Accelerated intelligent enterprise including timely vendor spend analytics

Abstract

To provide an intelligent enterprise system including timely vendor spend analytics and an associated method, which accelerate intelligence of an enterprise using artificial intelligence of analytics and provide transparent and prescriptive insights.SOLUTION: A computation system 100 comprises an input / output interface configured to receive information related to accounting information, a processor interconnected with a memory configured to store the information received from the input / output interface, and configured to analyze the information to provide timely vendor spend analytics, and a display device for displaying the provided timely vendor spend analytics.SELECTED DRAWING: Figure 1

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Patent Application No. 62 / 987,623, entitled ACCELERATED INTELLIGENT ENTERPRISE INCLUDING TIMELY VENDOR SPEND ANALYTICS, filed March 10, 2020, and incorporated by reference as if fully set forth. [Technical Field]

[0002] The present invention is directed to enterprise systems, and more particularly to intelligent enterprise systems that include timely vendor spend analysis. Summary of the Invention

[0003] Disclosed are intelligent enterprise systems and methods that include an input / output (IO) interface configured to receive information related to accounting information, a processor interconnected with a memory configured to store the information received from the IO interface and analyze the information to provide timely vendor spend analysis, and a display device configured to display the provided timely vendor spend analysis.

[0004] The system and method may include receiving information including spending by category, the system and method may include receiving information including spending by vendor, and the system and method may include receiving information including spending over a period of time, such as a selected fiscal year.

[0005] The systems and methods may include received information processed by a processor to find a perfect match using nearest neighbor propensity scoring. The systems and methods may include received information processed by a processor to identify a match using a fuzzy matching algorithm. The systems and methods may include received information processed by a processor to identify a match using at least two fuzzy matching algorithms. The systems and methods may include matching based on a cutoff score. The systems and methods may include parent company information identified for at least one vendor.

[0006] The systems and methods may include performing text tokenization, training an OEM based on the received information, and determining taxonomy categories based on the received information.

[0007] The systems and methods may include utilizing at least one hybrid model that combines support vector machines (SVMs), random forests (RFs), neural networks (NNs), and expert systems (ESs).

[0008] The system and method may include the display device being configured to select vendor actual spend, peer comparison, market considerations, and normative considerations.

[0009] A more detailed understanding may be had from the following description, given by way of example in conjunction with the accompanying drawings, in which like reference symbols indicate similar elements and in which: [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 illustrates an example of a computing system for determining vendor spend analysis.

[0011] [Figure 2]FIG. 2 shows an example screenshot of the system of FIG. 1 conveying vendor spend analysis by providing spend by category.

[0012] [Figure 3] FIG. 3 illustrates an exemplary screenshot of the system of FIG. 1 conveying vendor spend analysis by providing spend by vendor.

[0013] [Figure 4] Figure 4 shows the flow of the vendor rollup corresponding to this enterprise in Figure 1.

[0014] [Figure 5] FIG. 5 shows the flow for searching for OEMs from products in the database of FIG.

[0015] [Figure 6] FIG. 6 shows a screen depiction of the enterprise system of FIG. 1 showing navigation windows for selecting vendor actual spend, peer comparison, market insights, and normative insights.

[0016] [Figure 7] FIG. 7 shows a screen shot of the enterprise system of FIG. 1 showing a category spending landing page as described herein.

[0017] [Figure 8] [Figure 9] 8 and 9 show screen representations of the enterprise system of FIG. 1 showing vendor spend after selecting vendor spend in the window of FIG.

[0018] [Figure 10] FIG. 10 shows a screen shot of the enterprise system of FIG. 1 showing vendor spending for a particular category selected in FIG. 7 or FIGS. 8 and 9.

[0019] [Figure 11] FIG. 11 shows a screen depiction of the enterprise system of FIG. 1 showing additional details provided by selecting a particular vendor in the depiction of FIG.

[0020] [Figure 12] FIG. 12 shows a screen representation of the enterprise system of FIG. 1 showing additional details provided by selecting a vendor identified in the category of FIG.

[0021] [Figure 13] [Figure 14] 13 and 14 show screen depictions of the enterprise system of FIG. 1 showing user selectable toggles that allow depictions to show spending trends, strategic value, vendor risk, preferred vendors and exchange rates.

[0022] [Figure 15] Figure 15 shows a screen capture of the enterprise system of Figure 1 showing a depiction of actual spending for a particular vendor regardless of category.

[0023] [Figure 16] FIG. 16 shows a screen shot of the enterprise system of FIG. 1 showing a view when diving into a specific vendor's actual spend. DETAILED DESCRIPTION OF THE INVENTION

[0024] The present invention is directed to an intelligent enterprise that includes vendor spend analytics. The invention can accelerate enterprise intelligence using analytical artificial intelligence. The enterprise provides timely, transparent, and prescriptive insights. Users can make smarter, faster decisions with less risk by using the enterprise's collaboration and prescriptive intelligence. The enterprise may be used in conjunction with an existing financial and vendor management solution. The enterprise utilizes an agnostic overlay, a SaaS application that interfaces with all financial systems to categorize and visualize vendor spend.

