rate increases. Some of the better numbers were increases of 7 percent, 10 percent, 12 percent, and 23 percent.
In contrast, the typical projects that come through our office see conversion rate increases ranging from 200 percent up to 5,400 percent. Quarto-vonTivadar said he knows our staff is smart but doubts we’re five times as talented as the next guy. These types of numbers are the result of process, not talent.
To realize such increases requires a process and methodology that account for the thousands of variables that affect your ability to persuade a prospect to take action. Planning, measuring, and optimizing persuasion scenarios go beyond genius talent and move conversion rate optimization into a more scientific, holistic method that changes the focus from page analysis to scenario analysis.
When Quarto-vonTivadar was listening to people tell their conversion rate stories, he questioned if some of these increases didn’t reflect the standard deviation, or the typical fluctuations, of regular customer activity. Here’s an exert from his most recent whitepaper on A/B testing:
Let’s say that our friends, Bryan, Jeff, and Hal are standing in a room. Bryan is 6’1″ tall, Jeff is 6’2″, and Hal is 5’9″. Their average height (the mean) is 6’0″.
Now Bryan meanders into the next room and meets Antonio, who is 5’4″ and Shaq who is 6’7″. The average height of these three is also 6 foot — but what a spread in the numbers! Variance is how we represent this spread numerically. The more varied the range of numbers that went into the average, the larger the variance.
Variance leads directly to other measures, such as “standard deviation,” which you might think of as the average variance of each item. You’re likely to see it quoted after an average as the number following the phrase “plus or minus,” as in “6 feet, plus or minus 2 inches.”
Ever the scientist, he also wondered why more folks weren’t reporting their top-line conversion rates as “4 percent plus or minus 2 percent” to account for this deviation. (The white paper goes into much greater detail.)
Web analytics are extremely valuable and in next few years will become even more so to keep businesses viable and profitable. But they’re grossly underutilized and, in some cases, misused. To maximize your ability to measure online behavior, you must first plan persuasion scenarios. These scenarios will provide you with a predictive model of customer behavior that can be most effectively measured and optimized. (More on that in the next column.)
I’ll be excited to hear for myself what’s going in the analytics world when I present at the Emetrics Summit in London next week. And if you didn’t catch us in Santa Barbara, we will catch you at the Washington, DC, Emetrics Summit in October.
In the meantime, send me your biggest Web analytics challenges.
Sponsored content in collaboration with Marchex. When it comes to brand keyword bidding, most tests show that it makes sense to bid ... read more
Emily Ma, product director of Tencent’s advertising platform products department, was a keynote speaker at ClickZ Live Shanghai where she discussed the ... read more
The terms that customers type into your site search function can help you to gain an understanding of user behaviour and can be used to optimise ... read more