Archive for September, 2006
Posted on September 28, 2006 by Robbin Steif
How do you increase your conversion rate when your total budget is only $1500?
This is the question that I confronted recently when someone submitted a Contact Us form on our website. She made it very clear that she had very little budget to increase her conversion rate.
I thought about this for a while. Here are some of the things that owners of small businesses, who are not ready to spend much on outside conversion firms or consultants, might consider:
Do some cheap user testing.
Go find five to eight people who you know vaguely but who don’t know much about what you do. (For example, the parents of some of your friends, if they live in the same city. If you can get people who mirror your target customer, so much the better, but as long as your product is not so technical that the individual won’t understand the vocabulary, this is worth doing even with non-target users.) Sit down with them while they look at their computer and you are recording their thoughts (and tell them to talk out loud since you can’t read their minds.) For 2-3, ask them to do whatever your site does (e.g. make a purchase. Request more information on your site.) For the balance, ask them what they think they could achieve on your website, and see if they can achieve it. Take great notes, keep your thoughts to yourself while they are working, and give them something for their time. If they are distant friends, they may be uncomfortable accepting your money but might love a $20 gift certificate to Starbucks.
Cost: about $200, depending on whether you need to compensate anyone with real money.
Install some free analytics and read them
Sign up for a copy of Google Analytics (they are now open to everyone.) Ask your web guy or gal to put the code that Google gives you on every page, or better yet, in a file that gets included in every page. Read the Google help section here, and when you get to a part that you don’t understand, just skip it. (I’m serious. You’ll learn it in time, but if you stop and say, “Don’t understand it, can’t figure it out, can’t afford to hire someone,” you’ll lose all your momentum.) Every week, look at one new report so that you gradually learn what the analytics mean. Be sure to read all the little notes at the bottom of each report so that you really understand what the report means. Be sure that you create Google Analytic Goals — maybe you will have to stay with your analytics for a couple of months before you know how to do that yourself. Subscribe to the Official Google Analytics blog if you are just learning analytics. If you know them, subscribe to This Just In and ROI Revolution’s blog.
As you learn what your analytics mean, start to understand what changes they are suggesting. For example, where, in your most important (top) content is your bounce rate too high? People come to websites for lots of reasons – sometimes just to get a phone number – but a high bounce rate often signals a problem with that page.
Cost, $150 for your IT guy’s time.
Learn as time permits, as cheaply as you can
Buy a copy of the Eisenberg’s Call to Action and Persuasive Online CopyWriting and Steve Krug’s usability-for-newbies book, Don’t Make Me Think.
Cost, $74.54 if you buy them all new.
Subscribe to some blogs and email marketing. You already found this blog, so let me suggest The Site is Dead (that’s Matt Roche’s blog – he is the CEO of testing company Offermatica.) Other owners of blogs about conversion rate are welcome to get on and say, how about my blog? Blogs are wonderful for learning about a topic like this because you get a little bit of information at a time, not a book that you look at and keep meaning to read.
Cost: free
(Note: Steve Krug’s book is designed to be short enough to read on a long airplane ride. Call to Action is laid out in very short, bite-sized chunks.)
Do some testing using Google AdWords
Now that you see where some of your problems are (user testing and web analytics) and have learned what some solutions are (learning as time permits), you can do some inexpensive landing page tests. Take one page that your users had a really hard time with, and create an alternative to it that your new education indicates might be a winner. Create a Google AdWord campaign with two ads (they let you create multiple ads which they show in rotation, as you probably know). Be sure that one ad lands on the current “bad” page and the other ad lands on the page you are testing. Let it run for a few months (because if you are a small business with a small budget, you probably don’t have the kind of traffic to get statistical relevance in a week.) Measure which one “wins” — you have to define “winning” yourself, but you can measure in three different ways:
- Add the Google AdWords conversion code to your success page (often, a “Thank you very much” page)
- Change the titles of the ads every so slightly so that you can see them differently in Google AdWords (and I really mean slightly. Example, add a colon after the headline of one)
- Turn off autotagging in Google Analytics, create your own tags with a different utm_content tag, and measure in Google Analytics. Here is a link to doing that one.
No matter which way you go, be sure that you change your Google AdWords Campaigns so that it shows your ads evenly (i.e. do not let it “optimize” the one that gets the most clicks, since you care about conversion, not clicks, and you don’t want to see your test get skewed by high click-through yet low conversion ads.)
Cost for this one depends on how many clicks you have to get in order to learn something from your tests. If one of your tests shows your conversion rate going from 1% to 2% for people who landed on that page, you can get away with only 1000 clicks for each test version. If you can find a keyword at $.30 a click, your cost will be $600. Plus you will spend $500 for the creation of a test page.
