Apologies to subscribers - I had to change the name of this post twice and it was probably pushed through to you in your feeds. In any case, if you want to read Part II of this post, you can find it here.

Maybe you aren’t doing e-commerce. But your leads are worth money, right? So figure out how much and tell your web analytics — that will enable you to compute your return on your paid ad campaigns, banner ads, and anywhere else you pay for traffic.

Here’s how you do it, and let’s start with the easiest example: visitors only have one way to convert into leads. Let’s assume that they convert when they fill out a “Contact Us” form. So, sit down with your last 1000 completed “Contact Us” forms, compute how many of those turned into paying customers, and what an average converted lead was worth. To keep the example easy, let’s assume that when the visitor converted into a paying customer, they either license your $50,000 software or the souped-up version that costs $75,000. Three of the 1000 “contact us” leads purchased, of which two bought the less expensive version and one bought the more expensive version. So your average converted lead is worth (50K+50K+75K)/3=$58,333 Now, spread it over 1000 leads –
$58,333/1000 = $58.33. There you go — a lead is worth 58 bucks. (It’s not worth getting too precise when you don’t have much data - you should re-evaluate every month as you get more data.)

This gets a little more interesting when there are multiple ways to convert (they can sign up for an e-newsletter, download a demo, attend a webinar.) If everyone only did one of those conversion methods, you could just follow the instructions above and compute the value of an e-newsletter sign-up, the value of a webinar registration, etc. But the truth is probably muddier and most visitors who convert probably sign-up for more than one option, especially if the end result is a $50K purchase. There are many ways you could do this, but I always like to keep things really, really simple when you are guesstimating anyway. So I recommend that for starters, you consider each conversion to be an equivalent event, and then count by the events.

For example, let’s assume that the customer who made the $75K purchase signed up for all three events (e-newsletter, webinar and download.) The first $50K purchaser only signed up for the download, and the second $50K purchaser signed up for both the e-newsletter and the download. So that’s six events spread over $175,000 of transactions, or $175,000/6= $29K per conversion event that turned into a customer. Now you can spread that over all the 1000 individuals who took action on your site (whether they bought the software or not, since our real goal is to put a value on each software download, each webinar registration, etc.) Let’s assume that the 1000 “action-takers” actually completed two events, for 2000 total events. That makes each event worth $29K/2000 = $14 and change. So each of the three events (webinar, e-mail, download) is worth between $14 and $15.

Doing the analysis this way makes sure that you don’t doublecount the different conversion events and inflate their value. (You can also weight the events based on how often they occur with a conversion, but that has to wait for a different post.)

Don’t forget to put the value of the conversion into your web analytics package (if it accepts it.) If you are using the free conversion package that Google gives you with AdWords, you will definitely have a place to include this value and compute ROI. Google Analytics also gives you this capability, as do most of the intermediate and top-of-the-line packages.

Read part II of this post.

Robbin Steif
LunaMetrics

A year ago, I was dithering about buying a research report from MarketingSherpa, because it was $299 (for a book!) I couldn’t find it used anywhere. I didn’t want to spend 300 bucks just to find out that I had wasted my money. After hemming and hawing for two weeks, I noticed that Sherpa had an amazing return policy — if you weren’t satisfied for any reason, you could return the report within 30 days for a full refund, no questions asked. The reason the policy was so amazing was, you could purchase an online downloadable version and you could still get the refund — you just had to delete the report from your computer. Clearly, they had no way of monitoring your behavior. You were on the honor system. Obviously, MarketingSherpa saw the guarantee as a form of investment — yes, there would be a few cheaters, but the people who purchased because of the guarantee would far outweigh the money lost to cheaters.

I am always trying to convince my customers to have similar policies. They say things like, “If the customer returns the product to us, it becomes unsaleable. Or they allow returns, but make the terms really onerous. (Limited time, must be in the original packaging, restocking charge, etc.) I should lecture them on selling seconds on eBay, “open box” sales, but that’s a different post.

