If companies are to take advantage of the opportunities that doing business over the Internet opens up, they will need to deal with massive information flows from many sources, in a way they have not needed in the past. Previously, contact with other businesses was controlled and restricted by the human resources available. For example, the procurement agent controlled the number of offers to be considered by only ringing several likely suppliers, and then settling on one supplier after seeing they had the lowest price or the best delivery. A large amount was understood on both sides, so the telephone call could be restricted to a few details. The Internet opens up access to an overwhelming number of suppliers, most of whom will not understand the buyers needs in detail, and each of whom can send a bid which itself can contain hundreds of possible ways the bid can be structured from its components, with side deals on price, warranty, delivery, length of supply contract and other options.
An automated procurement system has to provide support for a wide spectrum of knowledge about purchases. At one end, little may be known about the object being purchased and the market dictates what is available, at the other end, a production line component is specified in massive detail in an MRP system, seemingly leaving little to the discretion of the system doing the procurement.
Buyers and sellers need some way of protecting themselves from information overload, while still allowing both sides to understand each others needs and reach agreement quickly on mutually beneficial transactions. With people largely out of the loop, it becomes important that the automated system can recognize and deal with change. Without this ability, what would have been small glitches in a procurement system managed by people will rapidly cause a system seizure.
Tupai uses its eCognition technology to deliver a business interaction system for procurement. It provides online bid generation, filtering, evaluation and negotiation. The system seamlessly combines analysis with experience obtained through data mining of the organizations databases and XML mining of responses from the electronic marketplace.
Understanding XML Messages
The messages between businesses are rapidly shifting from telephone and fax to XML, an extension of the HTML used by web browsers. XML allows information about the content to be embedded in the message, so the recipient can immediately make use of the content in other systems. The message structure is precisely specified for some marketplaces, such as electronics and vehicle components, and is dynamically defined within the message in others, but is always in the form of a tree, branching as the information becomes more detailed. The messages can hold dense and complex business information, making it essential that all the possible nuances are unravelled.
Tupais eCognition technology can be used to build business models having a much more flexible and general structure than the XML message, combining various forms of analysis with the organizations experience in the specific area. The resulting business model can then "swallow" any XML message, combining the message with its own model structure and analyzing and dissecting the composite model to make sense of it. Each time an incoming message is swallowed, the business model can potentially grow and change, providing an ideal base for evolution and development of business requirements. Alternatively the business model can produce an XML message by taking a particular snapshot of its current state, including in the message any constraints that are necessary to meet goals in the model, and the model using its business logic to control how much it exposes to particular message recipients.
If the procurement system has limited knowledge about a particular product, cellular phones for example, it will put out a na´ve request for bids on cell phones, then XML mine the possibly hundreds of responses. This way, it can come up with a detailed specification for cellular phones which lists general attributes that every supplier meets the frequency of operation, say - and specific attributes that are available from only a few suppliers an "active flip" or voice activation.
People do this now, by combing through catalogs and talking to supplier salespeople, then filtering the information back to the requestor of the product or service. When people do it, they are forced to filter out many opportunities. With catalogs soon to be in XML form, it will be essential to at least have an automated assistant skilled in searching through reams of XML and capable of "learning" what is required through searching by XML mining. What is learnt in this phase can then be immediately applied in the next phase of procurement, exactly what XML is about.
The product specification needs to be tuned to specific requirements the organization may have. Values are assigned to particular attributes to show limits and desirability - for the cell phones it might be standby time, talk time, call quality, maximum price. Constraints may be added delivery of 50 within 3 weeks, or a relation between standby time and weight. A bid request is prepared and sent out the request itself may still have alternative clusters, either GSM or CDMA, say.
Some means of filtering bids is essential if the request has gone to an open market. Bids may come from all over, from suppliers never heard of before, and be structured in imaginative ways to avoid directly stating that they do not meet the requirements. Without effective filtering, the procurement system would be overrun with possibilities that need to be examined in detail. The system uses the logic in the business model to reject nonconforming bids. The remaining bids are shortlisted in order of desirability on the current parameters and constraints, with nonconforming options removed.
A supplier may have seen an opportunity to put together a structured bid they will upgrade phones for executives to a more expensive model if an order is placed on them for the full quantity. The procurement system needs to recognize variations on the theme of the bid request without rejecting the additional logic in the XML message as superfluous.
The requestor may wish to change the bid parameters and constraints after viewing offers that match or exceed the desired specification.
The system can be used to negotiate with shortlisted suppliers to come up with the best deal, with the requestor either in or out of the loop. If the requestor is out of the loop, the system needs a "negotiating stance", so it knows how far to move on the various parameters before breaking off negotiations and moving on to the next supplier.
Tupaise Cognition technology has the ability to read and analyze complex incoming XML messages, move useful information in the messages into its model, write outgoing XML messages from its model, and negotiate away any differences between the buyer and seller positions. Its ability ranges from removing small glitches in a tightly specified communication, to negotiating away considerable initial mismatch between buyer and seller.
The eCognition technology avoids the need for reprogramming to handle the continuing flux of change. It provides the ideal base for efficient business to business interactions in a dynamic environment.