Daniel Buca

A book for a better online commerce

Daniel Buca — Sun, 06 Feb 2011 21:06:43 +0000

Actually there are 4 books, not only one, but the title sounds better like this

If you manage an online shop or you’re building software for online shops then these books are for you

1. Introduction to Machine Learning, Second Edition (Adaptive Computation and Machine Learning)
Link: http://www.amazon.com/Introduction-Machine-Learning-Adaptive-Computation/dp/026201243X/ref=dp_ob_title_bk
Price (Amazon): 37.50 $

2. Pattern Recognition and Machine Learning (Information Science and Statistics)
Link: http://www.amazon.com/Pattern-Recognition-Learning-Information-Statistics/dp/0387310738/ref=pd_sim_b_2
Price (Amazon): 66.25 $

3. The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Second Edition (Springer Series in Statistics)
Link: http://www.amazon.com/Elements-Statistical-Learning-Prediction-Statistics/dp/0387848576/ref=pd_sim_b_5
Price (Amazon): 61.32 $

4. Data Mining, Third Edition: Practical Machine Learning Tools and Techniques (The Morgan Kaufmann Series in Data Management Systems)
Link: http://www.amazon.com/Data-Mining-Third-Techniques-Management/dp/0123748569/ref=dp_ob_title_bk
Price (Amazon): 44.07 $

By reading one of these books (or all) you will get familiar with the subject “machine learning”.

How will “machine learning” help you and your business?
You will start thinking about:
- a strategy for better recommendations to your customers
- a strategy for better understanding of your customer’s needs and wants
- a strategy on how to get more from the data you already have

This would only be the start for a better eCommerce experience for everybody.

If You Liked It Then Bookmark It

Hide Sites

What we put on the first page of an online store?

Daniel Buca — Mon, 17 May 2010 20:32:17 +0000

If you ever designed / developed an online store I’m sure that you put asked yourself this question at least 2-3 times.
Nowadays there are so many opinions about the first that you do not know for sure which is the best solution so I thought to ask this question here.

Home page of an online store

Option 1: we show the products we want to promote to the clients or new products that bring us the best proffits or products not sold or even products that belong to Supplier X (we just made a deal with that supplier).
This option starts with the idea that we push to the home page products that have a particular value to us.

Option 2: we put the most attractive products, products most viewed or most ordered.
This version is based on the idea of best sellers, we promote what we sell better / what people are viewing.

Option 3: we analyze which categories / products the user visited before or which products the client ordered in a prior visit to our shop and we make a selection of products based on these criteria.
In this option we take into account the interests of the user.

But the question remains: what do we put on the first page of an online store?

Wich solution is correct?

Unfortunately, none of the three options is correct, as neither of the three options is wrong. Options 1 and 2 are based on a classic sales model wish, in some cases, can give good results and option 3 goes on a behavioral shopping model that can work for a while.

I, personally, would go on a mix between those three plus a flavor of seasonality. How about you?

If You Liked It Then Bookmark It

Hide Sites

RunClassifieds.com has a new owner

Daniel Buca — Thu, 11 Feb 2010 17:30:11 +0000

Since yesterday RunClassifieds.com has a new owner.
The has has grown constantly in the month that passed since launch and that confirms me, once again, that white hat seo work very good and there’s no need for spam or other illegal stuff to launch a new site.

Let’s wish the new owner good luck and many visitors!

If You Liked It Then Bookmark It

Hide Sites

News from RunClassifieds.com

Daniel Buca — Thu, 28 Jan 2010 19:46:32 +0000

I’ve been working hard on RunClassifieds.com in the last days and I had no time to update my blog but I have some news for you.

Now you can view the classifieds ads published on RunClassifieds.com in a new style, browse by images .
You can view the new functionality here.