A great return policy is a marketing investment. In the short term, you are betting that the profit from the customers who buy (and wouldn’t without the return policy) outweighs the cost of the few who do return the product. Over the long term, you have to factor in the benefits of acquiring a new customer and what their lifetime value will be, as well as the value of the customer who returned the product. Maybe she did return it, but for a good reason. She sees that you are a pleasure to work with, and will be back.

BTW, the new MarketingSherpa eCommerce Benchmark Guide 2006 shows that returns as a percentage of online sales, for web-only business, is 7.3%. They have lots of other fascinating statistics as well.

Robbin Steif
LunaMetrics

A while back, I heard a report on the radio. Scientists had found four factors that were associated with breast cancer. One of them was a high-level of education. So does that mean that if I skip college and grad school, I am less likely to have breast cancer?

The correlation between education and breast cancer is just that — it is correlation, not causation. We find them together, but that doesn’t mean that one causes the other. In fact, there is at least one (and perhaps more) variables that are lurking in the background which is really causing the cancer. For example, highly-educated women may be less likely to have children when they are very young, and it is may be the child-bearing act that affects cancer.

I use breast cancer to illustrate lurking variables because it is so clear — we know that we can’t skip college and avoid the problem. Online, the issues of causation and correlation (and lurking variables) are just as important, but often not as clear.

For example, I usually see that visitors who spend a long time online (over half an hour) are much more likely to convert than those who spend only 15 minutes. My very first question is about correlation vs. causation. Does the length of time on the site actually cause the conversion (”Well, I’ve wasted the last half hour on this dumb site, let me just buy what I need and move on”)? If that were true, we would work as hard as we could to keep visitors on our site before they convert, because it would increase the chances of them converting. Or does the conversion cause the long time on the site — it’s hard to make a purchase, or the individual needs to learn a lot before he can push the “submit” button, so the interested visitor ends up spending a long time?

Web analytics show us what, and not “why.” However, about a year ago, I posed this question to Dr. Alan Montgomery, who is a professor of clickstream analysis at CMU. His answer? “It’s a little bit of both.”

Robbin Steif
LunaMetrics

I was so surprised to get on my FeedBurner stats this morning and see a new tab, Live Hits.

Here’s what FeedBurner wrote about it, by way of explanation:

Why does it matter?

Hits don’t always correspond to subscribers, but they’re still important. Especially when they come from a web browser.

Potential subscribers are likely to “hit” your feed in their web browser when they click a “subscribe” link on your web site or discover your feed in a search engine. While these people are not counted as Subscribers (but will be the moment they get hooked on your content and start viewing it with a feed reader), they have nonetheless seen your feed, and we count these exposures as Hits

To really understand what this was, of course, I had to write John Zeratsky at FeedBurner (John, I hope your boss reads this so she knows how incredibly responsive you are…) John explained that anytime anything (a person or a bot) hits your feed, it shows up in a rolling, live fashion (they only show the last 25.) You can use it to be sure your feed is working. You can guess that someone is probably considering subscribing to your feed if the user agent is
Internet Explorer or Firefox Live Bookmarks (or other browsers. Those are the two that I have and could try.) John pointed out that it’s fun, too — you can click on your blog to start the subscription, refresh your other window (the one with your FeedBurner live stats) and see your own hit.

Robbin Steif
LunaMetrics

The challenge of creating personas has been nagging at me for some time, so I jumped at the chance to attend a recent Offermatica webinar, where Forrester research was going to speak about personas. (I would have published this post earlier, but I have been waiting… and waiting… and waiting… to hear from Forrester about permission to reprint their slides here.)

I feel like the problem of persona creation (or as Forrester calls it, Persona-lization) is even harder than we suspect because the personality type can be sliced along so many axes, and those axes change given needs. For example, there is a MarketingSherpa case study on Whirlpool’s creation of personas. Whirlpool breaks out personas based on need – did their washing machine break and they need a new one today , or are they building a new home and considering what kind of washing machine to purchase? And then we can take each of those need-based personas and break them down by personality type (is the individual who is purchasing the machine someone who needs to understand the difference between the various cycles, or are they just someone who wants to feel good about their purchase?)

