Reductive Analytics and Testing

Let’s say you’ve got a website. Consumer e-commerce. You get lots of visitors and you have lots of pages. Most of your traffic is from search engines, and your keyword range is wide, with a natural head-tail power law curve to it. Some of these people buy things, most do not and like any other business owner, you want to figure out how to make more people buy stuff from you. Nothing on the site is broken, or screams for help, like broken search or poor navigation and the site does sell a few things, just not a lot.
For me, the key has been segmenting visitors into behavioral cross sections. Like I’ll first look at the keywords the visitors came from and see if there are any keyphrases or root-keyword groups with abnormally low time-on-site averages or conversion rate. If one or many do then I would stop paying per click on those words if I was and I would exclude them from further research. I do this to rule out keyword-intent as a possible source of low conversion rate.
Then I’ll look at entry pages the same way. If I find an under performing landing page, I’d also expect to see a high bounce rate. If I didn’t have any landing pages relevant to the traffic the bad one was receiving, it would be come a candidate for testing.
The same analysis would then be applied to key pages in the shopping and checkout experience and if I found any un-replaceable pages I’d move them to the testing phase.
The pages selected for testing would be broken down into component pieces and variants of each chunk would be run through tests. Each variation would have large differences and I’d only test 2 or 3 of each, perhaps even using a “bad” one. Testing a low number of variations will allow me to run the test quickly and look for which page sections produced the biggest range in performance among the variations. These are the important parts.
Then tap the skills of experts in design or communication for interative testing and fine-tuning of the important page elements. When your improvements plateau, return to an earlier step and repeat the process.

Read More

North End Boston Searches

I live in the best neighborhood in the world, so I got curious today what people search for when they’re looking for info on the north end. Clearly lots of searches are going to only contain restaurant or festival names, but the searches below give me some indication of what he overall niche probably looks like. As expected lots of restaurant, apartment and festival schedule searches, but one surprise (At least to me) is the hotel searches. I don’t think I know of any hotels in the north end, and perhaps thats why there are so many people looking for them, a possible opportunity?
Here’s the top 34 terms as identified by keyworddiscovery. There are a bunch more but they’re all variations on the same.

keyword searches per month
north end 675
boston north end restaurants 257
north end boston 249
north end boston hotels 232
boston north end 202
north end restaurants 198
north end restaurants boston 169
north end boston restaurants 165
restaurants north end boston 152
bostons north end 70
north end boston apartment 38
north end pizza 24
north end cafe 18
north end boston ma 18
north end restaurants boston ma 14
north end boston massachusetts 8
restaurants boston north end 8
north end ma 8
restaurants north end boston ma 7
boston north end hotels 6
north end apartments 5
the north end boston 5
north end of boston 5
north end feasts 4
bostons north end pizza bakery 3
north end middle school 3
north end in boston 3
image north end boston.jpg 2
dolce vita north end boston 1
north end italian restaurants 1
restaurants in bostons north end 1
Read More

Frank Luntz’s Words that Work and Youtube Dial Sessions

I just got finished audio-booking pollster Frank Luntz’s new book Words that Work. While I’m very different from Luntz politically (he’s the guy who renamed the estate tax to the death tax, and is anti-“illegal immigration”) I’m absolutely fascinated by his work with language, specifically his testing methods. The book is a great read/listen especially for an online marketing professional who relies on words and images entirely to sell a product or a viewpoint, lots of great stuff in there.
But what really piques my interest in Luntz, and has since I first became aware of him when I watched the documentary “The Persuaders”, is his testing methods and how they overlap with the split and multivariate testing stuff I’ve been doing. Obviously testing wording variations with something like Google’s Website Optimizer is easy enough, but he uses another method that I think is missing from the internet marketing toolbox: dial sessions.
A dial session is shown in the movie and he refers to them in the book, but the idea is that a bunch of people watch a video and turn a knob at the same time. If they agree with the video they turn the knob one way and if they disagree the other way. I’m going to look into implimenting an ajax slider control in conjunction with youtube videos to create an online dial session system. I think the sheer breadth of content and test subjects available makes this a goldmine of useful data.
Obviously there are some challanges to address, like recording second by second data, correlating video content to slider position and interfacing with youtube’s flash widget, but I don’t reckon it’ll be too hard.
Stay tuned, I do plan to share my tool when developed.

Read More

Political Search Marketing Experiment Part 1

I didn’t notice it until today, the 3rd but the perma link page for my boston city council SERP probe is 5th at google but nowhere in msn or yahoo. I had expected the homepage to show up first, before the perma link, but the page has a freshtag of Feb 1, 2007, the day after I made that post.
Incidentally I also made it in the evening on the 31st.
I further expect this page to disappear for a few days when the freshtag goes away.

Read More