[0025] As referred to below, the term "processor" includes, but is not limited to, a single or multi-core general-purpose processor, a special purpose processor, a conventional processor, a digital signal processor (DSP), multiple microprocessors, one or more microprocessors in association with DSP cores, a controller, a microcontroller, one or more application-specific integrated circuits (ASICs), one or more field-programmable gate array (FPGA) circuits, other types of integrated circuits (ICs), systems-on-chips (SOCs), and / or state machines. As referred to below, the term "computer-readable storage medium" includes, but is not limited to, semiconductor memory devices such as registers, cache memory, read-only memory (ROM), dynamic random access memory (D-RAM), static RAM (S-RAM) or other RAM, magnetic media such as flash memory, hard disks, magneto-optical media, optical media such as CD-ROMs, digital versatile discs (DVDs) or Blu-ray discs (BD), other volatile or non-volatile memory, or other types of devices for electronic data storage. As referred to hereinafter, the term "memory device" is a device that can be configured to read and / or write data to one or more computer-readable storage media.

[0026] As referred to below, the term "display device" includes, but is not limited to, a monitor or television display, plasma display, liquid crystal display (LCD), or display based on technologies such as front or rear projection, light emitting diodes (LEDs), organic light emitting diodes (OLEDs), or digital light processing (DLP). As referred to below, the term "input device" includes, but is not limited to, a keyboard, mouse, trackball, scanner, touch screen, touchpad, stylus pad, and / or other device that generates electronic signals based on interaction with a human user. Input devices may operate using technologies such as Bluetooth, Universal Serial Bus (USB), PS / 2, or other technologies for data transmission.

[0027] FIG. 1 illustrates an example calculation system 100 for determining a vendor spend analysis. A service provider's vendor spend analysis indicates the service provider's performance in a manner that is easily compared to both ideal performance standards and the performance of other service providers in the same industry or organization. As described in more detail below, the vendor spend analysis is a weighted sum based on performance standards and multiple weighting factors. The vendor spend analysis may be calculated by combining information about spend data from a company's financial systems with business-related metadata. For example, vendor metadata such as risk scores, diversity ratings, and other data fields may be used to contextualize the spend data. These data points may be collected across multiple companies to provide peer comparisons of the collected information.

[0028] The exemplary computing system 100 includes a processor 165, a memory device 170, and a communication bus 180 for communicating data between the components. The computing system 100 optionally includes multiple components, including a graphics subsystem (not shown), a network interface (not shown), an input interface 160, and a printer driver (not shown). The graphics subsystem may provide data to and / or drive the display device 140. The network interface provides a wired or wireless network connection that may connect the computing system 100 to a networked printer or the Internet. The input interface 160 may, for example, be a wireless connection or a data port capable of receiving data from one or more input devices. The printer driver provides data to and / or drives a local printer.

[0029] The computing system 100 includes subsystems for evaluating service provider performance data, determining vendor spend analysis based on the performance data, generating graphical output data for displaying results, and analyzing the vendor spend analysis and related data. The subsystems include a performance data module, an aggregation module, a quantification module, a vendor spend analysis module, a display module, and a results analysis module. The modules may be implemented as software modules, dedicated processor elements, or a combination thereof. Suitable software modules include, by way of example, executable programs, functions, method calls, procedures, routines or subroutines, sets of processor-executable instructions, scripts or macros, objects, or data structures. Some or all of the modules may be implemented as modules or subroutines in a spreadsheet program such as Microsoft Excel®, a database program, or other types of programs, by way of further example. Data used and / or generated by the modules may be stored in the memory device 170. The processor 165 can be configured to operate and control the operation of the modules and to control communications between the modules and other components of the computing system 100, such as printer drivers, graphics subsystems, memory devices 170, network interfaces, and input interfaces 160.

[0030] Generally, accounting system 120 is used to summarize, analyze, and report the financial operations of a company. Accounting system 120 may include ledger-expense-credit-revenue system 128, payables system 126, assets 124, and budget 122, as described below. The accounting system 100 determines vendor spend analysis through information from ledger-expense-credit-revenue system 128, payables system 126, assets 124, and budget 122. Information in accounting system 100 includes data sourced from billing systems, purchase orders, accounts payable, invoice details, general ledger charts of accounts, and business hierarchies. Data from these sources is cross-referenced by the system 100 to organize, categorize, and enrich the data.

