where i’m at…

people search by

make your own Wink Widget!

Websites and virtual communities suffer from the digital equivalent.  Spam, Sock Puppets, Bot Attack, Slogs and Advertorial mix with aggressive affiliate programs, ads for debt consolidation, pay day lenders, porn, casino’s and pharmaceuticals without a prescription –   everything you encounter online that makes the web feel like a bad neighborhood have been given a name along with an interesting new site that wants to create a discussion about how to combat it- Virtual Blight .

Like it’s real world counterpart, blight left unchecked it will chase away the  inhabitants of a community and destroy the value of the site and the brand.

Modern applications are complex systems comprised of the application software, web servers and databases, operating systems and drivers, numerous servers, storage for application data, switches, firewall, and the networks that connect it all. Deploying and operating these systems requires significant expertise and scaling complex systems is a not-trivial problem. Applogic is trying to address these issues:

Applogic starts with grid operating system that puts together standard servers to creates a single supercomputer.  A grid can be built from commodity servers and may contain anywhere from 1 to 5,000 machines (so far, they claim they don’t know the effective upper limit).

3tera recently completed Grid Computing Benchmark Test along with dedicated server hosting provider Layered Technologies. The benchmark measurements show that Applogic really delivers performance.

An overall performance benchmark of 42,540 was achieved with the UnixBench WHT utility while consuming 443 CPUs. Based on the single CPU server result for the same benchmark of approximately 100, this implies that users can harness the equivalent of 420 single CPU dedicated servers through the utility computing service. This is no way represents the maximum performance of the system. During the benchmark we did not reach any fundamental architectural limit of the system, but ran out of hardware resources to add to the test.

Applogic makes very clever use of software virtualization, a process that can make almost any piece of Linux software or infrastructure hardware (Switches, Firewalls, Storage Arrays) and replicate them as a virtual appliance. A visual mapping tool allows the user to model complex IT application systems and transform them virtualized infrastructure — completely self-contained solutions that be propagated from one datacenter to another or seamlessly scale from a fraction of a server thousands of servers without reconfiguration.

We have yet to measure this Grid solution under real world tests, but the technology seems very promising

• Car salesmen lie to you on the phone to get you to walk in the door.
• Car dealers advertise a special in the newspaper and then tell you the vehicle is already sold.
• They tell you that the discount they advertised isn’t available in the model you want because “that car is hard to get.”
• They claim to have your car in stock and then try to sell you whatever they’ve got.

Most of us would rather get a root canal than walk into a showroom.

So, what does a dealer do when you catch the salesman red-handed in a bald-faced lie?
Apologize? Show Contrition? Pretend to care?

Berkeley Toyota simply refused to sell my wife a car.

What does this have to do with measurement? Call it the Stein Uncertainty Principle: what you choose to measure changes behavior with unintended consequences.

Toyota has a great reputation for customer satisfaction. They methodically follow up with every customer and have built much of their compensation for salesmen from the customer survey results. Salesmen know that one unhappy customer could affect their rating and cost them thousands of dollars. Does this system make salesmen treat customers better?

NO. Instead, this dealer won’t sell you a car if they think you might be unhappy with them. Catch them lying to you and they won’t sell you the car. Why risk letting anyone with a bad experience actually tell Toyota Corporate about it?

This saga begins when my wife decided that her 1992 Camry was ready for a replacement. The car had served her well for 15 years and over 100,000 miles, but time had taken its toll on the interior and exterior.

We began by visiting Broadway Toyota in Oakland, where I had leased a Prius the month before. We chose the model and color she wanted, but they did not have the car in stock. The Oakland dealer did search and told us that the touring package was only available in a Package 5, which contained the navigation system and was about $3k more. They offered to sell us a package 5 Prius for $1750 off of the sticker price, but said that they couldn’t find any in gray. We asked them to locate the car we wanted and give us a call.

After not hearing from them for three day, my wife decided to call Berkeley Toyota, where she bought the Camry in 1993. She spoke to a salesman named Javier and asked him if they had a gray Prius in stock. He said they had two available and made an appointment for us to come in.

My wife and I arrive 15 minutes later to discover that Javier hasn’t returned from picking up the car. When he finally arrives after 20 minutes, I explain that we were in a hurry; I took off work to come make sure my wife gets the deal we were promised and she needed to leave for class in two hours.

Javier stepped away for a few minutes to talk with another salesman and returned to tell us that it turns out that they sold one of the gray ones on Saturday and that the other one they have isn’t available because it just arrived and hasn’t been checked into inventory. He said that the car can’t be ready for delivery for another three days. When I questioned why it would take so long, he responded that the service department was very busy (this PDI process, according to the customer service manager at Berkeley Toyota, takes about 1.5 hours and can usually be done the same day).

