Blyth Incorporated, a $1.6 billion manufacturer of candles, fragrance and home decor products, has a policy of growing through acquisition and then permitting the businesses it has acquired to operate entrepreneurially. Successful though this has been as a way of increasing the top line of the business, the policy has tended to leave opportunities to exert purchasing leverage untapped, says Linda Peyton, director of procurement technology in the company’s Chicago-based global sourcing unit.
No longer. Recognising this weakness, in 2002, Blyth implemented a spend analysis application from Tigris, a company since acquired by Verticalnet. Cautious about spending investment dollars without knowing the resulting return, Peyton devised a strategy for measuring the tool’s effectiveness. Across a sample of $20 million of raw material spend, reductions in purchase price of the order of 10-15 per cent were consistently achieved, she says. The company’s annual report to shareholders even paid tribute to the huge return on the investment made.
So when Procuri – producers of another software application used within Blyth – announced a tie-in with analytics vendor SAS, resulting in a spend analysis tool that was not only more powerful, but needed less manual intervention, gaining the requisite approval was not an obstacle. “The functionality of the SAS-Procuri tool was far more advanced, and it automated the process far more than the Tigris tool did,” says Peyton.
At the moment it is too early for another testimonial in Blyth’s annual report – the new tool was only implemented last December – but Peyton has no doubt that one is richly deserved. “Spend analysis empowers our buyers, giving them the information they need to do the best possible job,” she says. “They can no longer say: ‘I don’t have the time or resources’ – all the information is there.
Yet such glowing reports are all too rare. Despite the fact that spend analysis is pivotal to effective supply management, analysts at AMR Research characterise spend analysis in practice as “plagued with functional silos, ad-hoc management, weak technology support and poor source data quality”. More than a third of companies have no visibility into over half of their spend, reckons Lora Cecere, a research director at the firm.
It is a damning indictment, yet one all the more remarkable because, in theory, spend analysis should be a low-hanging fruit long since plucked from the bough. Its premise, after all, is straightforward: figure out what you buy, and from whom, then look for opportunities to exert leverage through either grouping similar items together to form a bigger purchase, or combining purchases across a group of vendors into a single large order from one of them. Or, indeed, through both options at the same time.
But, in practice, achieving this has been difficult. While some major consultancy firms, such as Bain and AT Kearney, were carrying out spend analysis assignments for clients more than 20 years ago, the problem has been moving beyond largely manual (and expensive) one-off exercises staffed by consultants armed with spreadsheets, and automating the process enough to make it both affordable and repeatable at will.
The chief culprit: data. With true irony, the very businesses that will benefit most from spend analysis – disparate, multi-division, multi-company organisations, especially those that have grown by acquisition, such as Blyth – are those where the data available for spend analysis least lends itself to the exercise.
There are several difficulties. Even getting all the data together in the first place can be a challenge for some businesses, says Paul Noel, senior manager in Ariba’s Visibility Solutions division, under which the company’s spend analysis offerings fall. And even when the data is together, he notes, cleaning it to make it usable can prove a gargantuan task, especially when data is drawn from systems in countries with different languages. For example, although relatively straightforward to the human eye, the rules required to convert an error such as “Wodget 101” to “Widget 101” can prove complex to delineate. Language adds another difficulty: again, humans know that “Widget 101 (large)” is probably the same as “Widget (grosse) 101”, but computers do not come blessed with intuition.
Different part number ordering systems – perhaps using product codes, rather than the word “widget” at all – add yet another dimension. Even the most intuitive human might struggle to recognise that component 777389 in one system is the same 9-volt battery as 487919 in another system. It can be done, but only by pulling in additional data, such as component attributes and text descriptors.
Three approaches to the problem have emerged. The first, loosely speaking, comes from companies versed in data analysis and data cleaning. SAS, for example, has years of data analysis expertise. Translating its skills into a procurement context is a departure, but not a radical one – and in any case is eased by the tie-in with Procuri.
Zycus, a data analysis vendor, based in Mumbai, India, approaches the cleaning and classification problem from an artificial intelligence standpoint, taking samples of items and “training” the computer to recognise what they are. Clients include General Motors and General Electric, says its vice-president of marketing, Sandip Maiti.