[0031] Generally, the ledgers of the ledger-expense-credit-revenue system 128 may comprise a general ledger. The general ledger represents a recordkeeping system for a company's financial data, including recording debit and credit accounts that are confirmed by a trial balance. The general ledger provides a record of each financial transaction that occurs in the operations of a business company. The general ledger holds the accounting information necessary to prepare the company's financial statements, and the transaction data is segregated by type into asset, liability, equity, revenue, and expense accounts.

[0032] Ledger · Expenses · Credits · Revenues An expense in the system 128 is a record in a ledger that represents a payment or an incurrence of a liability in exchange for goods or services. The documented evidence resulting from an expense is a sales receipt or invoice.

[0033] Debits and credits are entered into an account ledger to record changes in value due to business transactions. A debit entry in an account represents a movement of value into that account, and a credit entry represents a movement out of the account. For example, a tenant paying rent to a landlord may debit a rent expense account corresponding to the landlord, and the landlord may credit an accounts receivable account corresponding to the tenant. Every transaction results in both a debit and a credit entry for each party involved, and the total debits and credits for each party for the same transaction equal each other.

[0034] Ledger · Expenses · Credits · Revenue System 128 revenue is the income a business receives from its normal business activities, usually from the sale of goods and services to customers. Revenue is also called sales or turnover.

[0035] Accounts Payable (AP) 126 is a general ledger account that represents a company's obligation to repay short-term debts to creditors or suppliers. AP 126 includes accounting for payments a company owes to suppliers and other creditors. AP 126 is the amount still owed to vendors or suppliers for goods or services received. The sum of all balances owed to vendors is shown on a company's balance sheet as the remaining accounts payable balance.

[0036] An asset 124 is any resource owned by a business. Any tangible or intangible thing that can create value through ownership or control and that is held by an enterprise to create useful economic value is an asset. Simply put, an asset represents the value of ownership that can be converted into cash.

[0037] A budget 122 represents a financial plan that includes both financial and non-financial information. Its most obvious features are estimates of revenue, sales, and expenses for a given organization. A budget can also include non-financial information, such as how many employees a company expects to need. While a budget 122 is a forecast document, companies also use it as a financial control tool. Financial control is a tool for monitoring the activities of their business.

[0038] The vendor spend analysis module may be included in data system 150 and / or accounting system 120. The vendor spend analysis may be determined by determining a weighting factor for each quantitative performance value, multiplying each quantitative performance value by its corresponding weighting factor, and summing the products of the multiplied quantitative performance values ​​and the weighting factors. The weighting factors may be based on user input received via an input interface. The weighting factors may be adjusted to focus the vendor spend analysis on aspects of performance criteria that the user determines are most important and / or to de-emphasize criteria that the user determines are less important. In one example use case, a user may set a performance factor to indicate that a particular performance criterion is extremely important. At a later point in time, the user may adjust the weighting factor to reflect that other performance criteria have become more important.

[0039] The display module may receive data generated by the quantification module and the vendor spend analysis module and generate data for displaying the received data. The display data may be communicated by the display module to a printer driver and a printer, which may create a printed version of the display data. The display data may be communicated by the display module to a graphics subsystem to a display device 140, which may render the display data to a user. Examples of visual depictions of display data generated by the display module and that may be printed by a printer or rendered by display device 140 are described in detail below.

[0040] The display data generated by the display module may have a variety of forms depending on the implementation of the display module and the computing system 100. For example, the generated display may represent the content of an image, one or more pages in a document, or other visual artifact. The generated display data may be compressed or uncompressed image data in a format such as Joint Photographic Experts Group (JPEG), Tagged Image File Format (TIFF), Scalable Vector Graphics (SVG), bitmap, or other format. The generated display data may be data or instructions interpretable by a window manager, desktop manager, graphical user interface, printer driver, graphics driver, word processor, document viewing or editing program, or other software or hardware component related to the implementation of a user interface and / or graphics.

[0041] The results analysis module may receive data generated by the vendor spend analysis module and determine whether parameters used by accounting system 120 and the vendor spend analysis module should be modified. For example, the results analysis module may compare vendor spend analyses generated for multiple vendors. If a majority of the vendors have vendor spend analyses above a threshold, the results analysis module may adjust parameters used in other modules to generate lower vendor spend analyses. Alternatively, if a majority of the vendors have vendor spend analyses below a threshold, the results analysis module may adjust parameters used in other modules to generate higher vendor spend analyses. Parameters that may be modified by the results analysis module include performance value criteria, performance value subcriteria, and weighting factors used by other modules.