Javier sat us down and asked my wife to fill out a credit application along with an offer sheet. He took the sheet and went into his manager’s office. Javier returned about five minutes later and told us Mr. Rios said his manager noticed that we “were upset or in a bad mood,” and wanted to be sure we would be happy. He then admitted that the gray car been damaged and needed repairs before they could sell it to us. After hearing we were not interested in buying a brand new car that had already been in the body shop, he tried to sell us on a different car.

Of course, we were not happy about coming in to the store to discover the car we wanted wasn’t available. We told Javier exactly what we wanted on the phone, we had already researched the car and we came in to close the deal. We didn’t come, not re-negotiate the deal or be offered a different color. They didn’t have the car they promised us available, so we left.

My wife phoned Mr. Rios a few hours later to ask what the manager might have meant by that comment about us being in a bad mood. She explained that at this point she was interested in ANY Prius with a Package #2 and a dark interior, and she was willing to come back in to discuss another color. Javier said that he would look into what was available and call her back.

Imagine her shock when Javier phoned back a few minutes later (presumably after discussing it with his manager) to say that he “feels uncomfortable selling us a car” since we were so angry and that her husband made him “feel like dirt” on the phone.

We were outraged. They lied to us about having two cars available. They lied about needing three days to prepare the car for delivery so they could fix the body damage. After all that, when my wife was ready to settle for another color to be finished with the agony of dealing with car salesmen, they said we weren’t happy enough to buy a car from them!

Toyota has gotten a lot of press about their customer satisfaction program, including a recent New York Times article, Toyota’s rise to world domination driven by customer satisfaction, we decided to write Toyota Headquarters.

We did hear back from Toyota, who said they would file a report with Toyota of Berkeley and that we would get a call back from them within 3 days. They did not offer any hope for resolution or provide anything we could feel positive about.

The next morning, we got a follow up call from Toyota of Berkeley. The customer satisfaction manager explained that he had heard about the situation and that the dealer was simply exercising their right to choose who to do business with. He agreed that Javier was trying to avoid us knowing about the car needing paint when he said it would take 3 days to PDI. He further stated that Toyota would require them to tell us about the body damage if the repair exceeded $500. He didn’t state that the damage was beyond the $500 threshold, but we speculate that the dealer told us about the body damage after they new thye would have to disclose it when they sold us the car.

He also said that this situation occurred because the salesman was afraid that if he sold us a car, we would respond unfavorably in our customer satisfaction survey. He said that one bad experience reported by a customer could lower the customer satisfaction index, costing the salesman thousands of dollar over the course of a year. Javier dug himself a hole and then claimed that we were in a bad mood and using profane language because he was scared we might trash him in a customer survey.

How is that for Irony? Because of Toyota’s customer satisfaction measurement system, Berkeley Toyota lied to us, got caught lying and then their management refused to sell us a car.

Unfortunately, that won’t ever show up in a customer survey.

Nielsen’s intent is laudable. He contends that the vast majority of users are consumers of content instead of creators and is urging site operators to make participation easier.

User participation often more or less follows a 90-9-1 rule:
* 90% of users are lurkers (i.e., read or observe, but don’t contribute).
* 9% of users contribute from time to time, but other priorities dominate their time.
* 1% of users participate a lot and account for most contributions: it can seem as if they don’t have lives because they often post just minutes after whatever event they’re commenting on occurs.

Nielsen goes afoul when he talks about blogs, “There are about 1.1 billion Internet users, yet only 55 million users (5%) have weblogs according to Technorati. Worse, there are only 1.6 million postings per day; because some people post multiple times per day, only 0.1% of users post daily.”

“There are about 1.1 billion Internet users, yet only 55 million users (5%) have weblogs according to Technorati. This analysis has three fundamental flaws.

1. Nielsen takes the broadest measurement of internet users and the narrowest definition of a blog, posts using one of the popular blog publishing platforms such as WordPress that happen to report postings to Technorati. The total number of internet users is not the same as the number of blog readers. To assume that even half of that 1.1 billion “users” would recognize a blog as being a blog defies credibility so it makes no sense to count them when measuring participation. Does someone who has never been to a site count as a lurker or passive participant?

2. Comparing measurements of different values from two different systems is unreliable. It is perfectly reasonable to compare the measurement of the same value from two different measurement systems to compare the measurement approach. You can’t do analysis with measurements tfrom two different system without normalizing the data.

• Technorati is an American-centric company that doesn’t claim to have an even global penetration.
• Blogging is a new phenomenon that has only been in the public eye since 2003. According to Technorati sources, the number of blogs doubles every six months. The August figure of 55 million will be around 110 million by February of 2007. It is approaching 70 or 75 million as of the middle of October.

