Database Development

Data is the foundation of your website

Most companies believe that they are in the "distribution" business... or the "retail" business... but the successful online companies realized a long time ago that they are in the "data" business also. Key information that is consumed and created by your industry can be useful to your company as it grows. It also can be used as a strategic asset giving you a competitive advantage against your competition.

As we investigate your data model we learn:

How your data is stored
How clean your data is
How distinct or "atomic" your data is
How your data is imported
How your data is exported
Who uses your data and for what benefit

Transactional data needs to be fast and flat (OLTP)

Data that is used for front-facing webpages must be fast and preformatted. Many websites that we consult with have as many as 21 database calls per individual page load. This puts a heavy burden on your database that powers your website. It also makes your website pages load very slowly.

You can make a lot of progress by tuning queries, optimizing query performance and speeding up database queries, but at the end of the day, the more "pre-thought" the data is... the better.

This technique of "flattening" data is generally referred to as "denormalization" or denormalizing the database. This makes your data flat and fast because the values are already pregenerated. As you engage in database development, you need to be aware of which databases should be developed to be fast and transactional.

Data warehouses and data mining applications use this frequently to do rollup reports or cube rollups where individual transactions for the monthly sales are simply added up into a fact table where a single row has the summarized sales for each month. By adding these up beforehand, any subsequent data requests or reports that need these sales totals don't have to be "re-added" each time.

Analytical data needs to be slower and highly "atomic" (OLAP)

Atomic data refers to the concept of data normalization. This means that your database is divided into highly categorized and organized fields and tables. Whether you are standardizing on third normal data, fourth normal data or even fifth normal data, the objective is to have data that is highly accurate, highly specific, and ultra-organized for reuse.

While your database development is going on, you need to be keenly aware of which databases that should be designed to be highly atomic for analytical use.

We have developed a concept called Atomic Data Modeling where we go far beyond a standard fourth normal normalization scheme to something much more precise, reusable and atomic. This atomic data structure allows us to regenerate data in virtually any format for any purpose.

Let's take a real-life example. We have data for Lance Armstrong "Live Strong" white and yellow cycling tank tops for the UCI Pro Tour. This long string of keywords are considered a "molecule" in our atomic data model. Just like H2O represents a water molecule (two hydrogen atoms and one oxygen atom), so too can we deconstruct our "product molecule" into very discrete data elements.

Let's deconstruct this molecule into its atoms:

Lance Armstrong (a player)
Live Strong (a brand)
White (a secondary color class)
Yellow (a primary color class)
Cycling (a sport)
Tank (a type of top)
Top (a type of apparel)
UCI Pro Tour (an entity or organization)

These atoms tend to make sense when combined in that molecular order... but now we've learned something important. If it makes sense to say:

"Lance Armstrong Live Strong White and Yellow Cycling Tank Top for the UCI Pro Tour"

Then it also makes sense to say other variations:

"Tim Tebow Florida Gators Orange and Blue Football Jersey for the University of Florida"

"Fred Taylor Jacksonville Jaguars Teal and Black Football Jersey"

and so on. By deconstructing the atoms in a keyword phrase, we are now able to reconstruct all possible variations of keywords that match the products that we need to sell.... all of it coming from our database.

Used data is accurate data

The more used data is... the more useful it becomes. Take the product catalog data from Ingram Distribution. They had the excel spreadsheet for all of the books in the universe, yet when Jeff Bezos took this data and launched it on Amazon.com, allowing anyone on planet earth to republish this data (through Amazon.com) to have their own bookstore, the central product data became that much more valuable. He then controlled access to new information like:

Which books were more popular
What price were the books selling for
What geographic areas had more views than others
What other products were bought by buyers of a particular book

You can see how the more bookstores were launched, the more valuable the core database become, making Amazon.com one of the most powerful stores on the planet which forced Ingram Distribution to sell off their company to Barnes and Noble in a firesale.

If you do not master the data for your business... you will also go out of business. Someone in your supply chain is going to dominate your industry data... at great cost to you.

Database Developers

Database development is a highly technical art form in its own right and should be done properly with highly trained database developers. We can help you develop your database in the proper form and suited to the precise purpose that you need it to serve. There is no one, single way to develop a database so we need to analyze your business and we can help you identify which database development and which database developers you need to do the job right... the first time.