Most technology approaches to the problem of better business intelligence are generic, but sometimes it is better to specialise. Large organisations struggle with getting decent information in many areas, but one that is a common problem is that of procurement spending. The data that is needed to enable good analysis of spending is typically locked up in a plethora of ERP instances and assorted specialist systems, so even answering a basic question like "how much did we spend with supplier x last year" can be problematic. Certainly there are projects in some companies which try to tackle this issue, but these mostly use general purpose technologies; in many cases the real solution used is Excel. One key issue is that the notion of a "supplier" is a slippery one. A company like "Unilever" has many subsidiary entities e.g. "Hindustan Lever" in India, and so a company buying on a global basis from Unilever may not capture and classify all the spending correctly. The world of products is even messier, and while there are some standards (notably UNSPSC) some of these categories can be counter-intuitive, and few companies diligently assign UNSPSC codes to all their products. A typical purchase order will more likely consist of a textual description of what is being bought, and this may be vaguely worded or ambiguous. For all these reasons, trying to analyse spending on "product category by supplier" is a challenge for most companies. This matters, since a good understanding of spending can help more strategic sourcing approaches, and enable more effective negotiation with suppliers, as well as enabling more effective handling of credit risk.
A relatively new market entrant is trying to tackle this head-on. Rosslyn Analytics is a UK company only founded in 2007, yet has already notched up an impressive set of customer names such as Capita, Rio Tinto, Novartis and Clifford Chance. It offers a web-services application aimed specifically at spend analysis. It has developed extraction technology, including connectors to common systems such as SAP, and a methodology meaning that it can quickly target the likely most relevant data in these systems e.g. purchase orders, invoices and supplier lists. Data is extracted from the client systems and fed to a server (an underling relational database is used) where consolidation of the spend data from various sources is carried out, including enrichment of the original data such as adding Dun & Bradstreet credit information to suppliers where needed. The company uses artificial intelligence to help automate the categorisation of products as far as possible, and then works with the client's in -house procurement team or category managers to supplement this and deal with exceptions. A range of analysis tools are then available to procurement staff.
One recent project involved dealing with 28 different spending data sources from ten subsidiaries of a global private equity firm, a project executed in eight weeks and achieving break-even in just three months. There are certainly other approaches in this area (modules of procurement vendors like Ariba, Emptoris and ERP vendors, some specialist companies like Aravo, CVM and Zycus) but my own experience in some very large companies tells me that there are plenty of companies that do not feel well-provided for when it comes to analysing their spending properly.
One important element of the approach used is to focus on the line item data and work directly with that in order to be able to produce a taxonomy, rather than relying on existing categorisation done within clients, which is often absent or ambiguous. Some other approaches I have seen focus on cleaning up supplier lists and standardising these (a worthy goal in itself) but this can leave the problem of poorly categorised spending. For example, while it is good to know how much a company spent with (say) IBM in a given year, what you really want to know is how much was spent on software, services, disks, servers etc from IBM (and other suppliers) by whom, and how much of this was within contract. It is the emphasis on the line-item categorisation that seems to me the key here.
It is unusual to come across a vendor that has developed a solid customer base within two years, along with multiple quotes from happy named customers, including quantified data on project payback in several cases. It is still early days, but the deep focus on the common problem of procurement spend, using the language of the business people being targeted, seems to be generating some early commercial success for Rosslyn Analytics, and I expect to hear more of this company in the future.