Kalido is another company parleying its business intelligence and data analysis skills into the procurement spend analysis arena. For Unilever’s Latin America operations, for example, a Kalido-based project analysed spend data from 34 Unilever businesses in 19 countries, “providing visibility on a global scale in terms of key raw materials and key suppliers, providing better information on opportunities to leverage scale, reduce costs and improve collaboration with key vendors,” according to a Unilever executive.
UK banking group HBOS, formed in 2001 by a merger of Halifax and Bank of Scotland, is another example of a Kalido project. HBOS procurement specialist Sharon Reason, who led the project to replace the two banks’ separate spend analysis systems with one Kalido system, cannot disclose the financial impact of the exercise. But she confirms the initial phase of the project – which saw the combined bank’s £1.7 billion annual spend consolidated on to 17,000 account codes, ordered into 27 commodities split into 142 categories – was designed to identify purchasing’s contribution to the £300 million of annual cost savings that the merger was designed to deliver.
But as Pierre Mitchell, head of the e-procurement practice at business process benchmarking company the Hackett Group, based in Atlanta, Georgia, points out, at the data-cleaning phase estimating any such savings is an act of faith. The result is a double hurdle that spend analysis projects must overcome to get corporate funding. Not only is the return on investment of the analysis phase itself uncertain – requiring purchasers to estimate in advance any likely savings – but companies must first fund a data-cleaning exercise that offers in itself no intrinsic return on investment at all. “There’s no direct ROI from having clean spending data – the return comes from how you leverage that data,” he says.
Quality data from day one
So a second approach to spend analysis – and one that is admittedly still emerging into the mainstream – seeks to sidestep this quandary altogether, by building spend analysis-quality data into the fabric of the enterprise system. “Master data management”, of the sort advocated by SAP, does generate an intrinsic ROI, yet delivers suitable data as a by-product, albeit not immediately, points out Mitchell.
Essentially aimed at companies with multiple instances of SAP running across the enterprise (in other words, a fair fit with companies likely to benefit most from spend analysis), master data management seeks to build a unifying “master” layer of data right across the business, making it easier to re-use definitions and attributes from one part of the business in another, and also providing the kind of harmonisation required for meaningful analytics.
Even so, many procurement executives will feel uncomfortable with this second approach to spend analysis – not least because it smacks of an IT initiative, rather than one aimed at delivering solid procurement benefits. And compared with the charms of a pure data analysis vendor, a niche procurement vendor at least offers interaction with people who speak the language of purchasing and understand procurement processes.
So the third approach to spend analysis is to engage with a specialist procurement vendor – a number of whom, conveniently, have been making acquisitions in this field over the past 18 months. Ariba, for example, has bolstered the spend analysis capability it acquired with its purchase of FreeMarkets by further acquiring Softface, specifically to gain access to the company’s data cleaning and classification technology. The resulting capability – in essence, the best of the three companies’ individual approaches to spend analysis pre-acquisition, but combined into one offering – has certainly transformed the ability of Pittsburgh, Pennsylvania-based paint manufacturer PPG Industries to undertake spend analyses, reckons Jim Polak, its director of general purchasing.
Hobbled by five ERP systems and no fewer than 23 different data sources – all of which had to be combined into a single instance of data and then categorised – the spend analysis phase typically consumed half of the 18 months or so that it took to run a combined spend analysis and sourcing consolidation event back in 2000, he says.
Even then, the results were often flawed. Polak will not forget the conclusion of a safety consumables sourcing event, for example, when the winning vendor casually remarked that PPG had underestimated its spend with the company by 50 per cent. And an event the following year, focusing on electrical MRO items, began with the premise that there were no more than 30-40 vendors, whereas in fact, says Polak, 307 were subsequently identified.
Although the final outcome, which involved concentrating this spend with just six suppliers, yielded savings of 20 per cent, the protracted timescale rankled. With a faster form of analysis, the savings would have been available more quickly.