Total cost: $1524. Remember that the biggest variables are in the Google AdWords testing: can you get the keywords that cheap, and is the difference in your new conversion rate going to be high enough to let you pay for only 1000 clicks per test version?
Robbin Steif
LunaMetrics
ps I published this again in an effort to get the feeds to work correctly. Sorry. RFS
View Comments (No Responses) | Categories: Conversion Science
Posted on September 25, 2006 by Robbin Steif
So I continue here with my Regular Expressions (“RegEx”) lessons. I am learning RegEx only because so many customers use Google Analytics, which throws the code at the customer with very little explanation.
This next lesson is about the question mark. This time, Google does a pretty good job of meaning what they say:
? Match zero or one of the previous expression
When they say, “The previous expression,” they mean, the character that comes right before the question mark. Since that is still pretty opaque, let me shine some light here.
Let’s say that you have an economics website and you only want to look at the referrers that have the word “labor” in their title. But some of those referrers come from non-US countries where they spell it “labour.” You could create a filter like this: labou?r
That way, it will match “labour” (which does have a “u,” which is the previous expression) and labor (which has zero of the previous expression, i.e. no “u” is included.)
You cannot use it like this : labo?r, or at least, not for the same purpose. It’s not a wildcard that you stick in between the o and the r to match any letter. The only matches would be to “labor” (zero of the previous expression) and “labr” (Thanks Serge for catching my error.)
Backslashes \
Dots .
Carats ^
Dollars signs $
Question marks ?
Pipes |
Parentheses ()
Square brackets []and dashes -
Plus signs +
Stars *
Regular Expressions for Google Analytics: Now let’s Practice
Bad Greed
RegEx and Good Greed
{Braces}
Minimal Matching
Lookahead
Robbin
LunaMetrics
View Comments (2 Responses) | Categories: Google Analytics, Regular Expressions
Posted on September 24, 2006 by Robbin Steif
I put CrazyEgg up on my site for a short period of time, mostly so that I could try it out before recommending it to a customer. And I loved it. First the explanation and then the editorial.
CrazyEgg is a website overlay on steroids, although they may be devastated to read that description (more below). Not only does it give you the standard overlay information that you can get from Clickmaps (or whatever overlay you have), but it also shows wherever people are clicking, even if that is not a link. For example, I saw this in my short test:
Notice how the words, “So now you have a website,” which are just a .gif, are more popular than the real links are.You can tell that at a glance because the overlay button is light blue, not dark blue — it is closer to red, and the redder the button, the more popular.
CrazyEgg makes this even more clear with the use of their heatmap. Notice how my “about us” link is my most popular one:
I always knew that About Us was my second most important page, but I never really “got it” quite as well as I did when I saw this heatmap. Pictures are worth thousands of words even for analysts, it appears.
Now that you see how cool it is, let me explain the issues.
The FAQs and other on-site Help on the CrazyEgg site don’t just suck – they are non-existent. Want to change the time period of your test? No idea how. Want to know what the classifications to all those clicks are? If you work with web stuff all the time, you’ll get it, otherwise you are lost. Want to know why they only show about 90% of the clicks that they report you have? No answers. Want to know what it means to archive four snapshots? Nope, no answers there either.
Furthermore, there is no way to add the code to an include so that you can just use it all the time, like you would an overlay. It would be the perfect solution for all those lost souls who are finally realizing that as great as GA is, the overlay is lousy, so they might as well pay $20/month for a great overlay.
I wrote CrazyEgg and asked why you have to tell them the name of every page you are adding your code to (and then still get the same code for each page). Hiten Shah (who was incredibly good about answering questions) wrote back very promptly three times and explained that they don’t want to be lumped in with all the WA and overlay providers. They are a testing solution, not a WA solution, he wrote. I think that’s just ridiculous (Sorry, Hiten). Scandanavian Airlines positions itself as a business airline, but they don’t make it hard for tourists to travel on them. CrazyEgg could easily position itself for testing and still allow people to use their services for day-in, day-out web analytic overlay work without compromising their marketing strategy.
Hiten tells me that he is a reader, so let’s see if he comments (and what he says.)
Robbin Steif
LunaMetrics
View Comments (5 Responses) | Categories: WA Tools, Web Analytics
Posted on September 14, 2006 by Robbin Steif
As regular readers know, I am learning about and sharing my lessons on Regular Expressions. In my last post on this topic, Part III, I wrote about the opening anchor, aka the carat symbol. It’s used at the beginning of some regular expressions. Today I am writing about its sister, this anchor $, which is sometimes used at the end of Regular Expressions (RegExen. I wish I had one of those “cool” smileys to easily insert here.)
