Coffee, Sun & Technology

For quite some time now I’ve been using analytical tools to slice and dice user data on “small sites” (Wambo gets about 50K visits a month roughly, and I have 2 other small B2C service sites that I’m managing on the side). In doing so, I’ve realized that the kinds of metrics you look at for these sites that need to grow fast are very different from traditional analytics.

While traditional sites look at “improving” existing business processes (increasing conversion rates, enhancing the customer experience, etc etc ), startup sites are laser-focused on finding the right formula for their site or service. Startups are always building. And speed is what matters.

For us early stage sites, we look at a completely different set of numbers. And, we compare data ALL the time. Month-over-month, week-over-week, even sometimes, hour-over-hour. So I’m thinking of another kind of analytics that’d be useful for me, let’s call that growth analytics. In an ideal interface, I’d like everything presented in the context of velocity. Velocity been the uber measure, similar to a session or a page view in traditional web analytics.

I’d want to see user acquisition velocity by hour, day, week and month. I’d like to be able to compare velocities for different time ranges (this week vs last week). I want to be able to track acquisition velocity  for different segments. I want to A/B test my site and see what the impact is on the velocity metrics. And I want to project in the future what my KPIs will look like if I can sustain the current velocity levels (i.e. if I keep growing my users by 3% a week, that will get me to the 1Million user mark by ___). A new calendar type but with dates in the future too, not just in the past.

Yes, you see where I’m going now. I think what’s a little broken with the state of analytics today is the fact that we spend 90% of our time trying to answer the “what happened” question. And that’s soooo yesterday

These days I’m running into more and more eMarketing folks with engineering backgrounds. I think this happens for a couple of reasons. First, e-business is about efficiency. Online you can track the effectiveness of your campaigns, promotions, changes in site design. Something you can’t do easily in the offline world. So it’s not too surprising to find people who can crunch numbers taking key roles in companies - even in the Marketing departments. Second, we live in a Google world. Google has become such a powerful force in e-marketing that it was able to push its spread its analytical/math based culture on to their partners and customers. It’s all about results as they say.

Coming from an analytical background, I certainly am happy to see this happening. As long as analytical folks realize they still need to work hand-in-hand with their more creative counterparts, this is a promising evolution.

Here’s a bonus… eBags.com first sketch, dated 1998, from Jon Nordmark and Peter Cobb.

I thought I’d expand a little on the process of innovation, in connection with my post on patents. We all look for ideas to do better, at all levels – reaching a higher conversion rate, releasing a new killer feature for our software product, etc. I’ve learned a few lessons as an entrepreneur, and maybe this could be useful to the blog’s audience.

First and foremost, the most important thing for creating innovation is to eradicate any cynicism in your organization. When I was a student at Stanford, I took a Computer Architecture class – it was one of my first classes after I arrived from Europe. The professor started his class outlining some of the pre-requisites/etc, then made a point about most ideas being bad ideas. I still remember the PPT slide with the words on it “There are lots of ideas out there, most of them are bad”. Well, frankly if you know that you are going to get shot down and called stupid everytime you have an idea the most natural reaction is to not talk about ideas.

Second, and related, be mindful of your own reactions when people come share their ideas with you. Sharing an idea, particularly with the boss, etc, is an act of courage. I know for example that building on someone’s idea, whether the idea is eventually dismissed or validated in its original form is the most effective encouraging and rewarding message you can send.

Ideas take time to mature. I’ll give you one example: at Fireclick we quickly figured our customers needed help interpreting their analytical reports. One thing that was missing was some kind of benchmarking, where customer X could compare its conversion rates for instance to other customers in the same vertical. Originally we thought we could have a best practices group, or organize online Webinars to share some of these stats, or even formally introduce customers to each other so they could discuss. None of these worked, until we actually decided to create the Fireclick Index, a benchmarking service for online businesses. That was the answer, which took us 18 months to figure out (we released it in 2003).

Ideas [typically] don’t come from brainstorm, solutions do. That’s just observation, and I could be wrong – I’ve just noticed that brainstorms are good for finding solutions to a *specific* problem. In the early days at Fireclick, we institutionalized a 2H brainstorm every Thursday at 7AM, horrible time of the day for this kind of thing (the intent was to avoid any disruption). In hindsight, these meetings were good because they help set the right corporate environment for innovation, but most of them would typically end up on a frustrating note with no decisions made. By contrast, whenever we have big problem to solve and organize informal brainstorm meetings we find the solution and everyone feels good.