In the webinar, Forrester started by defining the right number of personas (somewhere between 4 and 7). The pointed out that the first part of the process is to identify similarities across user segments to make the whole process simpler, and they used a financial services model:

Identify similarities:

Step #1: Interview people in the various groups. Their example was to interview individuals of both genders in four different segments: recent college grads, recently divorced, high-net worth and recent retirees. (I kept wondering about the high-net worth recent college grads, or the recently divorced retirees, but fortunately, all those webinars mute your phones and so I didn’t ask. If I had asked, I would probably learn that yes, there are rich recent college grads, but we are looking to create personas, not outliers.)

Step #2: Cluster interviewees along important axes. The axes they chose were amount of money in your account, how comfortable you are with managing money, and how often you log into your account. In their very simplified example, the recent college grads and recently divorced interviewees were all poor, uncomfortable managing their money, and rarely logged in. (Real life is never this clean cut, but it works for a one hour webinar.)

Step #3: Create personas based on attributes that are important for design decisions. So, in the simplified example, they created Joyce Tong, a 28 year old new lawyer with student loans, who is uncomfortable managing her money, who logs into her account frequently. Joyce has three goals: to pay off her loans, to build a little savings and to spent as little time as possible managing her account. The other persona they build was Bill Loftus, a senior manager who has lots of money, is comfortable with money and who logs into his account daily. His goals are to build wealth, enjoy a comfortable retirement and feel in control of his finances.

After creating the personas, Forrester built them out so that the personas felt like real people. Joyce already has a job, and now we give her a picture and narrative around her life, specifically as it relates to money. The narrative emphasizes her pet peeves (”After a tough week at the law firm, the last thing Joyce wanted to do was spend time working on her finances.”) It creates empathy by giving us enough detail to understand her (”Opening her browser, she went to the Favorites menu and took a moment to consider which bookmark led to the account log-in page of her company-sponsored investment site…”) The text works in design issues (”Once in, a look at her account menu baffled her.”) And it calls out the key attributes and goals mentioned above in Step #3. Finally, it includes a quote from Joyce or other involvement device (”I’m making more money but working 70 hours a week!”)

One of the most interesting (and fun) parts of persona-based design is keeping them alive. Forrester didn’t address this issue, but other companies have. Some companies create life-sized cut-outs and put the cut-outs in the room where the designers usually meet. I read that Yahoo! created personas and then created a tri-fold laminated brochure for all of their designers to use. And I often hear about companies where the designers refer to the personas at each step of the way: “Yes, but would that work for Joyce?”

I got on the Offermatica website to see if they are offering the webinar again but couldn’t find out. If everyone got as much out of it as I did, I’m sure they will.

Robbin Steif
LunaMetrics

By now, most e-commerce vendors know that if you require new visitors to sign in before they are allowed to purchase, you decrease your conversion rate. This morning, I experience a new version of this worst practice.

I couldn’t find the black sweatshirt I wanted on my favorite sites, so I started with Shopzilla and found a vendor I had never purchased from before, Apparel4sale. When I started the checkout, they demanded that I register. Well OK, I thought, another vendor who doesn’t know how to read their web analytics, but let’s get this done and over with. So I registered for their site, and then found that they had a PayPal shopping cart, which meant that I had to put in the information all over again. And then I saw that instead of directing the visitors to a “Thank you very much” success page on their own site, Apparel4sale allows PayPal to use a PayPal success page. This means that Apparel can’t measure their conversion rate.

Sure, they can take the orders they get and divide them by total visits or unique visitors (choose your poison), to get an overall conversion rate. But they can’t segment the way I wrote about yesterday — they don’t know which visitors who came from Shopzilla, like me, converted, vs. visitors who found them on a PPC ad, or any other way.