In the Google Analytics very unhelpful Help section on Regular Expressions, it says this about the dollar sign:
$: match to the end of the field
What they really mean is, don’t match if the string from my website has any characters beyond where I have placed the dollar sign in my Regular Expression. The dollar sign signals all the characters that I want to match to. (This is the hard part about the explanation, it’s always hard to explain what the target is and what the RegEx is.)
So let’s say that you have some pages that end in htm and others in html. You want to write a Google Analytics Step 1 (part of a goal) for your email sign-up form, but you only want the new .htm version. Your RegEx might look like this:
/email-form\.htm$
The dollar sign tells the Google Analytics, if the page on your site has anything after the final “m” in “htm,” it doesn’t count as a match to this expression. Notice that I also used a backslash before the dot so that Google Analytics interprets it just as a dot, not as anything special.
I understood what it does pretty quickly, but I had to put this post off for at least a week in an effort to understand why anyone would ever use a dollar sign. Here’s the problem: if you have any campaign code whatsoever that gets attached to this string when someone lands on your site, there is no match. For example, let’s say someone lands on your site and when he looks up at the address bar, it says this:
/email-form.htm?cid=123
As soon as there is campaign code, the dollar sign anchor will tell GA that there is no match. And it can be hard to remember everywhere that you would have campaign code and everywhere that you wouldn’t.
So forget the fanciful examples – why would anyone ever use it?
Once place you might use it would be with an IP address (for a filter, someone that you are trying to filter in or filter out.) You might have an IP addresss like this that you are screening out: 12\.34\.56\.78, which matches 12.34.56.78, but you want to be sure that it doesn’t match 12.34.56.789 — so you set up your expression to be 12\.34\.56\.78$ . And if you want to be sure that it doesn’t match 512.34.56.78 as well, you should use the beginning anchor ^ (how’s that for a reminder and a link back to the last part of this series), ^12\.34\.56\.78$
You should also be able to use it, Justin says, with the “on the fly” filter that you can do with every report (the little box near the top of every screen.) For example, if I wanted to know how many different variations of search terms end with my company’s name, I should be able to type in LunaMetrics$ in that little box on one of the search term reports (like Marketing Optimization > Search Engine Marketing > Overall Keyword conversion, and only see search terms like LunaMetrics or Call LunaMetrics or who the heck is LunaMetrics). I have never gotten that one to work. I do see potential there, though, because I can use a carat^ at the beginning of an in-report filter, like this: ^LunaMetrics, and I see all the search terms that start only with my company name.
And you can use it in the profile filters.
Am hoping one of the GA gurus will get on to say why the $ anchor doesn’t work in the in-report filters. After all, I am only taking lessons here, and sharing them with the world.
Backslashes \
Dots .
Carats ^
Dollars signs $
Question marks ?
Pipes |
Parentheses ()
Square brackets []and dashes -
Plus signs +
Stars *
Regular Expressions for Google Analytics: Now let’s Practice
Bad Greed
RegEx and Good Greed
{Braces}
Minimal Matching
Lookahead
Robbin
LunaMetrics
View Comments (No Responses) | Categories: Google Analytics, Regular Expressions
Posted on September 12, 2006 by Robbin Steif
I have an incredibly selfish reason for wanting everyone to go to the eMetrics Summit, Oct. 16-18 in DC. It’s not money (I don’t work for Jim Sterne); it’s not personal aggrandizement (by the time I speak at the end of Day Three, everyone will be gone but me); it’s not even business development. No — I just want to see all my friends.
I was thinking about this while finally reading The Long Tail. Our web analytics community is a perfect example of a tail market that wouldn’t exist without the Internet, for more than one reason. But it does exist and that’s how I get to share interests with people all over the world. Internet or not, though — sometimes I just want to have a drink with my friends.
Hey, I’m realistic – I really don’t expect Lars to come over from Sweden or Steve to come over from Australia or Mike to come from S. Africa. At least not to this Washington DC conference. On the other hand, I’m so looking forward to talking to Ian Houston and Tim Seward and Justin Cutroni and June Li and Joseph Carrabis and Avinash. And Matt Roche, what will I blog about if you don’t make it to DC?