3. The biggest problem is the definition of “blog”? Blogging is a term that embraces much more than the enabling technology. It isn’t about posting your opinion in Word Press; it is about adding your thoughts, opinions and analysis into the public discussion. Nielsen’s AlertBox, for example, is an opt-in electronic newsletter that is also posted to his site. The content he writes is widely discussed in blogs and gets its share of links from the blogosphere. The only reason AlertBox is not a “blog” is that the technology he uses to publish. It shouldn’t surprise anyone that mixing a broad definition of users and a limited definition of participants produces lopsided results.

Worse, there are only 1.6 million postings per day; because some people post multiple times per day, only 0.1% of users post daily.” Nielsen disappoints again on this point in three ways.

4. Nielsen defines daily participation as the threshold for determining whether someone is a regular contributor to the discussion. His definition stipulates that quantity of posts is the only thing worth analysis. Take Alertbox as an example again. Since 1995, Nielsen has published his column roughly twice a week or about 260 times over 11 years. By his definition of participation, one of the most widely read column on User Interface and Design, written by one of the leading experts on the topic, counts as an occasional contributor.

5. He ignores comments as a form of participation in the blogosphere. Most blogs have at least a few comments and many have hundreds or even thousands. It is hard to justify not including these as participation.

6. The last issue is how he crunches the numbers. Technorati’s measurements are “ancient” as judged by the historical rate of change. Still, if we accept 55 million blogs and 1.6 million posts/day, 2.9% of bloggers post EVERY DAY. If we use a more reasonable threshold like posting once a week or even once a month, participation in the blogosphere may be as high as 10 or 15%.

The final issue in this 90-9-1 analysis is that users are in multiple communities. A user who is a devoted participant in the Amazon community for example, one who has read thousands of books and posted reviews for each, can hardly be expected to be a daily blogger as well, but do you count him as a lurker?

You can prove almost any point with statistics and measurements tailored to your definition. If you want data for reliable decision making, you need to be more thoughtful in your approach.

ROI metrics and the sense of accountability they provide are addictive. Decision making is reduced to how much you are willing to spend for every dollar or customer you get back. It appears to take the risk out of marketing investments. Dave is absolutely correct…as far as he takes it.

Marketers who believe they can accurately quantify the impact of their online advertising, whether SEM or Brand, do so at there own risk. Every analytic system I have ever used is flawed; The assumptions and methodologies inherent in each approach creates measurement error. Managing campaigns on a strict ROI basis demands that you have at least two analytic systems and a detailed understanding of how each works and what they actually measure.

Determining ROI for search marketing requires you to understand user behavior. One very common issue to watch out for is how you treat keyword search and a brand search. Marketers will frequently separate ROI for their brand terms from keyword search. If a customer converts after searching for “red widgets,” that ROI is credited to keyword search for the term red widgets. If a customer converts after searching for “brandxxx widgets,” the ROI is credited to a brand search.

The problem is that customers may visit your site during the interest phase, the research phase and the purchase phase. Most analytic systems are not configured to track multiple visits through CPC and/or SEO channels, so the return doesn’t appear to make the investment.

• Customers will find you searching for keywords and then return searching for your brand or some variation/misspelling of the brand.
• Customers will find you on one computer and return via direct navigation or brand search from another computer to purchase.
• Customers will enter the URL’s in the search box instead of the address bar (perhaps as high as 15% of users) so direct navigation shows as brand search.
• Brand searches are frequently latent conversions from keyword searches.
• Depending on the cookie setting of your analytics system, you may or may not preserve each customer touch source. You also may not credit the touch with the conversion.
• Referrer data isn’t always preserved through caches and browsers. Firefox users on MSN, for example, will show up as direct searches instead of tagged with a natural search keyword.

I just finished a three month contract for a startup. Despite deploying sophisticated, redundant analytic systems (Google Analytics and ClickShift’s Statistical Bid Management, only 60% of the orders in the first three months were tracked and many were reported as direct navigation or searches for the brand.

You might expect this for a mature brand with a large repeat customer base, but it defies logic for a startup that was still only using CPC for marketing. Since we had a small data set, I was able to research the orders individually and attribute the source and term for each record in the customer table.

You wouldn’t want to try to repeat that method with 10,000 orders, but the result was a 160% increase in the reported ROI for the CPC campaign. Individually or combined, the analytics systems didn’t produce accurate enough data for decision making.

The only way to really understand ROI from each channel and search term is to find ways to induce customers to login as quickly as possible, while the referral data is as fresh and accurate as possible. Incorporate that referral/source data directly into the Customer table and import all sales information into internal systems to produce the ROI measurement.

If you do not have an initial source associated with a customer record, make it a goal in every customer interaction (survey, customer service call, etc.) to obtain that information. This allows you to accurately attribute revenue to the marketing investment and track every additional touch point that generates a visit regardless of source, medium or computer. With that kind of data on hand, you have a baseline to begin to understand the value of each advertising channel. Then your ROI based decisions can be good ones.