I have something else too that might interest you.
I wrote on the RunClassifieds.com Blog an article about how people browse classifieds websites and about how you can make a classified ad that can help you promote efficiently and free your business.
Here it is: How to write a free classified ad that really works for your business

If You Liked It Then Bookmark It

Hide Sites

RunClassifieds.com reviewed by KillerStartups.com

Daniel Buca — Fri, 08 Jan 2010 19:08:13 +0000

RunClassifieds.com has just been reviewed by KillerStartups.com . You can find the review here.

To give just a small quote from the article: “individuals couldn’t really ask for a lot more”.

The review from KillerStartups.com makes me believe that I made a good job with the website and that I’m on the right track.

If You Liked It Then Bookmark It

Hide Sites

How to promote your business through classifieds

Daniel Buca — Thu, 07 Jan 2010 10:19:05 +0000

Your business, be it online or offline, may have many needs but one of the most important needs is promotion.
You need to let the people hear about your business, know what you do and why you are better than others.

There are many ways for promoting your business online and one of them is through classifieds.

You can post classifieds very easy and mostly free to many websites.
The are general classifieds websites, there are niche or local classifieds. There are also niche forums that have a classifieds section where you can post your ads.

Step 1: Writing the classified ad
Your classified ad should be simple to read and understand and short enough not to be boring and long enough to offer readers the needed information about you.
You should not focus on general information about your business, you should better focus on products or services you offer, even if you need to write more than 1 ad.

Offer good contact information and identify yourself.

Also, do not forget to add images and video, if the website allows, if not at least post links to images or video.

Step 2: Where to post it
This depends very much on what is your business about.
You should make a list of websites that you update at least every 3 months.
The list should have:

general classifieds websites (world wide and regional / local)
niche classifieds websites (world wide and regional / local)
forums that have classifieds sections

Search Google, ask your friends, ask on Twitter and build your list. Then post your classifieds on those websites, it doesn’t matter if the website is big or small, what matters is that some of those visitors can become your clients.

Step 3: Measuring results
Always include a link in your classifieds.
The link should go directly to your website, on a page related to what you wrote in the classified (ensure, in this way, the relevancy of the link).

You can use, then, Google Analytics to track referrals for your website, you can even group the websites from the list and, in this way, see the results of the whole classifieds campaign.

—

Promoting your business through classifieds can only bring you good results, either big or small, but can never harm your business.

If you start promoting your business through classifieds don’t forget about www.RunClassifieds.com where you can post your classified real fast (no account needed, just confirm your email), easy and free. You can even define a nice and unlimited images gallery for your ad.

If You Liked It Then Bookmark It

Hide Sites

Thoughts for 2010

Daniel Buca — Thu, 31 Dec 2009 19:34:18 +0000

2010. A new year, a new challenge.

For 2010 I want more discipline in everything I do, I want to get involved in less projects but I want to do more for each one.
I will try to give more to the people around me, family or people I work with, and I will expect more from them.
I will not expect people to do the right thing; I’ll just get sure they do.
As for online projects: I have a few surprises for this year, stay close, you’’ like it.

I wish you a happy new year, full of good things and I wish you to fulfill your wildest dreams.

Random Posts

If You Liked It Then Bookmark It

Hide Sites

RunClassifieds.com is live

Daniel Buca — Mon, 28 Dec 2009 22:11:15 +0000

I’ve been working lately on a new project. Some might say it’s the wrong project to work on , some might say it’s worth to try.
I’m talking about RunClassifieds.com , a new classified ads website that’s live starting today.

Let me tell you a few words about this project:

you can post your classified ad free, with no registration, in 3 easy steps: define ad, add images, confirm email
you can add in the image gallery as many images as you want
you can choose from a wide variety of categories and also you can choose the country, state/region and city for your ad
you can propose us to review you ad in order to promote it, all for free (reviewing and promoting your ad)
very simple and intuitive navigation
fast loading

Where do I want to go with this project? I’m going to explore the social classifieds market and this project will be developed in this direction. More details on the next release

If You Liked It Then Bookmark It

Hide Sites

Value of experiments in online commerce

Daniel Buca — Tue, 22 Dec 2009 16:26:55 +0000

I have written several times about the need to adapt online commerce and online commerce solutions to the users’ needs. There is still a question: how do we realize what the online stores customers need?