So Dylan — are you coming to the Summit? (I’ll buy the first round…)
Robbin Steif
LunaMetrics
View Comments (4 Responses) | Categories: Industry News
Posted on September 10, 2006 by Robbin Steif
This is the third in a series of lessons I am taking (and sharing) on Regular Expressions. This one is on the use of the anchor, symbolized by a carat, like this: ^. My tutor, Steve, writes about the dollar sign as well; I will handle that in a future post.
(Useful factoid: the people who work with Regular Expressions all the time call them RegEx. I have no idea how they make that plural.)
Here is what Google Analytics’ incredibly opaque Help says about the carat anchor:
^ — Match to the beginning of the field
I really understood every individual word in that sentence, I just couldn’t understand what they mean all strung together. (So I have a personal tutor.)
Here is what it means:
^ — If anything comes before this character, the string is not a match to this Regular Expression
For example, let’s say that I have two pages on my website, http://www.mysite.com/secondpage/contact/, and http://www.mysite.com/contact/.
Usually, Google Analytics, which is where I use RegEx (RegExes? REs?), perceives those two pages to be called /secondpage/contact/ and /contact/. That’s because GA already knows about the domain, www.mysite.com, and usually only cares about it if I have a subdomain (and have added the code, a technicality we won’t deal with.)
If I want to find all the strings that start with /contact/ (the second option) but just put in that same line, /contact/ for my Regular Expression, I will get everything that can possibly match the string, which will include the one I don’t want, /secondpage/contact/. This is something that has taken me a while to understand with Regular Expressions — they match everything that they possibly can, so you have to use the special characters to keep them from getting out of control.
If I only want to match http://www.mysite.com/contact/, I can use Regular Expressions like this:
^/contact/
That’s it. That’s how you use the anchor. And now, you are done. Everything after this is a clarification of one big nagging question that I had: Why would anybody use an anchor carat anywhere except here:
^http://www.mysite.com/etcetera/andso-on.php
Answers:
1) GA already thinks in terms of relative urls. It assumes the http://www.mysite.com, so when you ask for ^/contact/, it will come back and correctly show you strings that say /contact/, and you are usually saying to your boss, “They mean www.mysite.com/contact/.”
2) Anchor carats are useful in other places besides just urls. Let’s say you want to create a filter for the entire range of IP addresses in your company. However, your IP addresses all start with a two digit number, like 64.xx.xx.xxx, so you wouldn’t want to filter out something that looked like this: 164.xx.xx.xx. To solve that problem, you can use a carat: ^64 etc
Not sure if your regular expression will match the string you need it to? Use this handy tool from Epikone. You put in the string you want to match to and then the regular expression your wrote, hit enter and see how well you did. Many thanks to Justin (from Epikone) for help with this post.
Backslashes \
Dots .
Carats ^
Dollars signs $
Question marks ?
Pipes |
Parentheses ()
Square brackets []and dashes -
Plus signs +
Stars *
Regular Expressions for Google Analytics: Now let’s Practice
Bad Greed
RegEx and Good Greed
{Braces}
Minimal Matching
Lookahead
Robbin
LunaMetrics
View Comments (7 Responses) | Categories: Google Analytics, Regular Expressions
Posted on September 7, 2006 by Robbin Steif
I hear this question a lot. What do we web analysts do with our time? To make this tractable, I am going to focus just on web analytics – no split path testing, no data mining, no user testing, no… no.. just what one considers to be “traditional” web analytics. (Maybe an oxymoron?)
I think web analysis has four goals:
1. We use web analytics to troubleshoot the site and the business
Is a conversion rate down significantly, are page views/visitor way off? Is some other performance metric not performing? If so, where is the problem coming from — from a certain section of the site, from a certain product? Is that page or piece of the site broken in some way – code missing, a link not working? If the site is not broken, is the problem coming from traffic sources – an AdWords campaign that is landing in the wrong place, a referrer that no longer refers, an affiliate you forgot to pay and so conveniently forgets to send you traffic? Notice how I went from the most general information, such as conversion rate or page views, down to specific information, such as whether an affiliate is angry at us.
And if nothing is broken — how can we account for the problem? Is the metric down across all sources and all products? Is it down vs. last year (or are we just seeing seasonality?) Is there new competition, or better competition (and with that, I have to leave off, since I promised to limit this to web analytics.)
Don’t be fooled into thinking that there is no ROI here. When you find a problem and fix it quickly, you avoid days worth of opportunity loss — revenue that might have been left on the table or leads that registered with a competitor, were it not for you, the analyst.