Customer ideas are typically higher quality. We all know this – but it’s so true. Maybe because customers actually point you to a problem you can creatively solve, which is easier.

Communication. This is where I personally struggle the most, partly because I’m a very impatient person and I never take the time to fully “package” my ideas before I share them. In smaller structures, we typically do not have the time to fully research ideas and find the right words to describe it. But it’s an essential step. Make your idea attractive and people will love to discuss it.

Last but not least, take action. It’s ok if you fail, because you’ll have learned something along the way and your next idea will be better. Personally, I’ve failed a lot, with some success along the way that makes it all worth it.

More on feeds While I have fewer and fewer direct visits to the site, the feed circulation has been steadily increasing since mid-November.

These are 7-day moving averages

We’ve discussed earlier the hypothesis of a 10 to 12 day offset between direct traffic to the site and feed circulation. But looking at this chart I’m prompted to make another hypothesis: the reason why my direct traffic is down might be because visitors to the site subscribe to the feed and don’t come back to the site directly.

I think I’m just harvesting my November traffic generation efforts - and slowly depleting my pool of prospect subscribers. What’s interesting in all of this is the Google Adsense Click-through rates, which have increased as well. I have no explanation for this yet - even the number of clicks has gone up.

[Read other posts for this series: part 1 part 2 part 3 part 4 part 5 part 6 part 7 part 8 part 9 part 10 part 11]

This blog is about… Web Analytics. I haven’t talked much about the Google Adsense program results yet, because I’ve been waiting to get quality data before reaching any conclusion. I’ve recently added some text links, right under the post title. These create revenue when someone clicks on then and clicks on one of the links on the landing page. I will share some results later, but 6 weeks into this Web analytics project I am a little frustrated with Google Adsense. Couple of reasons:

First, I have no control whatsoever on the ads. I can’t even pick the theme. I noticed that to keep the ads relevant, I have to mention “Web analytics” or name a vendor in the latest post otherwise some crazy ad will be served. (Siebel, Bacardi, Coach…). Clearly silly, why not have some kind of API where we pass to Google Adsense a keyword or two for targeting purposes?

Second, horrible native reporting capabilities. At the very least, give me some basic idea of who’s clicking on my ads. Referring domains and Entry page are two great starting points. This should be a walk in the park for the Google Analytics guys and provide incredible value to publishers.

Now in spite of all, Google Adsense has been working OK. I technically can’t talk about results, but to give you an idea, it pays for my Starbucks every morning (not the venti capuccino yet, just a tall coffee drip…)

[Read other posts for this series: part 1 part 2 part 3 part 4 part 5 part 6 part 7 part 8 part 9 part 10]

Whoever wrote this Internet Retailer article got a little carried away with the title: Net-A-Porter gets 20-fold boost in sales through “Tell Santa” e-mails. The title seems to suggest the “Tell Santa” feature (see below) multiplied Net-A-Porter sales 20 times:

Reading further, it says that users who got to the site via the email are 20-times more likely to buy the article than someone just visiting the site (at least that’s what I understand from it). I can see that - but it’s a little disappointing especially after reading the catchy title, because the overall effectiveness of the “Tell Santa” feature depends on 2 additional factors not mentionned here:

- Percentage of overall site visitors who click, fill out the “Tell Santa” link and send it
- Click-through rate on these emails

Obviously if we’re talking about 100 email visitors/day and 1,000,000 visits/day the impact isn’t as great as suggested in the article. This article would be a lot more interesting if it gave a little more detail for what seems to be an interesting case study. Maybe the Net-A-Porter team can comment on this.

Playing with feeds. Eric Butler’s post on Feedburner prompted me to explore some ideas with feeds. Feed readers are my #1 segment and I haven’t paid too much attention to them. I plead guilty.

Here are a few things I did:

First, I consolidated all my feeds to be delivered using Feedburner. The goal is to get better Feedburner stats, and gain access to their suite of tools. Since I use Wordpress, I just downloaded this cool plug-in from Steve Smith which does this very easily.

