The Vanishing Paperwork Trick
Malcolm Wheatley investigates the role of ERP in improving administrative business tasks
History doesn’t record the precise reaction of the unfortunate 1960s customer of the (then) British Gas Board, who opened their quarterly bill to find a demand for £1 million. But that, and blunders like it, did much to popularise a saying that had hitherto been known only to computer engineers: garbage in, garbage out.
Fortunately, technology has removed one source of garbage in. Punched paper-tape, punched cards, and low-grade magnetic tape are now to be found in computer museums, not factories. But other sources of error remain. No matter how good the recording media, if users still type in garbage, then garbage out will be the inevitable outcome. And here, the improvements in technology haven’t been so kind to manufacturers.
In the old days, computer systems were usually little islands of automation. They handled specific tasks, such as shopfloor scheduling, or costing, and the people using them were in the main ‘power users’ who could recognise errors, and work around them. Today’s enterprise systems are all-pervasive, and an error typed in by a temporary clerk in the sales administration department can have huge knock-on effects.
One solution to the problem has been a focus on detecting and correcting errors. “If ERP systems contain errors, users lose faith in them,” says Ed Wrazen, European vice-president at Trillium Software, a firm specialising in data quality and data cleaning issues. “For ERP to work effectively, people have to be able to trust that the data they are using is accurate. And if they can’t rely on the data, then they will continue to rely on paper-based legacy systems that they do trust.”
A Trillium Software solution that allows manufacturers to bring about this level of trust is marketed as a ‘plug in’ to SAP R/3, and has been successfully deployed by Canadian aerospace and rail transportation manufacturer Bombardier. The system is used to ensure the veracity of the product, material data, inventory data, and supplier and customer records held on Bombardier’s two large SAP databases and 70 subsidiary databases underpinning the company’s manufacturing operations at around 50 plants around the world.
The sheer scale is impressive: each week, some nine million new or changed records are reviewed, including the inventory records for over 2.8 million inventory skus held around the world. Not only does the system trap errors, automatically converting mistakes such as wodget 101 to widget 101, but it also intelligently interprets free-form text in up to 180 global languages – including Asian double-byte character sets – and can understand that widget 101 (large) is the same as widget (grosse) 101. “It’s an intelligent tool that we can teach how to interpret our data,” says Dr. Claudio Gruler, Bombardier’s SAP integration manager.
But while such approaches make perfect sense for the aerospace industry, where lead times are long and the cost of errors is high, error-trapping is less effective in fast-moving manufacturing environments where the error may well have reached the customer before it is detected.
Again, technology has a role to play in helping to prevent errors at source. In most factories today, job tickets and operation tickets are read by barcode instead of being typed in by Molly one morning after a hard evening out on the town. The switch to barcoding was probably cost-justified on the basis that is enabled them to dispense with Molly’s services – but the upshot has been a huge improvement in data accuracy.
And in any case, barcodes – or even RFID, widely touted as a replacement – can only go so far. Job tickets and inventory movements are one thing. But no barcode can adequately be used to automate the fact that an employee has changed their address, or wishes to book a holiday. Likewise, from the perspective of the customer or supplier, technologies such as EDI which have tried to impose an error-free structure on transactions are widely regarded as too restrictive for general release outside a few highly structured industries.
Raw automation, intelligently applied, can help. Automotive manufacturer Peugeot-Citroen, for example, turned to European software company ITESOFT in order to process the invoice payments that EDI didn’t cover. A relatively small proportion of the total, 25 per cent of all invoices payable, they still equated to over half a million invoices a year.
Essentially, says ITESOFT’s marketing director Tom Fox, the approach involves automatically scanning paper invoices, conducting optical character recognition (OCR) conversions to turn them into digital documents, and then applying basic ‘reality checks’. Do the numbers add up? Is the date valid? Can an order number be recognised? If the answer to these questions is ‘yes’, then payment is approved. If not, then manual intervention is required.