2. We use web analytics to figure out how to make the site better.
This one is an all-nighter topic, and it is midnight already. So here are some easy examples. We might compare the conversion funnel from Lead Form 1 and see it converting from the form itself to the Thank You (success) page at 25%, yet the conversion funnel from Lead Form 2 converts at only 12.5% — so what are the differences in design? We might just look at bounce rates (single entry rates) and compare similar pages that bounce at very different rates. We look at our successful keywords in our Finding Methods report to figure out how to optimize our site for those, since those are clearly some of the ways that customers are using to search for our site. We look at conversion rate by source to decide which paid campaigns to pour more money into and which to cut. We segment our customers all the way down until we figure out what the most profitable customer looks like and where she comes from, and then we pour more resources into finding more customers just like her.
This is the area that gets the most attention from top management (unless the site is broken, in which case, they are furious.)
3. We spend time making our web analytics perform better.
Software isn’t perfect, but if we are going to invest in it (and even if it’s free, our time is an investement), then let’s get the most out of it. To use the form example from the point above – if our software supports it, we can track each spot in the form and see where potential leads bail. Better analytics can help us do a better job of spotting online opportunities. We also fix our web analytics when they break — because, why would you invest in a $50K/year package (not to mention the analyst you have to pay), to have broken analytics and learn nothing? Technically, this third goal is just a way to achieve #1 and #2, but we (I?) spend enough time at it that it was hard not to call this one out.
4. We present the information to management in a format that makes them care about the information and take action.
This is perhaps our hardest job. We get down so deep in the data that it is hard to remember the needs of management (or customers.) They have other things to do so if the analysis doesn’t come to them in a meaningful and easy format, they stop caring. I should also point out how important consistency is. Once your CEO or customer understands the metrics, you can improve upon them gradually, but don’t change the format (or heaven forbid, the metrics) arbitrarily.
I will be using some of this at the Summit, so please comment and criticize!
Robbin Steif
LunaMetrics
View Comments (8 Responses) | Categories: Web Analytics
Posted on September 2, 2006 by Robbin Steif
Since the whole SEO industry is still dealing with the departure of Danny Sullivan from SES etc, I won’t go there and will just complain about conversion on his new blog/podcast (well, commenting. Maybe he doesn’t define comments as a form of conversion.)
I checked out his new site today, and I was pretty surprised when I got on the site and saw post after post, with no comments. Hmm, I thought, maybe that’s because loyal Searchcast listeners are just that, listeners. And you get the subscription so you don’t need to go to the site… still and all, a little weird. And then I tried to comment (and all I want to tell the guy is, could you please read my post about 10 mistakes that podcasters make and recognize that mistakes 3-6 were inspired by you?) But I was unable to comment, and here is why.
First, I had to create a TypeKey membership. Already a stumbling block (like asking your buyers to sign in before making a purchase). Then I went back to the site and tried to comment but I wasn’t signed in. OK, I can do this, I though. So I clicked to sign in (with my shiny new sign up) and got this:
It didn’t matter which choice I made, either way, I got the same error message:
An error occurred: This weblog requires commenters to pass an email address. If you’d like to do so you may log in again, and give the authentication service permission to pass your email address.
Eventually, in the course of writing this post, I went back to my profile, changed it to pass my email address, and now I get the error message automatically. I don’t even have to make a choice.
When the data look strange (like your 3% conversion rate just dropped to .002% or you are one of the most popular podcasters around but no one sends you comments), it’s a red flag. Maybe something really is wrong.
Robbin Steif
LunaMetrics
View Comments (3 Responses) | Categories: Usability
Posted on September 1, 2006 by Robbin Steif
Don’t you just hate when you forget your password? Or worse yet, you forget both your password and your user id.
If you are desperate enough, you’ll be able to retrieve them. Plenty of sites make it easy to get them again (I lost my Offermatica password twice today and both times, got a new one within minutes. But that’s because they made it easy.)
The real problem for websites is when the user really does need to sign in (think online banking or looking at frequent flyer points online, for example) and can’t, because she forgot her information. So instead, she gets her customer service the old fashioned way, which is exactly what these companies are trying to get away from.
I’m not arguing for no passwords or everyone having “weak” passwords. But why can’t sites that require a password signin remind you of their format? If the user id is an email address, always, why don’t they tell you that? There is no security issue — anyone who wants can proceed to the signup screen to learn the format. If the password must be longer than 8 characters and include at least two non-alpha characters, why don’t they tell you that after your password fails, instead of just failing? We all have a “pool” of favorite passwords so that we don’t have to remember every word in the alphabet – if you tell me what the format is, I can probably narrow it down to three passwords and possibly succeed before I have tried too many times.
Robbin Steif
LunaMetrics
View Comments (1 Response) | Categories: Usability
Subscribe to this blog's feed
