The consumer goods industry has been actively discussing collaboration, RFID and global data synchronisation. Yet, how can the promised efficiency gains be realised if basic data quality is mostly poor. In fact, is the industry not at risk of actually achieving the very opposite as collaborating companies continue to infect each other with poor data.
Elsevier Food International Vol.7, No.3, September 2004 Pascal Kuipers

Global data synchronisation. Sir Terry Leahy, chief executive of Tesco, called it “a boring subject with exciting consequences”. The excitement of a seamless flow of information enabling the industry to drastically lower costs and increase efficiency will, however, be strongly delayed due to poor product data quality.

“Accuracy and consistency of product data are lacking at both retailers and manufacturers. In fact, more than half of the items in company systems contain incorrect data,” reads the survey Internal Data Alignment: Learning From Best Practises, published in May 2004 by Capgemini and the Global Commerce Initiative (GCI) at the ECR Europe Conference and Marketplace in Brussels.
Incorrect data causes inefficiencies, higher costs, lower customer satisfaction and – when passed on in collaborative efforts – these negative impacts of poor quality data will even be magnified as one company infects the other. ECR-inspired collaborative initiatives will fail to deliver wide-scale efficiency gains and the so wanted supply chain transparency through the use of EPC-enabled RFID-tags will remain an unattainable dream.

The problem in perspective
The survey includes an example of the UK-based data-quality company UDEX, which in November and December 2003 analysed all new products – both consumer units (eaches) and traded units (cases) – processed by its UK Quality Assurance service. UDEX found that half of the consumer units were incorrect and one-third of traded units were inaccurate. Nearly three-quarters of suppliers of consumer units had incorrect items and 57 per cent of suppliers of traded units had errors (see sidebar ‘How accurate is the current product data?’ below). This problem is not limited to the UK only. “We have followed this analysis up with other analyses in the USA with similar conclusions,” says Robin Howe, CEO of UDEX.
To put the problem further into perspective: UDEX limited its UK analysis to a subset of only 17 attributes that are part of the GCI-endorsed EAN*UCC global data dictionary (GDD). This GDD, however, currently contains 151 attributes (e.g. product dimensions, weight, number of products in pack, brand, sub-brand, etc.) and this number will increase. The problem is therefore even much bigger than the UDEX analysis indicates.
The number of 151 attributes per product is bound to increase. What is an acceptable level of data (attributes) per product? And does the industry not run the risk of looking at too detailed a level of product data?
“The acceptable number of attributes is the minimum number of attributes needed to deliver the business processes that synchronised data underpins,” comments Nigel Bagley, Unilever’s GCI programme manager. “Today we focus on item maintenance processes, including – typically – new item introduction. As more order-to-cash processes [see table 3 – ed.] are addressed, more attributes will be needed. When we make the leap from efficiency benefits to growth benefits and start using data synchronisation to drive CPFR, collaborative differentiation and good old category management, we will need to increase the scope of attributes.”

Potentially catastrophic results
Insufficient data quality is not a recent problem. In a presentation for retailers held in August 2003, GCI stated that “Both retailers and manufacturers at present deal with significant in-efficiencies and are restricted in further collaboration improvements due to poor, non-standardised product-data management within and between trading partners.”
“Most companies have been aware that they have had a problem with product data for many years but the existing processes – that invariably involved manual intervention along the information chain – allowed them to work-around these failings,” says Bagley.
Also Howe admits that the industry is pulling its own leg. “Problems with data quality are currently masked by manual workarounds,” he says. “Data is distributed across many systems within a retailer. The data is often corrected at the point of detection (e.g. in the Warehouse Management System) rather than in the master data system so there is a lack of alignment within the retailer. This means that the retailer cannot assign responsibility for errors to the supplier – the error may have been introduced internally within the retailer. When these manual workarounds are removed as processes are automated, the errors will be exposed with potentially catastrophic results.”
“This has, of course, highlighted the problem which is felt first by the manufacturers who are responsible for the basic product data,” says Bagley, who is also chairman of GCI’s global data synchronisation implementation group. “Very few manufacturers that I talk to have ignored this issue. The problem is that it is a really big issue and it will take some time for manufacturers to get it right. In Unilever, for example, we have been implementing a Master Reference Data Repository programme for some time but it was very internally focused. We now are re-working a lot of that programme to make it externally focussed. Another example, given by Nestlé at the ECR Europe conference, showed that over 50 per cent of the data records in their systems were either redundant or wrong. Simply put, it is a lot of work to find the problems and then to fix them. Sara Lee estimates 1.5 hours per SKU. Even when that job is done, work is still needed in changing the organisation and information management processes to ensure better maintenance in the future.”