[0042] The accounting system 120 may also perform electronic commerce or banking functions based on data received from the quantification module and / or vendor spend analysis module. Data indicating past payments to the service provider may be available to the results analysis module in a memory device and / or via a network interface. Payments to the service provider may be contingent on subsequent vendor spend analysis, performance metrics, and / or performance sub-metrics, and rules relating performance to compensation may be stored in the memory device. The accounting system 120 may determine that a debit or credit should be incurred against the service provider's account based on the compensation rules and data received from the quantification module and / or vendor spend analysis module. The accounting system 120 may then initiate an electronic transaction to generate the determined debit or credit. For example, if the results analysis module determines that the service provider exceeded the required vendor spend analysis or service performance metrics or sub-metrics, the results analysis module may initiate a bonus payment to the service provider. If the results analysis module determines that the service provider was overpaid for work performed, the results analysis module may initiate a debit to the service provider's account. The results analysis module may initiate credits or debits via a network interface using technologies such as Electronic Funds Transfer (EFT) or other suitable services or protocols.

[0043] The accounting system 120 may also analyze data received from the quantification module and / or vendor spend analysis module based on triggers and notify a user when a triggered event occurs. For example, a trigger may be established based on whether a service provider exceeds a required vendor spend analysis, one or more performance metrics, or one or more performance sub-metrics. Alternatively, a trigger may be established based on whether multiple service providers have a required vendor spend analysis, one or more performance metrics, or one or more performance sub-metrics. If a trigger condition exists, the results analysis module may generate a notification email to send to a user indicating that the trigger condition has been met. Alternatively, the accounting system 120 may communicate with a display module to generate display data indicating that the trigger condition has been met. A corresponding notification may then be displayed on the display device 140 for the user. A user may use the notification, for example, as a basis for renegotiating the terms of an SLA with one or more service providers.

[0044] Computing system 100 may be implemented, for example, as a computer, a system-on-chip, or other computing or data processing device or apparatus. Processor 165 is configurable to execute instructions that specify the functions of the modules, such as those described above with reference to FIG. 1. The instructions may be stored in memory device 170 and / or one or more additional computer-readable storage media (not depicted).

[0045] FIG. 2 illustrates an example screenshot 200 of the system 100 of FIG. 1 conveying vendor spend analysis by providing spending by category 210. Screenshot 200 shows that the company spent $149.9 million 220 in fiscal year 2019. This spend data is then broken down into specific spend categories and depicted using display color and size to indicate the direction and magnitude of the amount spent in each category. As shown in screenshot 200, application 230 has the largest spend, corresponding to the largest depicted box and corresponding to $66.9 million spent. Big data 240 is the second largest spend and is depicted using the second largest box representing $27.1 million spent. Calculations 250 is the third largest spend, representing $26.5 million, and is slightly smaller in size than the big data representation because it represents a slightly smaller spend. Additional spending is shown for other categories of spending, including Non-IT 260, Platform 270, End User 280, Distribution 285, Security 290, and IT Management 295. Each box is sized according to the category and corresponding spending amount.

[0046] For example, the overall display may be used to represent total spending, from which each category may be sized according to a percentage representing the share of that category's spending in total spending.

[0047] In addition to the size of the depicted box, the color of the box may also be used to provide additional information, which may include trends in spending for the category.

[0048] FIG. 3 illustrates an example screenshot 300 of the system 100 of FIG. 1 that conveys vendor spend analysis by providing spending by vendor 310. Similar to the depiction in FIG. 2, the screenshot 300 of FIG. 3 again shows that the company spent $149.9 million in fiscal year 2019 320. This spend data is then broken down into specific vendors and depicted using the color and size of the representation to indicate the direction and magnitude of the amount spent with each vendor. As shown in screenshot 300, the first vendor 330 has the largest spend, corresponding to the largest depicted box and corresponding to $36.3 million in spend. The second vendor 340 is the second largest spend, depicted using the second largest box representing $22.6 million in spend. The third vendor 350 is the third largest spend, representing $18.2 million, and is slightly smaller in size than the representation of the second vendor 340 because it represents slightly less spend. Additional spend is also shown for other vendors, including a fourth vendor 360 through vendor x. Each box is sized according to the amount of spending associated with the vendor.

[0049] As an example, the overall display may be used to represent total spend, from which each vendor may be sized according to a percentage representing the vendor's share of total spend.

[0050] In addition to the size of the depicted box, the color of the box may also be used to provide additional information, which may include trends in spending for the vendor.

[0051] FIG. 4 illustrates a vendor rollup flow 400 corresponding to the present system 100 of FIG. 1. In the vendor rollup illustrated in FIG. 4, vendor data is extracted from database 41 of system 100 of FIG. 1. This extracted vendor data enables the visualization of FIG. 3, which shows a user, such as a CIO, CMO, or CFO, how much a company is spending in total across multiple vendors that have been rolled up into a single company through a corporate acquisition, such as a merger or acquisition activity. Vendor rollup may also be referred to as parentage.

