Both spend analysis and general-purpose BI tools derive their power from the ability to query large datasets quickly and efficiently. These queries aren't keyed in by an expert, as would be the case with a conventional database system; instead they are deduced, built, and executed automatically by the tool, as the result of simple user actions such as clicking on an item. This ease of use empowers business users to produce valuable insights from data without the assistance of costly and difficult-to-schedule IT and database experts.

For the purposes of this discussion, let's consider a simplified spend dataset -- basic payments from an accounting or ERP system. Each row of data contains a date, an amount, the vendor (supplier), the cost center or business unit, a GL code, and other related fields. Every one of potentially millions of rows of data contains the same columns.

Analyzing the Data without Mapping

Loading a dataset of this type into a BI tool lets us answer useful questions immediately. We can fully explore the relationship between Vendor and GL, Vendor and Cost Center, and so on. We can build hierarchical views of our Cost Center structure, so that we can see how Vendor spending affects each level of the hierarchy.

But (and this is a key ‘but’) we can neither alter the structure of the data, nor infer new information from the existing columns. Any such work has to be done offline, with other tools — historically spreadsheets, which are not only difficult to work with, but can only handle a subset of the millions of rows we're facing.

Analyzing the Data with Mapping

Spend analysis tools add the power of mapping to the analysis toolbox. With mapping, users can create additional columns in the existing dataset, and populate those columns with information that is inferred from the contents of other columns. For example, we may know that a payment to a particular Vendor, when booked to a particular GL code, always means an IT Contract Labor expense. Neither the Vendor by itself, nor the GL by itself, supplies that information — it is, in effect, new information, a "Commodity" column.

Mapping also helps us clean up the data. For example, we may know that "I.B.M." and "IBM" are the same Vendor. The BI system doesn't know that, though, because those names are different and computers are very literal. However, armed with mapping, we can create a new column in the data that we'll call "Vendor Parent." We can enter into this new Vendor Parent column the name "International Business Machines, Inc." for any row of data with vendor "I.B.M." or "IBM" or other permutation of the name. Now we can aggregate spending with IBM and drill that aggregated spending by Cost Center and by our new Commodity column. Suddenly there are new insights to be found everywhere.

Summarizing the Difference

A BI system can answer many questions about the data. For example, it will tell us who billed us, for how much, where the spending was booked on the ledger, and who bought it. But we don't know "what" was bought, which is key to managing demand and prioritizing sourcing initiatives. With the data mapping facilities of a spend analysis system, we can deduce the commodity purchased, allowing us to launch sourcing initiatives such as reverse auctions. We can take steps to reduce spending by controlling the number of vendors we're using for the same commodity, allowing us to negotiate better terms. And, for the first time, we can characterize the demand for commodities across the business, and set long-term goals for controlling and monitoring that demand.

The insights coming from just a casual mapping of spend data can pay immediate dividends as stakeholders begin to realize the extent to which company spending can be optimized. Finding spending that should be on contract, minimizing bypass spending with unapproved vendors, and starting "what can you do for me" conversations with newly-aggregated vendors whose significance had been previously unappreciated can pay huge dividends in return for not much effort. In fact, we may find that we've actually hit higher discount levels with vendors than was thought previously, since we've now grouped together all the disparate names that have crept into the ERP system and aggregated the spending. It is not unusual to discover that one or more vendors owes an immediate credit.