Knowing that user experience can greatly influence purchasing decisions I think we should not ignore this.

This article tries to answer this question: how do we realize what the online stores customers need?

What should be improved?

User experience starts when a visitor reaches a page on your online store. This page is not always front page of the site. The first aspect to be improved is this page, which starts a user’s visit. Remember: this page is not necessarily the index page of your online store.

We are in the moment when a new visitor entered the online store. He now needs to go as fast and as easy as possible for products that interested him, those that are relevant to him. We must improve, then, the path from first page of his visit to the page he wants to see.

When the user has reached the product or products he’s interested in there are still two aspects that need to be improved:

the path from initiating an order to ending the order
the registration process of a new user

Generally, these are the directions where we need to look when we want to improve the user experience on an online store. How we do this is, however, another story.

How to find out what are the users desires?
There are several ways in which we can find out what they want.
We can do feedback forms, we can do surveys, we can do a contest to stimulate feedback but we can also do experiments with the online store.

I think that any way of collecting feedback we use, if we let the user decide what information he will provide, we can never be sure that we have real and relevant information available. Why do I believe this? Primarily because when a user sends us his feedback, being aware of that, he will provide us with subjective information. Second: this sort of collecting feedback generates reactions from extreme situations, the very happy and very unhappy people, but we are interested in an objective opinion of the majority of an average user.

We remain, however, with the experiments that we can run a tool that can help us find out what they want and we can be almost sure that the user provided objective and relevant information.

Note: a subjective opinion of a direct feedback from the user means that the information provided is different from the actual experience of that user.

What is an experiment with an online store?
An experiment on an online store is not an easy process but I think the results worth it.
A simplistic description of an experiment should be: modify sections of the store and study what happens, how users react and how the KPI (Key Performance Indicators) changes.

What are the steps necessary to make an experiment:

identify the area where we want to do the experiment: it is better to choose one area at a time in which to do an experiment
define what will be affected by the experiment: if we want to play with the product listing page and run some experiments on the product filters, change how it works and how it looks or their position then we write that down, change it and then look for the results.
define a time and users interval when to run the experiment: the experiment can run several hours, several minutes or even several days, well, we can define and if it will run for all users or only for part of them.

Now you only have to implement your experiment and to measure the results.

If You Liked It Then Bookmark It

Hide Sites

Database sharding – when and how we implement

Daniel Buca — Thu, 15 Oct 2009 14:47:59 +0000

I started building an application for the education system and it will have to use huge amounts of data.
Some of the issues I have to address are size of the database, real-time reporting and speed of access to the database. Considering this aspects I started looking into database sharding so here are my first conclusions, a short presentation on what database sharding can do for you.

Solutions like should be planned before project starts in order to have an effortless scaling in time.

What is database sharding?
A decomposition of a database into multiple smaller units that can handle requests individually; A breaking into shards; A shard is a method of horizontal partitioning in a database or search engine.

What is database sharding beyond these definitions and how can it help us in real life when building online applications

When we plan an online application we have to think about it’s potential to grow. The growth translates, in general, into more requests to the application and huge resource consumption. From a certain point the resource consumption evolves exponentially, not linearly.
Even if today hardware is cheap and prices are going down in time there comes a time when investing in new, more powerful, hardware doesn’t bring too much performance and the cost per percent of performance increased is growing exponentially.

Where do problems appear more often?

web server: problems with accessing files or too many concurrent requests; you can deal with this kind of problems by using a CDN for file storage and by using balancers for web servers; there are a lot of solutions for this issue, both hardware and software, both cheap and expensive
database server: problems arise both at writing to database and reading from it; at a certain point even a replication model (few write servers and more read servers in replication, or whatever model suits you best) can be inefficient when problems are generated by huge tables and indexes.

