Was the Larkin Building the first RDBMS ?
The Larkin Soap Company was founded in 1875 by John D Larkin. Like Sears & Roebuck, mail order stores preceded the department store. The railroads made cross country shipment of consumer goods possible as soon as the tracks were laid. In fact Sears did not build stores until late in the game opening its first store in 1925 in Chicago. One could say that E bay and Amazon are redeployments of 19th century business models reworked to operate in the cloud. Larkin faced the same two-fold database management problem as its modern day descendants. On the one hand there is the need to maintain accurate information on customers such as address, orders, payment clearance, shipment, on suppliers such as inventory, purchase price, sales price, restocking, on personnel, on op-ex, cap-ex and so on. Distinct from such operational details is the problem of marketing. From the first it became clear that the assembly line factory was always going to produce more than it could sell. Out of this condition emerged two tactics: Advertising on the one hand, and Regulatory Capture on the other. The beautiful fin de sicle posters exemplify the first. Wondrous lithographs blur the distinction between art and industry (from Wiki Commons):
The sensationalist 'yellow journalism' of the 'muckrakers', exemplify the second. The large meat packers for example favored the complex and cumbersome regulations proposed for their industry as a way of keeping out new firms. Established firms already had the bureaucracy in place to ensure regulatory compliance. While expensive and a nuisance, it was not a deal breaker as it was for startup firms. Regulatory capture has only accelerated in the last century culminating in such political structures as the EU. The EU in fact is best understood as a defense against competition mounted by established business and industry. Its army of lawyers and bureaucrats serve to protect certain firms from having to compete with others, as well as making market entry prohibitively expensive.
At any rate marketing has proved to be a more demanding problem even than accounting from a relational database perspective. Googles vast and complex data systems in the end are in virtue of digital marketing, an endeavor which generates billions in revenue for the company every year.
There is now a whole industry devoted to programmatic advertising in which companies bid for the little advertising windows one sees while surfing the internet. In fact companies hire experts such as TradeDesk to handle the complex details for them. Which brings us to the main point.
Larkin revolutionized the handling of corporate data. Its accountant Darwin D Martin, was a pioneer in the design of the corporate relational database management system. Martin was familiar with the Dewey Decimal System recently created by a fellow New Yorker. Martin reworked Dewey's sheme from a single index system to a multiple index system for multiple sets of data indexing not just customers but products and orders as well. At the same time Martin grasped the marketing side of data, using it to run promotions to target existing customers based on past purchases - much as internet marketers market to those who have bought other products or have searched for them online.
One surmises that Martin must have developed the physical equivalent of the computer based RDBMS with its sets of tables and indexes. He wrote a book on the subject. The first to make a card ledger: Story of the Larkin card indexes – January 1, 1932 which I've searched for online but have been unable to track down. The basic idea of the RDBMS can be grasped from examining this excellent graphic prepared by Julie McMurry for free use on Pixaby.com:
As can be seen each primary data table has a primary key. These primary keys are written into derived tables as foreign keys. In this way data tables are normalized, having no redundant data entries. Darwin D Martin evidently devised a similar system for actual paper files - index cards actually - where the cards were indexed and sorted by their index number. If you want to know what a customer ordered, you first look up his index in the customer file. Then you find that index number in the indexed orders file. That file card in turn gives you the index into the actual orders file. The orders file gives you the actual order file you can pull out to get an itemized list of products ordered. So far so good. The difficulty arises from the marketing side of things: Suppose you have inventory of a product you want to try to sell to your existing customers. How to determine which customers are likely prospects to market to - especially if you have millions of customers ?
Reading between the lines it sounds like Martin's system was not terribly robust when dealing with the marketing aspect of the database problem. In fact even for computer based RDBMS system using the Structured Query Lanquage, SQL, this problem has proved difficult and SQL systems apparently don't scale well as customers reach into the millions. Search engines like Google have had to abandon the conventional RDBMS in preference for the graph database. From wiki commons:
The graph database models the real world in a much more straightforward way. Nodes represent the physical objects of the world, the nouns. The properties of the node represent the adjectives. The arrows represent relations between nodes, the verbs. The relationship properties represent adverbs. As you can see its quite simple to add nodes and retrieve information: Alice is 18. She knows Bob and has since 2001. She is a member of a chess club as is Bob. Google, Amazon, and the big corporations helping companies use their data for marketing have transitioned to graph databases. One can only imagine Larkin might still be in business had such technology been available.
As it was it relied on the conventional RDBMS data model which Martin had devised before the Larkin buildings construction. Martin explained his system to Wright, who proceeded to layout a plan which allowed the building to act as a kind of human powered database engine. If you look carefully at old photos you see walls of file cabinets for index cards of various sizes. As I say I've not been able to track down a copy of Martin's monograph on his card ledger system but I welcome info explaining its workings.
For further reading about the history of the Larkin Company see this excellent article by architectural historian Zeynep Çelik Alexander.
For more info on graph databases check out neo4j.com which has great educational resources as well as a browser based system called the sandbox which you can use to get your CRUD on, i.e. learn how to use neo4j's amazing CYPHER SQL-like language to Create, Read, Update, and Delete database nodes and relationships and the key:value properties of each.
The Sketchup Model used for the illustrations can be downloaded from the 3dwarehouse .