[0052] Vendor rollup identifies all manufacturers of purchased items in OEM disclosures and further allows for linking to OEM ownership. Purchasing companies may be notified in an automated, real-time or near real-time manner that some vendors have been acquired by other companies. This feature also allows for notification of potentially redundant, wasteful, and ineffective business negotiations.

[0053] For example, a large financial institution spends $30 million annually on VMware, $5 million on Pivotal, $25 million on EMC, and $20 million on Dell. A purchasing professional in the corporate procurement department may not be aware that Dell acquired VMware, Pivotal, and EMC. Therefore, rather than negotiating each contract individually for a smaller amount, the large financial institution may view the $70 million lump sum spend with Dell as significantly larger than the individual amounts. This larger spend may provide a perspective for negotiating better pricing with each of the commonly owned vendors. Significant savings, potentially amounting to tens of millions of dollars, may be achieved.

[0054] Identifying the actual company requires normalization 420 of supplier and OEM company names. The method of FIG. 4 allows for company names to be imported using abbreviations, synonyms, DBAs, or alternative names (AKAs). A further refinement to the method may be to consider companies simply located in different states or countries as the same company. Names may also be partially or fully stemmed to create normalized names. These normalized names are retained in the database of FIG. 1 and associated with the appropriate company for reuse.

[0055] Additionally, a multi-stage fuzzy matching algorithm 4301, 4302 is used to further normalize the results. The algorithm utilizes statistically tested cutoff scores to provide accurate matching results. Figure 4 shows a detailed diagram of the process flow.

[0056] Specifically, FIG. 4 shows the organizational database of FIG. 1 providing input along with new companies and vendors to search to a first step 420 where a propensity score matching algorithm identifies nearest neighbors to find a perfect match. The output of this test is provided to an algorithm 450 for determining the parent company of the vendor or company. The algorithm also outputs a perfect match to a second algorithm 4301, 4302 using fuzzy matching, as described above in this specification. The output of fuzzy matching 4301 is provided to a second fuzzy matching algorithm 4302, which filters items 440 according to a predetermined cutoff score. These values ​​are then input to algorithm 450 for determining the parent company of the vendor or company. The results are then fed back to organizational database 470, which outputs a final dataset 460 for the vendor-OEM rollup.

[0057] The system intelligently determines and exposes vendor parent-child relationships that arise as a result of mergers and acquisitions. This solution exposes critical relationship data in less time than any existing process or solution because current methods fail to normalize company names. This failure leads to inaccurate associations, even if they are made. The system enhances the parent-child information to include all known past parent organizations and dates of parent-child change, enabling the ability to analyze spend over a specific time period. This is especially important when we consider benchmarks and trends.

[0058] Similar to what was described above for Figure 4, accounts payable, invoice, and purchase order data are used to determine a company's spend with vendors at the OEM and vendor parent level. Artificial intelligence and machine learning are used on the vendor taxonomy to determine the original OEM, regardless of who the goods or services were purchased from or whether the customer's data includes OEM data. The combination of artificial intelligence and machine learning creates "reseller spend visibility." This visibility reveals and visualizes a company's actual spend with original equipment manufacturers ("OEMs"), which is hidden from most systems due to purchases from and through multiple and countless resellers.

[0059] For example, a company may be spending over $2 million per month ($24 million per year) on Cisco products, but this is not apparent in any of the company's financial records because purchase invoices, purchase orders, and invoices show Cisco products purchased through numerous VARs (value-added resellers) and resellers. Larger organizations can negotiate better terms by showing where a vendor is located in the company's vendor list. In this example, if Cisco is one of the top five vendors, for example, they may apply leverage based on this relationship. This information would not be apparent in existing financial systems if an OEM requires an organization to purchase their products through a VAR (value-added reseller) or reseller.

[0060] The system performs vendor aggregation similar to that described above. This occurs not only at the OEM level, but also consolidates "non-OEM" vendor spend. The need for vendor aggregation arises from multiple vendor instances within the source data (inconsistent trading partner vendor tables) and corporate dependencies that exist outside of the source data. Spend amounts may be categorized into standardized technology business management categories to promote transparency in order to show how monies paid to vendors are allocated or used within the organization.

[0061] In the database of Figure 1, a core OEM hierarchy table is constructed. Tables link to themselves to form a parent-child hierarchy. Spend data from customers is ingested and matched with potential machine learning naming that may be inherent in supplier names (vendors of purchase records), OEM names, manufacturing part numbers, model numbers, and customer data. This approach uses enhanced NLP (natural language processing) to process the spend descriptions. The resulting spend descriptions are then passed to a deep learning model, including an inference engine with heuristics that provide speed and scale. Linear support vector machines and random forests are also used to process the final output. The final output 460 is used to determine whether the input OEM name matches an existing OEM name. If the system 100 does not have sufficient certainty of the match, the combination of the purchase data and the input OEM name is passed through a neural network. If there is still not a high enough certainty of the match, the input OEM name is passed to a human curator for analysis. The human curator attempts to match with an existing OEM name or indicates that this is a new OEM. The system creates new OEMs as needed. When a match is indicated, either manually or by a neural network, the input OEM name is added as an alias to an existing OEM, eliminating the need for subsequent human curation. Once spend data is analyzed and matched, spending can be aggregated by parent OEM appropriate to the OEM ownership hierarchy at the time of spend.