Second, I turned on Google Adsense for feeds in Feedburner. It’s really simple to do in the Feedburner interface, but you have to apply for this with Google (they’re in beta), and after a few forms and emails, I didn’t get selected so no Google Ads in my feeds. I don’t have time to fight this right now so I’ll just pass.

Third, I turned on a few of the FeedFlare features which essentially allow me to insert useful links into my feeds such as “Email the author”, or “Add to my de.icio.us” etc. Cool!

In this process I have broken my feed links at least once or twice (Eric had the same issues I think), but it seems worthwhile. If you have a Feedburner account I encourage you to check FeedFlare out. I’ll be watching for results and publish them to this blog.

[Read other posts for this series: part 1 part 2 part 3 part 4 part 5 part 6 part 7 part 8 part 9]

Segmenting my reader base. Well it took me about 4 weeks to figure out that my readers can be divided in three key segments: (1) Feed readers, (2) Loyal site visitors, and (3) Casual site visitors.

Feed readers are the most loyal by far. Once they subscribe to the feed, they tend not to unsubscribe (as demonstrated by the data Bloglines and Feedburner provide me). Loyal visitors directly access the site using the URL www.coffeesunanalytics.com, or, they go to Google looking for “Coffee Sun Analytics” or “Xavier Casanova” - i.e. they know what they’re looking for. Casual visitors on the other hand are typically one-time users, they find my site from a referral site or a search engine.

I’m still unsure about what’s the appropriate mix of (1) vs (2) vs (3). Currently, I’m at approximately 125-150 feed readers (total for all feeds), 200 loyal site visitors/week and about 200 casual visitors per week. Roughly 30/35/35%. Clearly the goal here is to convert the casual visitor into either a loyal site visitor or a feed reader - and I further segmented the new visitors I acquired over the last 4 weeks by referring domain.

Playing with data a little I’m learning two new lessons. The first one confirms the importance of the referral source as a key conversion driver: according to my segmented reports, first time users which were referred to me by other Web analytics bloggers (for example Eric Butler and Eric Peterson ) do come back - whereas first time users who got to my site from Google or another random referral aren’t coming back. In other words, the (3) to (2) conversion rate greatly varies depending on the referral source.

Second, I do not believe posting everyday is a productive use of my time, for this blog. In weeks when I’ve posted 7 or 8 posts, I have received tons of natural search traffic - actually I even got “digged” . This effect tends to last for 10 days or so, meaning that if you spot posting for a while you still get some inertia going but it dramatically slows down eventually. I (of course) will not complain getting 30 or 50 Google Search users on a given day, except that these never come back.

So unless you are a professional blogger/doing this for money, don’t waste your time over-blogging.

[Read other posts for this series: part 1 part 2 part 3 part 4 part 5 part 6 part 7 part 8]

Interesting fact about RSS feed users. What happened to my traffic? The last week or two have been kind of depressing as far as visitor activity on my site. Unpredictable patterns, some days strong, other days slow, and little correlation between how often I’m posting and how many visitors I’m getting.

One thing did surprise me however. I use Feedburner for formating and delivering my RSS feeds, and watch my feed circulation stats almost daily. And quite surprisingly, these numbers have kept going up even though I have posted much the last few days and my direct site traffic has been cut in half.

Running Fireclick session numbers vs Feedburner circulation stats for the last 8 months, I noticed a very interesting phenomenon which happened in September. My site has been down for 3 straight weeks (I was in Europe, unable to restart my router in my basement) - which of course brought the direct session count to the site down to 0. The feed circulation however kept going for a few days, took a dip, then started going back up.

Taking a closer look at the stats I think the lag between the direct site traffic and the feed circulation is about 12 days. (Red line represents the Feedburner stats and the blue one the Fireclick session stats - day by day since April 3rd ‘05).

The other interesting property of RSS feed users is loyalty, which I’m sure will be of interest to the Web 2.0 community. I don’t have enough data yet to prove this, but if you look at the chart above the red line does exhibit a slow start then a consistent uptrend - while the blue line is a lot more erratic. The vertical scale by the way is normalized according to an overall average.

[Read other posts for this series: part 1 part 2 part 3 part 4 part 5 part 6 part 7]

Gus Kormeier from Altrec pointed me to the Stanford Guidelines for Web Credibility, an interesting page recapitulating some of the key ingredients for a credible Web site - with references to research papers to illustrate/prove the recommendations.