It’s the originating paper document that is the weak link in the chain, and arguably a better solution is to eliminate the paper. Tools such as ‘employee self-service’ and ‘supplier portals’ and ‘customer portals’ are presently regarded as among the best ways to do this, says Nick Rawls, director of product marketing for enterprise and manufacturing systems at Peoplesoft. What’s more, he points out, the advantages go beyond those popularly supposed. Yes, data entry is cheaper if employees and suppliers type it in themselves. And yes, the people closest to the data – the employees, customers and suppliers directly affected by it – are likely to do a better job of keying it in than Molly.
But especially in fast-moving manufacturing environments, he says, self-service applications can bring about an element of real-time that would otherwise be lacking. “The data goes straight to the core data warehouse, and not through any other intermediate systems or databases,” he says. “Employees’ holiday or sickness notifications can update shift-planning systems immediately. A new customer order, or order change, can trigger email alerts or text messages as appropriate. And supplier advance delivery notifications can spark changes in shopfloor schedules.”
But how reliable are such systems in practice? Acid tests don’t come much caustic than at German chemicals giant BASF, where over the last five years no fewer than 33,000 manufacturing employees have begun to use an employee self- service system. Linked to BASF’s central SAP R/3 system, and accessed through the company’s intranet, the system is credited with improving not just data accuracy, but also achieving substantial cost savings. Some 28,000 employees access the systems at least once a month, in order to update personal records, or make requests.
Developed by BASF’s own semi-autonomous IT arm (which trades in its own right) the system is seen as a showcase, says Dr Wolfgang Littman, manager of BASF IT Services’ human resources unit. “Department by department, several hundred secretaries used to spend much of their time dealing with such things,” he says. “Now, we estimate that only five secretaries play an active part in processing employee data.”
But one of the major benefits of the system, it seems, is also one of the most difficult to quantify. Having one single system, it turns out, forces a single standardised approach – which can pay hefty dividends in terms of process simplification and error prevention, says Ian Gotts, chairman of Nimbus, a process mapping design consultancy and software firm. “Start with a small group of users and a whiteboard, and brainstorm how things should be done,” he urges. “Then build that process into the workflow: it’s classic re-engineering, with consistency and improved efficiency.”
But useful though all these approaches are, observers are pointing to the promise of what could yet prove to be a revolution in how businesses and their ERP systems interact – both internally, and with the wider world.
Termed ‘e-forms’ technology, the idea is to create quasi-intelligent documents with an element of business process workflow built-in, explains Matt Roadknight, a managing consultant with Egham-based business and technology consultancy Conchango. “It’s not just about collecting or creating data, but then doing something with it – passing it onto another party, or to a business application,” he says.
Adobe, better known for its PDF files and Postscript printer language, is one of the leading e-forms vendors. Its Intelligent Document Platform, which works with most ERP platforms but is best integrated with SAP, generates intelligent electronic documents. These can take an ERP output – a purchase order requirement, for example – and send it to a supplier in the form of a PDF file containing not just the requirement, but also embedded data capture and data validation capabilities. The supplier enters information into fields such as the despatch date, and transmits it back as an order acknowledgement. “Enter the wrong information, and the error is flagged in real-time, rather than when the document has been returned to the customer,” says Mark Wheeler, Adobe’s senior enterprise marketing manager.
AstraZeneca, which employs PDF files as a global document standard for its 50,000 worldwide employees, is presently experimenting with the intelligent document format, says Wheeler. But its flexibility can be best seen at a Canadian space hardware factory, COM DEV Space, of Cambridge, Ontario, Canada. COM DEV Space began using Adobe’s intelligent forms not for purchase order or invoices, but to provide a workflow capability used to control the raising of the thousands of new part numbers it generated each year.
The Adobe technology has quite literally transformed what used to be an error-prone and labour-intensive paper-based process. “What used to take hours now takes minutes or even seconds,” he says. The productivity improvement has been substantial. “Calculating the true saving is difficult: it’s not just an improvement in productivity, but also a reduction in errors. The ROI has really been so significant that there’s been no need to precisely quantify it – it’s just a non-brainer,” says design automation manager Perry Edmundson.