[0062] The OEM hierarchy table is kept up to date through a combination of human curation and automated ingestion of organizational hierarchy data from corporate and open ownership sources such as LexisNexis, Dun & Bradstreet, and Moody's. The OEM hierarchy is also tracked by date period as corporate ownership of organizations changes over time.

[0063] Most companies fail to accurately report vendor and OEM spending because accounts payable and purchase order systems are not designed to report at this level of specificity, manual data entry, such as product information, when creating purchase orders is often unstructured, and financial reporting systems are financial reporting-driven. Accounts payable systems do not necessarily capture detailed accounting line items as part of their payment records for product information. Financial reporting systems typically do not include vendor details for expenses that are accounted for and processed through fixed asset systems. Financial system reports may also aggregate data based on general ledger accounts but generally do not maintain sufficient vendor details to provide accurate vendor reporting. The present system 10 intelligently identifies OEMs through a combination of technologies to achieve a higher match rate in a shorter time than any existing process or solution.

[0064] Accurate vendor spend aggregation is a massive task. In large enterprises, vendors are often duplicated in accounts payable and purchase order systems due to multiple remittance addresses, global regional instances of a given vendor, user biases such as inconsistent abbreviations in vendor names, and general inconsistent vendor table maintenance. The system normalizes and streamlines inconsistent vendor names by "name compression." For example, in the accounts payable vendor table in Figure 1, IBM appears in multiple forms. These may include, for example, IBM, IBM, International Business Machine, and International Business Machine SA. Training the system to recognize each name as the same organization is essential for efficient, accurate vendor aggregation.

[0065] When processing a transaction, the first items to resolve are the supplier name and manufacturer name. These may vary in spelling accuracy, pseudonym use, and abbreviations. For example, Dell, Dell Technologies, Inc., and Dell Corp ltd should all refer to Dell. For example, HP and Hewlett Packard are the same, but HPS are different. The system 100 matches input vendor names to existing hierarchies by interpreting and normalizing various variations in company names. The system 100 also simplifies the dataset through a visualization engine.

[0066] FIG. 5 shows a flow 500 for searching for OEMs from products in the database of FIG. 1. Item descriptions are used to train a model to recognize words to classify OEMs. Manufacturers use phrases created to differentiate their products. This phrase is not necessarily captured at the purchase order line item level, but is often captured at the invoice line item level. In topic modeling, the concept of word vectors may be used to search for topics of interest, which is used in this system to search for OEMs from product descriptions. This approach requires a large corpus of data from OEMs to achieve a high degree of certainty.

[0067] The implemented flow 500 takes an incremental approach to learning the OEM: the OEM is validated and the model is updated to improve the next training iteration.

[0068] Flow 500 may further leverage online learning by creating a robust foundation for incremental improvement that takes into account the fact that words describing categories may change over time. Each word is taken as a random variable, and the combination of multiple words is the conditional probability that the OEM is a given OEM. The feature set, which is the words used, may change over time. This feature set is defined by Equation 1:

number

number

[0069] While other organizations use catalogs of products manufactured by OEMs, our system uses a combination of artificial intelligence and machine learning, along with purchase item data, reseller information, and spend classification, to determine OEMs based solely on product descriptions. Our system also differentiates the use of product-level information through a visualization engine.

[0070] Specifically, in step 510, FIG. 5 shows that the data is in PDF format. , CSV (Computer Specific Values ​​V), or other file formats. In step 520, a text description of the invoice PO line item is determined. Random forest methods may be used to learn categories, and variable importance may be used to prune the dictionary of terms used. The system may also convert the text to terms in step 530, and then use term frequency and inverse document frequency to remove N-grams that do not support class separation.

[0071] Tolerance may be performed from the text description in step 540. This may include removing stop words and creating n-grams spanning words, with lengths of one to three words, to preserve order. The enrichment data from step 530 may also be fed into tolerance.

[0072] In step 550, the system may learn which N-grams match categories, from which it may learn OEMs in step 570, and in step 590, it may learn taxonomy categories to identify OEMs from descriptions.

[0073] In addition to providing transparency into "who" a company is spending money on, an equally important data point is providing transparency into "what" a company is purchasing. The use of pre-defined and recognized taxonomies combined with the ability to view data sets via taxonomies allows the system to provide standardized context for spend aggregation. Standardized taxonomies enable analytics such as benchmarking and anonymous peer-to-peer comparisons. Custom taxonomies via customer portals may be configured to allow customers to create additional data views.