(Sidebar: If you look in the comments from that post yesterday, you will see Ohad Gliksman’s two cents on landing pages. Among other things, he highlights the practice of having specialized landing pages per affiliate — like Shopzilla. )

Now let’s see if this sweatshirt is as bad as the site and the analytics must be…

Robbin Steif
LunaMetrics

When I am confronted with a new website, especially an e-commerce site, my first goal (after I do whatever the customer thinks is the most important) is usually to figure out what drives the conversion rate.

If a conversion rate is 2%, that means, some visits are converting at 10% and some visits are converting at .001%. So I attack the problem by slicing and dicing the data to figure out what a 10% conversion rate “looks like.”

It is a backwards problem. The territory would be easier if the questions were, “We made this change, how did it perform?” or, “We’ve started a new campaign, how did it perform?” With the backwards, slicing and dicing problem, the question changes to, “Something performs well here and something performs poorly — what are those ’somethings?’”

I usually look at a handful of indicators (and always love when I have a powerful package like Omniture to do this with):

Conversion rate by landing page: This is especially helpful when a site is well-optimized for the search engines, because the landing page says a lot about the search terms themselves (and you don’t have to do the analysis for the 800 variations of the search term.) The equation here is orders that started with a visit to a specific page on the site, divided by all visits to that page, all in the same time frame. (And if you are interested in the visits vs. visitors debate, you should read Matt Belkin’s blog.)

Conversion rate by type of visitor (new, returning or loyal): Yes, we know that returning visitors should have a higher conversion rate than new visitors, but how much higher? Today I started working with a new site that had tiny, tiny new customer conversion rates, and returning conversion rates that any catalog company would be proud of. This becomes very interesting when you learn that over 1/4 of the site’s visitors are returning customers — so you’ve got all those new visitors who almost never buy but when they do, they suddenly become really valuable. Equation: New customer orders/Visits from all new visitors.

Conversion rate by organic search engine. You might think this is ho-hum, but all sites are not created equal and all search engines aren’t either. It doesn’t necessarily follow that if organic search results from Yahoo! convert better than those from Google, I should spend pay per click money on Yahoo! but I’m sure going to start there. (I add this last comment because I want to remind everyone, it’s not worth running numbers if you aren’t going to do anything with them. )

I wrote a white paper on this topic some time ago, you can get it here.

Robbin Steif
LunaMetrics

In March, I wrote about Google Sitemaps, and pointed out that you can learn from the analytics that they give you. (Plus, if you have client side analytics, you don’t know when the crawlers come to visit, since client-side analytics don’t talk to the automated traffic. At the very least, you can learn when the Googlebot last visited from Sitemaps.)

As part of the Sitemap thing, it is nice to include a Sitemap so Google will do a better job of crawling your site. The downside is, Google’s xml conversion tools are not very user friendly. I made the mistake of sending the techies at a customer a note yesterday, “It’s easy! Just use Google’s little tools,” to which they replied, “If it’s so easy, you do it.”

So I found this tool, xml-sitemaps.com, that makes creating a sitemap in xml actually easy. You, too, can look like you know what you are doing in front of your boss and your customers when you use this free tool (or you can pay the $14.99 and get added functionality.)

Robbin Steif
LunaMetrics

This just in from Avinash Kaushik, the web analyst from Intuit who calls them like he sees them:

Just wanted to drop you a note that my level of blog envy finally reached a critical mass and I finally have a blog of my own .

Robbin Steif
LunaMetrics

I see, or get requests for benchmark data all the time. What’s a good conversion rate? What’s a good clickthrough rate? Mostly, I want to tell people not to covet their neighbor’s rates, because the neighbor is in a different business.

Having said that (and heard that, over and over again from other conversion professionals), I want to point out that there is a time and place for benchmarking:

• When you are starting a new business and need the numbers for your business plan and your investors
• When you are rewriting your resume
• When you are asking for a raise

There are probably many more, these just come to mind quickly.

Robbin Steif
LunaMetrics