How can database sharding help us in this case?

When using sharding as a solution for database horizontal partitioning we have to think about logical distribution of the database on multiple servers.
The structure of the database will be the same as for a normal architecture in most cases but on each server we will have smaller databases with fewer records.

Now let’s give an example to translate all that I wrote until now

Let’s consider we have an application that manages users and the messages exchanged between those users.

At start let’s consider we will only need 2 database servers.
The structure for the tables in the 2 databases will be the same, the only difference is that on server Alpha we will put the users that have and id that is even and on server Beta we will put the users that have an odd id.
On the Alpha server we will store also all the messages related to the Alpha users, both sent and received and on Beta we will do the same for Beta users.
If we do it this way we will double the number of messages in the entire system but we gain performance when doing readings from the database.

This is a simple solution for a simple problem. For more complex problems there are other solutions and I will write about those in the following lines.

One more thing about this example: how do we work with this database setup from the point of view of the developer?
Simple: all the requests to the database layer will have to contain the user id too and the database layer will decide, based on the user id, to which server it will send the request.

Moving on: types of database sharding implementations

There are many ways to implement database sharding for the application we want to develop.
Each one has its own advantages and disadvantages and it’s our mission to decide which one fits our project best.
Some ways of database sharding implementations can do us a lot of good but some can destroy our application so always plan smart.

Vertical partitioning
We split the database into logical entities that can live independently from others. Some entities will go on shard while others on different shards. We can also have a dedicated shard for one entity.
Example: we have one shard for user profiles, one shard for messages and one shard for image gallery.
In this way, when we want to show images we will send requests to the image shard. All we have to do is to maintain in all shards a reference to a central entity, in this case the user.
This set-up works well in certain cases and it’s easy to implement but can’t handle individual growth of entities.
Interval partitioning
This set-up is best suited for financial applications and means that we choose the shard we write to/read from by checking in which interval a certain value fits.
Example: all financial data for year 2008 is saved in shardX and for year 2009 in shardY.
ID based partitioning
In this case we copy the same database structure on all shards and we define the main entity of the application, for example the user.
Considering we have n shards available we will have the following rule of writing / reading: for user id 1 we use shard 1, for user id 2 we use shard 2, for user id n we use shard n, for user id n+1 we use again shard 1 and so on.
Such a solution is a good option when we can estimate the growth of the database.
The good: all shards will be evenly loaded.
The bad: is hard to increase the number of shards used and re-balance the shards.
Directory based partitioning
Think of a system where for each database request we also send to the database layer the entity identifier and the id (for example: entity – users, id – 567 or entity – images, id – 23).
We also build a directory that relates entity + id to shard.
Choosing such a system is the best solution when not knowing how your application will grow and when. It’s also the hardest to implement because for all database operations you have to update information’s in the directory.

When not to use database sharding

There are still a lot of situations when we are not forced to use database sharding, there a lot of other solutions we can use until we really need database sharding.
We can use sql cache, sql proxy, storing entire record sets with memory with memcache. Or we can use better hardware.

Such an example comes from the 37signals team, read here.

Before the end

Here is a nice presentation from NETLOG about database sharding, enjoy!

Database Sharding At Netlog

View more documents from Jurriaan Persyn.

If You Liked It Then Bookmark It

Hide Sites

Daniel Buca

A book for a better online commerce

Related Posts

What we put on the first page of an online store?

Home page of an online store

Wich solution is correct?

Related Posts

RunClassifieds.com has a new owner

Related Posts

News from RunClassifieds.com

Related Posts

RunClassifieds.com reviewed by KillerStartups.com

Related Posts

How to promote your business through classifieds

Related Posts

Thoughts for 2010

Random Posts

RunClassifieds.com is live

Related Posts

Value of experiments in online commerce

Related Posts

Database sharding – when and how we implement

Related Posts