[0074] It is basic for businesses to define a set number of levels for their taxonomy, rendered as one row in the database. The system allows for any number of levels to accommodate all trading partners, providing maximum flexibility. The system may be used to encompass both IT and non-IT expenditures.

[0075] A taxonomy may be established first. The model must then learn to automatically classify expense account items into appropriate categories. This automatic classification is achieved using a hybrid model that combines support vector machines (SVMs), random forests (RFs), neural networks (NNs), and expert systems (ESs). Random forests may be used to detect n-grams that correlate with categories. This process reduces the number of variables from 6,000 to between 200 and 500. These variables may then be used to build SVMs and NNs. In SVMs and NNs, the results are scaled to indicate the probability that an item actually belongs to a class. This probability is used to determine which items require manual review for proper class assignment. Using both SVMs and NNs is equivalent to having two experts review an item and reach expert consensus to confirm that it belongs to a particular class. If this method produces inconsistent results, a class is selected for the item, depending on the reliability of the method and the difference in reliability, or the item is sent for human review.

[0076] The expert system, ES, is for items that have good inference rules to determine the correct class label. This is a preliminary process until SVM, RF, and NN have enough capacity to learn categories and perform automatic classification. RF has a high execution time cost, but it is very consistent with the inference method used by experts to determine class labels.

[0077] This approach utilizes both a hybrid approach and one where data is prepared through training to successfully develop a taxonomy. A trade-off is necessary to incorporate inference rules from human experts and expert system rules to feed the results to the engine for training. The initial data volume is modest and requires a means of obtaining good labeled data. Expert systems provide this bridge.

[0078] Figure 6 shows a screen depiction 600 of the enterprise system 100 of Figure 1 showing navigation windows for selecting vendor actual spend 610, peer comparison 620, market insights 630, and normative insights 640. This screen may be used to create and utilize aggregate data from the database of Figure 1. The system launches the actual spend analysis by selecting the actual spend navigation selection button.

[0079] Figure 7 shows a screen depiction 700 of the enterprise system 100 of Figure 1 showing a landing page for the category spend 610 described herein above. It uses the taxonomy described above in Figure 2, including Applications 230, Big Data 240, Compute 250, Platform 270, End User 280, Delivery 285, Security & Compliance 290, IT Management 295, and Non-IT (not shown). A presented IT pattern is provided, in this example, annual spend 710. Each of the variables may be changed to alter the presentation, for example, to use different colors for the trends or different display time frames.

[0080] 8 and 9 show a screen depiction 800 of the enterprise system 100 of FIG. 1 showing vendor spend after selecting Vendor Spend 610 in window 700 of FIG. 7. FIG. 8 depicts screen depiction 800 partially transformed from screen 700 of FIG. 7, while FIG. 9 depicts screen depiction 800 more fully transformed from screen 700 of FIG. 7. Screen depiction 800 again includes Applications 230, Big Data 240, Compute 250, Platform 270, End User 280, Distribution 285, Security & Compliance 290, IT Management 295, and Non-IT (not shown) depicted in FIG. 7 to provide spend data based on vendors within the taxonomy of FIG. 2. Specifically, the size 810 of each vendor box is assigned based on the spend level within the taxonomy. Colors such as red and green may be used to indicate increased spend (red) and decreased spend (green), with shades of each color used to indicate the degree of increase or decrease. Additionally, the identified vendors may describe resellers that may be involved in the chain. As is evident in comparison with the depictions of Figures 8 and 9, the vendor boxes replace the spend values ​​depicted in Figure 7. Each of the vendors that make up part of the taxonomy of Figure 2 is represented by a size and color that provides an underlying meaning to the data, as described herein.

[0081] FIG. 10 shows a screen representation 1000 of the enterprise system 100 of FIG. 1 showing vendor spend for a particular category selected in FIG. 7 or FIG. 8 and FIG. 9. In this representation, a calculation 250 that is part of a taxonomy is shown selected. Screen representation 1000 provides vendor spend for the selected calculation 250. As is evident from a comparison with FIG. 9, for example, this expands the view of the calculation taxonomy 250 and the underlying vendors corresponding to the calculation 250 category. Screen representation 1000 highlights details about the vendor consolidation benefits that can be achieved.

[0082] Figure 11 shows a screen depiction 1100 of the enterprise system 100 of Figure 1 illustrating the selection of additional details and example calculations 250 taxonomies from Figure 10, and further providing details by selecting a particular vendor 1110 in the depiction of Figure 10. For example, selecting VMware 1110 provides additional details regarding the top spenders, identified by name, among companies in this category and the total amount spent in the category. Such information may provide information regarding contacts within the company for a given vendor.