Roots of the problem
Complexity is also rooted in the organisation, where it is not clear who in the end is responsible for data quality. “The problem with internal data alignment within the companies is that most of them do not have a central system for their master data,” says Jörg Pretzel, CEO of the Germany-based data quality company Sinfos. “The person or department who is in charge of master data alignment ‘collects’ the data from various colleagues and departments for further handling. The data is used in various IT Systems. Too many interfaces, too many different systems prevent a high data quality. If they align the data bilaterally with their retailers they also have lack of unified standards for the outgoing data.”
Bagley confirms this with an internal example. “When Unilever launches a new item, the information that collectively makes up the ‘item file’ comes from many different sources,” he says. “The formulation comes from our research departments, the packaging information from our logistics departments, the ‘on-pack’ claim from marketing and, when we get to exchanging price through GDS, the price will come from our sales teams. Of course, this is a simple view, in reality even more functions are involved! We don’t have a single research department, we have research departments for different categories. The same for marketing. Logistics is run, for good reasons, geographically and logistic information will differ between regions and so on…”

Getting and keeping data clean
There is obviously no easy way to solve this problem. According the Capgemini/GCI survey, cleaning up the mess is the first part of the solution. “Internal data alignment is a two step process,” it reads. “First it is about ‘getting data clean’ as an initial one-time activity to clean up the current inaccuracies. Secondly, it is about ‘keeping data clean’, ensuring that the root causes of data inaccuracies are addressed in a permanent manner.”
“Getting data clean is about mapping, checking and aligning data, which is a manual, labour-intensive effort,” the survey continues. “This will take most companies between six and 12 months to accomplish. Keeping data clean is about aligning processes, organisation, standards, IT solutions and trading partner collaboration. This will take most companies at least two years to develop, implement and, most importantly, institutionalise these changes.”
The survey states that “This issue is largely about people and their behaviour.” Dealing with this human factor is time consuming. Internal data alignment affects nearly all business processes throughout the organisation and all business employees are responsible for it. At the same time, centralisation of all data management activities is needed to guarantee, control and maintain a consistent set of accurate data.
A delay of some two to three years at least should therefore be taken into account, before companies can experience the excitement of a seamless flow of information. Howe (UDEX) thinks that pressure from retailers may well speed up this process. “Leading edge retailers will insist on quality assured product data within the next 12 to 18 months, because they will not allow this problem to jeopardise big, data dependent projects such as RFID. Suppliers will be score-carded on data quality as a key dependency on these projects.”
Bagley thinks it will take more time for companies to internally align product data. “Global data synchronisation and the Electronic Product Code [enabling technology for RFID – ed.] have brought this issue to a head. Over the next two to five years all companies will make significant improvement to the quality of their outbound and internally used information.”
Pretzel (Sinfos) says it depends on the size of a company, the quality of the IT structure and the knowledge of EDI. “Some companies can improve the alignment within a few weeks only, some need more than a year,” he says. “The solution internally lies in a central system that all departments being in contact with article master data use as core system. If this system is linked with an external solution for the master data alignment between business partners on the industry, logistic and retailer side, a big step towards smooth flowing ECR processes is made.”

 

---

How accurate is the current product data?

 

What percentage of the items are inaccurate?

	Consumer units	Traded units
% incorrect items	51%	34%
% suppliers with incorrect items	72%	57%

Source: UDEX

From all incorrect items a breakdown was made for the different attribute types where these inaccuracies occurred. Among consumer units, the greatest percentage of inaccuracies were found in the areas of ‘weight/volume actual on pack’ and ‘product range’. Among traded units, inaccuracies were highest with regard to ‘pallet height’ and ‘pallet gross weight’.

Breakdown of inaccuracies by attribute type

Consumer units
Attribute type	%
Product dimensions	11%
Weight/volume actual - on pack	28%
Number in pack – actual	16%
Product range	28%
Sub-brand	17%
Total	100%

Traded units
Attribute type	%
Product dimensions	18%
Product gross weight	6%
Pallet configuration	16%
Pallet height – mm	27%
Pallet gross weight – kg	31%
No. of consumer units in traded unit	2%
Total	100%

Source: UDEX

---

Source: “Internal Data Alignment: Learning From Best Practises” by Capgemini and Global Commerce Initiative, May 2004. A short version can be read in Elsevier Food International’s sister publication Executive Outlook, Vol. 4, Number 2 (June 2004), page 34-45.

---