Teorem's Blog

Jay Janssen, who you have maybe seen on the pictures of the Mysql Conference started a blog named MySQL Guy some days ago.

I think you will find valuable information in his posts, for instance regarding mysql clusters.

It seems to be the thesis developped by Brian Smith in this post: Why I'm at Affiliate Summit 2007.

The conversion rate is obviously very important for merchants, it's the true indicator of the effectivness of the money spent in online avertising and define clearly the ROI (return on investment). And the revenue shared model ensures the merchants that what they are paying is really providing them more earnings.

From my point of view, the main problem in such a model is the trust you can have in the partners.

With the CPC model, it's quite easy for both, the merchant and the shopping engines, to record the traffic sent and received, and then have analysis done and resolve conflicts.

With a shared revenue model, the traffic sender is more or less at the mercy of the guy receiving the traffic. So, when we are talking about the affiliates of Google AdSense or Yahoo Search Marketing, it's not really a problem: the big guys (G and Y) won't cheat on the small guys (the affiliates). On the contrary, they have to put in place means to avoid the small guys cheating on them by sending fake traffic.

But putting in place means to verify that revenue has to be shared because of the traffic sent to a "small guy" would be difficult if not impossible because once the user is on the website, the small guy has complete control of what the user browser can do: for instance, it can avoid to put in place the beacons allowing mechanisms like tradedoubler to record that a sell did take place.

Thus, the intermediary (the shopping engine) must trust the traffic receiver (the merchant) with regards to the numbers of sells, thus on how much the merchant must pay.

I'm wondering how well the investors have trust in business models based on the faithfulness of human beings in general.

You surely heard about the VC firm called Sequioa, only because of their last success: selling YouTube to Google.

One of their partner is explaining what's make a great success of a start-up in an article in Forbes.com.

So, it seems that I have to find an idea answering the needs of customers in a small market that will be big in a few months/years to be successful ... again :-).

If I take a look back at what we did at Kelkoo ... he is right.

After seeing some advertising in London some months ago that really surprised me (limited download per months, high prices) with regards to french prices and offers, I wanted to check if it still the case and where is the best living place for Internet addicts (like me).

Let's summarize the french situation: all providers have ADSL2 (look for ADSL Standards on wikipedia) offers that you can really use when you are sufficiently near the DSLAM. Outside France Telecom (now selling under the brand Orange), all offers are around 30€, including access to television, unlimited phone and unlimited internet. Most of the time, you will have to pay more only if you want access to Canal Sat or TPS, that are providing more Television channels. Orange (because of the regulation) provides the same offer for around 34€.

I tried to found a comparable offer in the UK, and I found only Orange (Unlimited Internet for around 20 pounds (35€) and only at a 8Mb/s bandwidth (no adsl2 possible). Tiscali is asking you 22 pounds for the same service. And the similar offer from BT is 27 pounds with download limitation.

You will find more information on this ADSL Guide. With two providers having ADSL2 offers BeThere and UkOnline (I don't want to discuss the price).

As for Germany, my german language skills being not that good, I had some difficulties to find something interesting. T-Com, the subsidiary of the historical phone company provides an offer at 45€. On the other hand, you have 1und1 providing an offer very similar to the french ones.

It seems that my first idea was right: I prefer to stay in France.

The current issue of BusinessWeek (October 2006) has a cover story about "Click Fraud" that will surely put more pressure on the major Search Engines, Google first and naturally Yahoo.

If you don't know what is click fraud, please take a look at what Wikipedia is saying about it. They already have the link to the businessweek article.

There is not much more in that BW article than what was described in the last january issue of Wired. Stories showing the problem for small to medium companies, description of the future threats and confirmation that we (Yahoo) take the problem very seriously and invest to implement solutions.

There is at least two interesting information item anyway. First, the announce that most of the click fraud comes from affiliates and not our network, and that a very good percentage is coming from "parking web sites". Those are web sites hosting thousands of domain names their owner don't want to use for the moment and making them serve ads from Yahoo or Google as a way to benefit from the free hosting. Second, the importance of groups of people clicking on ads for payment, those people being a lot more difficult to discover than an automatic fraud click system.

It is natural, like any other successful new business, that the CPC model haves its predators profiting from the system without providing any value to anyone outside themselves. We will do whatever we can to stop them, but I hope anyone will remember how many wonderful and useful services the CPC system is funding before saying that it should be stopped because of fraudulent clicks.

Pascal Borghino, working in my team, is posting about Modeling, UML, databases and the apropriate tools to use. He is mostly ranting that MagicDraw UML is not providing him the extended functions he needs in a Database Modeling tool.

While I understand his frustration, I think he is mispelling the problem: MagicDraw is not a DB Development Tool !

For the realization of a software or a db scheme, there is always a moment when you have to stop modeling and start developing the stuff you want to deliver. At that moment, you also change the tool you are using in order to have the appropriate features for the development activity. While working with Java, I will stop using MagicDraw and start using Eclipse JDT. For a DB Scheme, you will start to use the development tools provided by the database vendor or any other third party (preferably open-source, naturally). Usually, those tools have bridges allowing to translate a UML model into the corresponding elements in the target environment.

To know when to stop modeling and start developing is difficult. It's common to see software developers stoping too soon and then having a model without sufficient information to understand their software. The opposite trap is also common : stopping modeling too late and then having a model so complex that it's sometimes easier to read the code directly. To avoid that second trap, changing the tool you are using in your work is then an advantage because it ensures that you will update the model with only the necessary information. Moreover, it ensures that you will use a tool that is perfectly suited to the task in each step of the realization.

I think that when you reach the point where the target database becomes a very important question to continue your work, you already are past that moment.