[0083] Figure 12 shows a screen depiction 1200 of the enterprise system 100 of Figure 1 showing additional details provided by selecting a vendor 1110 identified in category 250 of Figure 10. This allows a specific order 1210 to be displayed in category 250 and vendor 1110. Interaction with screen 1220 may also provide the user with additional information 1230 about the specific order, such as order number, date, cost, product, etc.

[0084] 13 and 14 show a screen depiction 1300 of the enterprise system 100 of FIG. 1 , showing user-selectable toggles 1305 that allow the depiction to show spending trends 1310, strategic value 1320, vendor risk 1330, preferred vendors 1340, and exchanges 1350. The toggles 1305 allow other metadata to be selected and provided to the user. The visualization system may provide an output depiction of any other metadata found within the system 100. For example, vendor risk 1330 in the taxonomy of FIG. 2 , which again includes applications 230, big data 240, compute 250, platform 270, end users 280, distribution 285, security and compliance 290, IT management 295, and non-IT (not shown) depicted in FIG. 7 , may be displayed and color-coded by selecting this additional toggle 1305 metadata.

[0085] 15 illustrates a screen depiction 1500 of the enterprise system 100 of FIG. 1 showing actual spend 1520 for a particular vendor 1510 regardless of category. As depicted, each vendor 1510 is identified and presented based on the amount of spend 1520, taking into account underlying affiliates as described herein. The amount of spend 1520 is represented via the size of the box, and the trend direction of the spend 1520 is represented by color coding.

[0086] Figure 16 shows a screen depiction 1600 of the enterprise system 100 of Figure 1 showing a view of actual spend for a particular vendor 1610. This shows dependent spend incurred by the enterprise against the vendor.

[0087] Although features and elements are described above in particular combinations, those skilled in the art will understand that each feature or element can be used alone or in any combination with other features or elements. Additionally, the methods described herein may be implemented in a computer program, software, or firmware embodied in a computer-readable medium for execution by a computer or processor. Examples of computer-readable media include electronic signals (transmitted over wired or wireless connections) and computer-readable storage media. Examples of computer-readable storage media include, but are not limited to, read-only memory (ROM), random-access memory (RAM), registers, cache memory, semiconductor memory devices, magnetic media such as internal hard disks and removable disks, magneto-optical media, and optical media such as CD-ROM disks and digital versatile disks (DVDs). A processor in association with software may be used to implement the systems and methods described herein.

Claims

1. 1. An intelligent enterprise system, comprising: an input / output (IO) interface configured to receive information relating to accounting information; a processor interconnected with the memory configured to store information received from the IO interface and analyze the information to provide timely vendor spend analysis; a display device configured to display the provided timely vendor spend analysis; Equipped with A system characterized by:

2. the received information includes spending by category; 2. The system of claim 1.

3. the received information includes spending by vendor; 2. The system of claim 1.

4. the received information includes expenditures over a period of time; 2. The system of claim 1.

5. the period is a selected fiscal year; 5. The system of claim 4.

6. the received information is processed by the processor to find perfect matches using nearest neighbor propensity scoring; 2. The system of claim 1.

7. the received information is processed by the processor to identify matches using a fuzzy matching algorithm; 2. The system of claim 1.

8. the received information is processed by the processor to identify matches using at least two fuzzy matching algorithms; 2. The system of claim 1.

9. the matching is based on a cutoff score; 2. The system of claim 1.

10. Parent company information is identified for at least one vendor; 2. The system of claim 1.

11. Text tokenization is performed, 2. The system of claim 1.

12. an OEM is learned based on the received information; 2. The system of claim 1.

13. determining a taxonomy category based on the received information; 2. The system of claim 1.

14. At least one hybrid model that combines a support vector machine (SVM), a random forest (RF), a neural network (NN), and an expert system (ES) is used.

14. The system of claim 13.

15. the display device is configured to select vendor actual spend, peer comparison, market insight, and normative insight aspects; 2. The system of claim 1.

16. 1. A method for providing vendor spend analysis, comprising: receiving information regarding accounting information via an input / output (IO) interface; storing the received information via a processor interconnected with a memory; and analyzing said information to provide timely vendor spend analysis; displaying the provided timely vendor spend analysis via a display device. A method characterized by:

17. the received information includes at least one of spending by category and spending by vendor; 17. The method of claim 16.

18. analyzing the information includes at least one of processing by the processor to find perfect matches using nearest neighbor propensity scoring and processing by the processor to identify matches using a fuzzy matching algorithm; 17. The method of claim 16.

19. an OEM is trained based on the received information; and a taxonomy category is determined based on the received information; 17. The method of claim 16.

20. In the analysis, at least one hybrid model combining a support vector machine (SVM), a random forest (RF), a neural network (NN), and an expert system (ES) is utilized.

17. The method of claim 16.

Images