Ariba spend analysis has transformed the process, Polak says. The savings have certainly proved in line with expectations formed during the earlier, largely manual, exercises. But the timescale involved in achieving them is radically different, he enthuses. Sourcing teams are now given just 90 days to run an event from start to finish – from beginning the analysis, in other words, to deciding upon the final choice of vendors.
This has been achieved without having made any changes to the IT infrastructure: “We still have the five ERP systems, and we still have the same 23 data sources. But we now have the power to efficiently and effectively analyse the information that we extract from them,” says Polak.
Emptoris is another procurement vendor offering its customers spend analysis. It has done this for some time, notes AMR Research’s Lora Cecere, but has effectively subcontracted the task to Intigma, a company that it acquired in June. Intigma’s expertise lay in cleaning and codifying data using proprietary scientific algorithms, explains Emptoris’s director of product marketing, Kevin Potts. Send Intigma a free text “dump” of a company’s spend, and back it would come, structured and classified.
Emptoris customer Dow Corning has been using the Intigma service since April. The company is already confident that the expected annual returns of 20-30 per cent on its investment in the technology will materialise, says its procurement programme leader, Michael Lanham.
But the Intigma acquisition is noteworthy for another reason: instead of “spend analysis as a one-off project”, Emptoris bundles the technology into its software offering – effectively turning it into “spend analysis as business as usual”. It’s a move prompted by customer demand, reveals Potts: some customers were re-running their Intigma spend analyses with Emptoris as frequently as every two weeks.
And this is likely to be the future direction of spend analysis. Supply chain optimisation vendor i2, for example, which in the late 1990s acquired pioneering spend analysis firm Aspect Development (and, arguably, subsequently allowed the technology to become moribund) is these days promoting spend analysis as part of workflow processes such as new product development. A tie-in to this effect with French product lifecycle management company Dassault Systèmes was announced on 28 June, points out Samir Bhargava, head of i2’s supplier relationship management group.
Enterprise vendors are moving in the same direction, notes James Yearsley, a partner in the supply chain management practice at Deloitte. Implementing a procure-to-pay solution from the likes of Oracle or SAP, he argues, enables companies to run on-demand analyses based not on invoices – traditionally the raw data source for spend analysis – but purchase orders, a “cleaner” data source but also a faster one. All that is required is pre-coding items that are ordered with the appropriate commodity and category codes, something that later versions of enterprise systems such as Oracle are ready-configured to do.
After a long gestation, and difficult birth, spend analysis may at last be moving into the mainstream.
CASE STUDY: MITTAL STEEL
A test of metal
Having decided in 2004 to deploy a spend analysis solution, Mittal Steel, the world’s biggest steel maker, with headquarters in Rotterdam, faced an awkward choice: which solution, precisely? Having mulled over its requirements, says procurement business analyst Caspar Vrensen, Mittal issued a request for quotation in September 2004. Evaluating the responses was illuminating, he reports: with a long history of acquisitions, Mittal was sure that plenty of low-hanging fruit were there to be picked, and was surprised at the lengthy timescales and hefty data cleansing projects proposed by some vendors.
By the end of 2004, a shortlist of three companies had emerged, each of which was issued with a sample of Mittal spend data to analyse and report back on – and given just a week to perform the task. “Even though each company assured us that it was up to the job, we wanted to see how it performed in practice,” says Vrensen.
One clear leader emerged, not so much in terms of the sample exercise itself, he says, but in the thoroughness with which it approached the task, and the quality of the insights contained in the three-hour presentation each contender was invited to give. “It was clear that they’d worked hard, even though they had lost valuable time through a holiday period,” says Vrensen. The vendor in question was Verticalnet.
Mittal accordingly signed a deal with Verticalnet in February, and is currently working hard to prioritise the opportunities thrown up by the analysis, before rolling them out over its global operations. “We’re looking for a 10 per cent reduction in our MRO spend – but even a 1 per cent return would be acceptable, given that our MRO spend comes close to $3 billion,” reveals Vrensen.
Malcolm Wheatley is a freelance business and technology journalist who writes for leading UK and US publications
